How to review local-business data quality

Four checks with different meanings

Check	Current evidence	Remaining uncertainty
Business relevance	Category, name and location from the returned record	Adjacent categories and incomplete market coverage
Maps-to-website association	Matching name, address or phone on inspected pages	One sufficient signal can pass; this is not a two-signal guarantee
Person role	Published name, role, quote, source and capture time	Founder/executive does not prove current ownership
Public contact	Contact published on the inspected source	Mailbox deliverability and person control can remain unknown

Maps search versus domain upload

Maps mode checks the returned business against the website. Current rules accept a sufficiently long matching business name, matching address, or matching phone; one can be sufficient. Ambiguous matches become needs_review. Review same-name businesses and shared chain sites manually.

Domain-only mode checks readability, same-domain scope and parked-site conditions. With a supplied business name, it also uses available matching evidence. An unnamed readable domain is not proof that it belongs to a separate company name in your spreadsheet. See domain enrichment.

A cleaner-list review sequence

1. Check category and city before keeping the business.
2. Open the source website; compare address/phone and the branch represented.
3. Review every needs_review record and exclude unresolved mismatches.
4. Read the person's role and supporting text; leave unsupported names blank.
5. Check public mailbox category and source. Use independent delivery verification when needed.
6. Export full CSV with evidence and record excluded reasons.

An unreadable site is not proof that no website exists. No listed website is not proof that the business lacks one anywhere online. A discovered email is not a verified deliverable email.

Deduplication boundaries

Within one campaign, CID takes priority, then Place ID, then normalized name plus address. Separate stores of one company remain separate businesses. There is no global CRM deduplication or automatic company-level root-domain merge. Domain uploads deduplicate within their own batch. See multi-city leads.

What is actually measured?

The public evidence sample explains its source and limited scope. A small website check does not establish production owner precision, Canadian coverage, email hit rate, time saved or customer conversion. No comparable production benchmark is published here. To evaluate providers, use the same dated industry/city sample, report missing fields and keep independent mailbox validation separate.

Count returned businesses, relevant businesses, correct websites, supported person records, public-email rows and person-linked-email rows separately. Calculate cost per usable row from the whole workflow, including optional checks. Credits and billing explains the billable units.