Which lead sources earn the next dollar?

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: A retained provider Maps record or organic search result. Includes duplicates, directories and excluded candidates. Original cache acquisition volume was not recorded.

Observed code reference: exact-source-inputs.json and retained provider datasets

Pass test: Source candidate admitted with stable identity, attributable company website/contact and source/history predicates. This is not an import-ready result.

Observed code reference: six_lanes.py / six_extensions.py

Pass test: A completed saved outcome, including email failures and other held results.

Observed code reference: six_autonomous.py / existing process()

Pass test: Current strict-valid exact email receipt, complete fresh exclusion/contact history, unique company/email/domain, exact saved activity/commercial quotes, safe greeting and comma-domain. Imported and held rows excluded.

Observed code reference: readiness_six.py independent predicate

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

Pass test: Company was admitted under the shared identity/history rules. Admission is not a ready result.

Observed code reference: six_lanes.py, source attribution in window-final.json

Pass test: Saved processing has a completed outcome. Email holds and other failed outcomes are processed too. Pending is counted separately.

Observed code reference: six_assess.py, window-final.json

Pass test: Independent source/hash, commercial quote, current strict email and complete-history checks all pass at the exact checkpoint. Human review and campaign import remain separate.

Observed code reference: readiness_six.py, assessed frozen code

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

512 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

The 512 raw Maps row occurrences across the 16 executed datasets.

What the script actually does

The audit groups records by placeId and attributes each place to its first current query in saved index order.

Plain-English pass criterion

One occurrence per distinct placeId survives this counting step. This is deduplication, not a website, contact or business-fit check.

Failures and pending route

31 repeated occurrences are counted as repeats, not as failed businesses. Admitted companies are separately counted once using their saved source attribution.

What this count means

481 distinct Maps places. A place identifier is not yet a verified company/contact identity.

Observed code reference: report build_funnels.py:62-74

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

150 places with matching saved text that passes the usable-page heuristic. At family level these are distinct places first attributed to a current query. Saved chart note: Distinct place IDs with a matching saved usable scrape, first attributed to earliest current dataset. No fresh scrape was run for this audit.

Distinct place IDs with a matching saved usable scrape, first attributed to earliest current dataset. No fresh scrape was run for this audit.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

99 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

96 admitted companies in this source. 35 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 96 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

35 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

61 completed processing outcomes of 96 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

45 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

38 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

38 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

1,346 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

297 admitted companies in this source. 264 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 297 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

264 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

33 completed processing outcomes of 297 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

6 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

6 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

6 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

6 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

1,271 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

394 admitted companies in this source. 185 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 394 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

185 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

209 completed processing outcomes of 394 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

21 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

5 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

3 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

3 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

180 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

51 admitted companies in this source. 15 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 51 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

15 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

36 completed processing outcomes of 51 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

13 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

6 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

6 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

6 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

14 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

4 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 4 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

4 completed processing outcomes of 4 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

1 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

1 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

1 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

1 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

543 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

195 admitted companies in this source. 49 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 195 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

49 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

146 completed processing outcomes of 195 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

17 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

6 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

4 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

4 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

2,341 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

877 admitted companies in this source. 571 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 877 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

571 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

306 completed processing outcomes of 877 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

53 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

36 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

32 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

32 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

18 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

7 admitted companies in this source. 1 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 7 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

1 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

6 completed processing outcomes of 7 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

1 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

1 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

1 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

1 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

118 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

52 admitted companies in this source. 52 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 52 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

52 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 52 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

0 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

4 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

2 admitted companies in this source. 2 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 2 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

2 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 2 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

0 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

503 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

127 admitted companies in this source. 127 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 127 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

127 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 127 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

0 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

47 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

25 admitted companies in this source. 25 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 25 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

25 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 25 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

0 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

Historical source collection that predates the additional-leads run.

What the script actually does

The retained records do not establish the original scrape volume. The report deliberately leaves it unknown.

Plain-English pass criterion

This is historical context, not a current pass/fail test. Exact actor settings are shown only where retained input receipts establish them.

Failures and pending route

Do not derive rejection rates from a missing original denominator. Historical inputs without receipts remain unknown.

What this count means

Unknown means not recorded in the inspected evidence. It does not mean zero, and it must not be replaced by the retained-record count. Saved chart note: Historical original scrape count, separate from the current retained candidate records.

Historical original scrape count, separate from the current retained candidate records.

Observed code reference: report build_funnels.py:52-58 and historical provenance mapping

What enters

Previously saved source records from the raw pools assigned to this family or dataset.

What the script actually does

The current run loads the retained pool, preserves source attribution and scans it for unused candidates. The original actor is not rerun.

Plain-English pass criterion

A record is present in the retained input scanned by this run. Admission, email verification and business-fit checks happen later.

Failures and pending route

The next stage removes history conflicts, duplicate identities, unusable domain variables and unproved email-to-company identity. This bar alone does not claim new or qualified leads.

What this count means

338 current retained input records, not original historical scrape volume. The separate historical bar may be unknown. Saved chart note: Candidate records scanned by the current run. Not an original historical scrape count.

Candidate records scanned by the current run. Not an original historical scrape count.

Observed code reference: new2000.py:134-154, 176-183 · scale1200.py:209-222

What enters

Retained records with source identity, domain and email evidence.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Processed candidates in this chart whose exact email has already passed the completed ok check.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

A valid-email candidate with no usable saved text is held from personalization. Do not call it a no-website business without a separate check.

What this count means

0 is the intersection of processed + valid email + usable saved page. It is not the total number of scraped sites in this source. Saved chart note: Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Intersection with all preceding stages. This is a readiness cohort, not chronological order for Maps.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

111 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

30 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

16 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

16 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 16 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

16 completed processing outcomes of 16 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

12 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

12 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

12 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

22 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

3 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

2 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

2 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 2 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

2 completed processing outcomes of 2 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

2 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

2 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

2 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

29 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

3 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

0 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

13 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

2 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

1 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

1 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 1 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

1 completed processing outcomes of 1 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

1 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

1 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

1 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

31 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

7 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

3 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

3 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 3 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

3 completed processing outcomes of 3 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

3 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

3 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

3 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

10 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

0 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

0 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

18 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

4 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

2 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

21 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

3 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

2 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

1 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 1 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

1 completed processing outcomes of 1 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

1 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

1 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

1 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

15 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

5 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

3 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

3 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 3 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

3 completed processing outcomes of 3 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

3 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

3 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

3 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

11 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

3 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

0 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

11 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

5 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

4 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

4 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 4 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

4 completed processing outcomes of 4 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

4 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

4 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

4 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

6 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

3 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

3 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

3 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 3 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

3 completed processing outcomes of 3 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

2 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

2 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

2 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

4 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

1 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

0 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.

Observed code reference: new2000.py:273-285

What enters

Site-proven Maps email candidates.

What the script actually does

The run normalizes company, email, domain and name identities, checks the complete suppression/history snapshot and prior prepared/paid work, requires identity_receipt.clear and a valid domain variable, then appends only unused non-overlapping candidates.

Plain-English pass criterion

The source identity is clear, email belongs to the company, domain is usable, and no prior-use or current company/email/domain conflict is found. These are admission checks, not paid email or AI-fit passes.

Failures and pending route

Conflicting identities stay excluded. Admitted rows that have not completed processing remain pending. Do not count pending rows as failures.

What this count means

0 admitted companies in this source. 0 are still waiting at the frozen snapshot.

Observed code reference: new2000.py:134-154, 176-183, 330-336 · scale1200.py:209-222

What enters

The 0 admitted candidates in this source.

What the script actually does

The eight-worker batch runs email checking, website/model work where applicable, and saving. The report counts a row as processed when a saved processing_status exists and is not pending.

Plain-English pass criterion

A non-pending processing outcome has been saved. This is a progress state, not a success test. Email-held, model-held, scrape-unavailable and worker-held rows can all count as processed.

Failures and pending route

0 admitted candidates have no completed processing state here. Some may have already started or even have an email receipt. They remain pending, not failed.

What this count means

0 completed processing outcomes of 0 admitted candidates. Later bars identify which of these actually passed readiness checks.

Observed code reference: new2000.py:206-231, 248-255, 418-428 · report collect_sources.py:138

What enters

The exact candidate email from each completed processing row in this source.

What the script actually does

The run sends the address to MillionVerifier once or reuses the saved receipt for that exact address. The current wrapper checks email before buying personalization.

Plain-English pass criterion

The saved verification attempt is completed, provider result is ok, there is no provider error, and the address matches the candidate. Catch-all is not treated as ok.

Failures and pending route

Catch-all, invalid, unknown and uncertain outcomes are held. A pending candidate's valid email stays outside this processed cohort until its processing state is complete. The current wrapper reuses a completed verification without an age test. The proposed final validator must restore the 24-hour freshness check.

What this count means

0 valid emails inside this source's processed cohort. Across the run there are 163 valid results, but only 157 are in completed rows. Six belong to pending rows. Saved chart note: Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Completed processed candidates only. Valid emails in still-pending rows are shown separately.

Observed code reference: new2000.py:206-231 · production_pool.py:410-421

What enters

Candidates inside every preceding chart cohort, with a valid email and usable saved company text.

What the script actually does

One paid Luna Max request classifies the business, writes the short Hungarian activity phrase P1, selects the greeting/business noun and supplies exact quotes. Python checks the schema, quote presence, phrase length and name/legal evidence. There is no second model judge.

Plain-English pass criterion

There is a saved generation ID, nonempty P1, an accepted construction_services or other_services category, fit=fit and no EXCLUDED status. P1 must complete ‘Láttam, hogy … foglalkoztok’ and stay within 45 characters. The validator is not a full semantic guarantee.

Failures and pending route

Unsupported business identity/activity, excluded categories, invalid output and empty or capped model output are held. A weak person/legal quote falls back to Szép napot or vállalkozás. Human language and fit review is still needed.

What this count means

0 supported fit/copy rows inside this chart's prior passing cohort. Run-wide, 107 rows have supported personalization, but 16 fail the email/ready gate, leaving 91 in the nested ready cohort.

Observed code reference: scale1200.py:253-267, 345-434 · fresh100.py:95-116 · pilot100.py:377-429

What enters

Supported fit and personalization with completed valid-email and usable-source evidence.

What the script actually does

The final eligibility check reuses the exact email receipt, checks current exclusions and prior-run source identity, stores an eligibility receipt and saves the ready flags.

Plain-English pass criterion

Generation ID and P1 exist, category is accepted, review_ready is true, outreach_status is INSTANTLY_READY, email result is ok, receipt and verification timestamps exist, and the saved email matches. These stored-count checks do not parse a maximum email/identity age. The later source review confirmed that the active wrapper does not enforce the legacy 24-hour rule.

Failures and pending route

A failed final identity/history/email condition is held. A local database-save failure is a separate recovery state and must not trigger a duplicate paid model call.

What this count means

0 stored review-ready rows at the morning cutoff. This counter is not independent proof of refreshed eligibility. Ready does not prove human approval, campaign import, message sending, a reply or a sale. The 2,000-row final acceptance had not been reached.

Observed code reference: new2000.py:196-204, 233-255 · scale1200.py:492-516

What enters

One saved Google Maps query/city actor dataset, or all 16 executed datasets at family level.

What the script actually does

The stored compass/crawler-google-places build 0.14.759 input requests Hungarian results with websites, up to 200 per search. Reviews, images and contact enrichment are off. Retrieval is rejected if it reaches the 1,000-record download bound. The report counts saved rows before local deduplication and exclusion.

Plain-English pass criterion

This is an acquisition count. A returned, saved actor row is counted even when it is a repeated place or later fails a local check.

Failures and pending route

An ambiguous actor start is held rather than blindly retried. Returned records that fail local country, website, closure, identity or history checks do not enter the candidate pool.

What this count means

6 raw place occurrences in this chart. Raw records are not unique leads. The 16 query datasets overlap by 31 occurrences.

Observed code reference: new2000.py:287-329 · report build_funnels.py:69-102

What enters

Maps results remaining after local prefilters, plus any matching run-local scrape cache.

What the script actually does

The crawler reuses a matching saved scrape or tries the own-domain homepage, HTTP/HTTPS/www variants and observed about/contact/service links. It aims for up to three usable pages and eight requests, with a nominal 35-second crawl deadline and a separate fallback of up to 12 seconds. Company identity is checked further during AI classification.

Plain-English pass criterion

At least one saved page has 240 or more characters after trimming leading/trailing whitespace and at least 30 words. It must not match the scraper's listed hosting-suspension, default-hosting, Outlook-login, domain-for-sale, PHP-error or browser-challenge patterns. This is a text heuristic, not proof of company fit or a full website-quality check.

Failures and pending route

Missing usable evidence includes records excluded before a crawl and records whose crawl did not yield usable text. 150 of 481 fresh places have usable saved text, but 331 missing usable results are not evidence of 331 companies without a website. The executed actor required website=withWebsite.

What this count means

1 places with matching saved text that passes the usable-page heuristic. This query can reuse saved evidence from earlier queries. Query raw scopes overlap. Saved chart note: Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Matching saved evidence within this raw dataset. May include cached sites from earlier queries. Per-query raw scopes overlap and must not be added as unique companies.

Observed code reference: run_trial.py:302-323, 341-422 · fresh100.py:133-154

What enters

Usable saved pages from the candidate's own domain or permitted domain variant.

What the script actually does

A regular expression extracts visible email addresses. The script accepts the company domain/subdomain or an allowed free-mail address shown on that page. It selects one address: company-domain first, then info/iroda/kapcsolat/office/ajanlat role mailboxes, then lexical order. It saves the email's source URL and page hash.

Plain-English pass criterion

At least one syntactically valid eligible address is visibly present on a usable same-company page. No guessed address can pass. Deliverability has not yet been verified.

Failures and pending route

With no eligible published address the place does not become a candidate. The current extractor does not retain a verified fallback sequence of every alternative email.

What this count means

1 site-proven email candidates before the final admission/history check. An email found on a page is not yet a valid-email pass.