Public Business Directory Leads

Bounded public-source lead extraction actor for local business prospecting.

Best fit

Use this actor when you already know which public pages you want checked and want reviewable lead rows instead of raw fetch output.

Best-fit operators:

• local agencies validating small lead batches before outreach
• operators normalizing direct business sites plus official public records
• human reviewers who need explicit source diagnostics when a directory blocks automation

What it does

• fetches only the explicit public sourceUrls provided in the run input
• optionally replaces blocked or timed-out primary sources with bounded fallbackSourceUrls
• executes the Proteus public_business_directory_leads_v1 workload
• ranks normalized lead rows into a generic public output shape
• pushes lead rows to the default dataset
• stores the full result envelope in the default key-value store under OUTPUT

Input

• sourceUrls is required and must contain one or more explicit public URLs
• fallbackSourceUrls is optional and should contain direct business sites or official public records you want Proteus to try when a primary source fails with a blocked or timeout condition
• region, businessCategories, maxResults, pageLimit, and idempotencyKey are optional

Quick Start

1. Put the main public pages you want checked in sourceUrls.
2. Put backup official records or direct business sites in fallbackSourceUrls.
3. Keep pageLimit close to the number of URLs you submitted unless you want extra retry room.
4. Start with a small maxResults like 5 while you validate the source mix.

Recommended source mix:

• use sourceUrls for the main business pages or directory pages you want checked first
• use fallbackSourceUrls for official records or direct business sites that can still produce a useful lead if a large directory blocks the run

Recommended first run:

• 1 blocked-prone directory page in sourceUrls
• 1 direct business website in sourceUrls
• 1 official public record in fallbackSourceUrls
• maxResults = 5
• pageLimit = 6

Paid Actor Contract

Stable marketplace-facing input fields:

• sourceUrls
• fallbackSourceUrls
• region
• businessCategories
• maxResults
• pageLimit
• idempotencyKey

Stable marketplace-facing output sections:

• records
• errors
• source_diagnostics
• quality_summary
• execution_plan

Current limits:

• at least 1 primary source URL and at most 25
• at most 25 fallback source URLs
• maxResults between 1 and 25
• pageLimit between 1 and 25

Current non-goals:

• no login-backed scraping
• no CAPTCHA bypass
• no hidden-browser anti-bot evasion
• no enrichment beyond the submitted public URLs and bounded fallback queue
• no guaranteed email extraction

Output

Dataset rows contain public lead fields such as:

• business_name
• business
• category
• city
• state
• phone
• website
• source_url
• fetched_source
• listing_url
• listing_page
• review_summary
• website_weaknesses
• opportunity_score
• opportunity
• verification_status
• verification

Default dataset views:

• Lead Review shows rank, business, contact path, fetched source, verification, and run metadata in one table
• Buyer Review trims the table to the main buyer-facing columns for quick scanning after a run

Dataset-row alias rule:

• raw row fields stay present for compatibility
• buyer-facing aliases are added beside them so dataset exports read more like a public product

The default key-value store OUTPUT record preserves the full adapter envelope, including:

• job_id
• status
• record_count
• error_count
• records
• errors
• source_diagnostics
• quality_summary
• buyer_presentation
• execution_plan

Additional key-value store artifacts:

• RUN_SUMMARY contains the compact run summary plus output_key
• ERRORS is written only when the run returns one or more top-level errors

How to read the run:

• review the dataset for the ranked lead rows
• review OUTPUT when you need the full result envelope
• use buyer_presentation when you want buyer-facing labels and a one-line review headline before reading the raw sections
• review source_diagnostics to see which submitted URL blocked, failed, or produced records
• review quality_summary to decide whether the run is good enough for manual follow-up
• use Buyer Review first when you want the shortest operator scan
• use Lead Review when you need the fetched source, run metadata, or listing path beside the lead row

Common blocked-source pattern:

• a large directory may fail with directory-fetch-blocked
• a direct business site may still return a usable lead row
• an official record in fallbackSourceUrls may preserve a second lead row instead of wasting the remaining page budget

Guardrails

• public web only
• no login-backed scraping
• no CAPTCHA bypass
• results are bounded to the submitted URLs