Multi-Source Business Verification

Brand Verification is an optional website module. It reuses the staged website crawl, returns confirmed and possible brands with evidence URLs, and can remain limited to the product categories selected by the user.

An independent Apify Actor that verifies already-qualified business leads through Google business evidence and optional refinement sources.

Best-in-class verification foundation

The first approved upgrade phase is active:

• A versioned Actor registry separates production sources from controlled pilot candidates.
• Every source record follows a common evidence contract with its Actor, schema versions, evidence URLs, collection date, and expiration date.
• Checkpoints are exact to the Apify tenant, business identity, source, selected product scope, source-data version, and analysis version.
• Legacy checkpoints are reused when still valid and automatically migrated to the new key without repeating the downstream check.
• Pre-run and final estimates separate new checks, reusable checks, downstream Actor expense, Furniture Retail Intelligence fees, and the maximum expected total.
• Shared Supabase tables persist the registry, exact source checkpoints, and cost estimates for multi-user continuation.

Pilot Actors are cataloged but are not promoted into the production path until controlled comparisons pass.

• Google confirmation and Google rating/review count are the default.
• Yelp and BBB are optional source checks.
• It does not use Google Maps for discovery; it only confirms leads already returned by Overture.
• Google product evidence runs first. Only unresolved selected product groups are sent to the dedicated staged Website Evidence Actor.
• Website evidence reuses category checkpoints, tries inexpensive HTTP first, renders a browser only when needed, and can optionally use a residential retry.
• A blocked website is reported as unable to verify, never as proof that the product is absent.
• Facebook remains a fallback when prior evidence does not confirm the requested products.
• Every business/source pair is checkpointed to avoid paying for a fresh check twice.
• Google Business checkpoints are reusable across later product-category selections and analysis-library updates until they expire. Website and Facebook checkpoints remain product-category-specific.
• maximumSourceChecks is the per-run paid-search guard.

Cost behavior

Google and optional review lookups use their configured downstream source. Website use is bounded by selected leads, selected product groups, and one to five targeted pages per website. Residential proxy use is off by default. A transient search timeout is retried once by default; the run hard cap remains the final spending limit for charge-capped downstream Actors.

Creator pricing standard 2026-08-02.1:

Completed new event	Creator fee
Google business check	$0.003 per business
Website refinement check	$0.003 per business
Facebook fallback check	$0.003 per business
Yelp, BBB, Houzz, or Trustpilot check	$0.003 per source/business
Company registry check	$0.003 per business
Product-category analysis	$0.001 per business/category
Brand Verification bundled with another website module	$0.003 per business
Standalone Brand Verification, including its website visit	$0.005 per business

Overture discovery, downloads, failed work, and current reusable checkpoints have no creator fee. Downstream Actor, proxy, and platform usage is separate and is shown before a run. Creator charging requires both the run request and the owner-controlled FRI_CREATOR_BILLING_ENABLED=true environment switch. Keep that switch off until these exact events are configured in Apify pay-per-event monetization.

During the beta phase, creator billing stays off. Each run still publishes creatorChargePreview and projectedCreatorFeeUsd so observed costs can be compared with this standard before public charging is enabled.

Matching policy

The Actor compares the lead name and location with search evidence. Strong name and location agreement is confirmed. Ambiguous evidence is marked for review instead of being silently accepted.

This Actor is designed as a verification stage. It does not decide whether a business fits a selected furniture category.

Gold-standard benchmark gate

No pilot Actor becomes a production source based only on a successful run. The Phase 2 benchmark compares it with the current route using a human-reviewed United States and Canada case set. Promotion requires at least 92% precision, 85% recall, no more than a 5% false-positive rate, 95% evidence completeness, balanced country coverage, no precision regression, and a controlled cost ratio. The template is in benchmarks/gold-standard-template.json; scripts/evaluate-benchmark.mjs produces the promotion report.