Look up one company
prebuilt/company-to-info uses People Data Labs company enrichment. Its result
has nullable fields and explicit misses; it does not combine several sources.
Inspect its current contract before an authorized lookup, which can use
Deepline credits:
company_found and miss_reason, then compare
company_name, company_domain, and the profile context with the intended
entity. A filled response alone does not establish a match.
Normalize the website without losing evidence
Preserve the submitted value and extract its hostname with a URL parser. Separate paths and tracking parameters:https://example.com/products/widget
can identify the same website as its homepage, but does not prove which legal
entity or sales account owns the product.
Keep meaningful subdomains until your identity policy permits collapsing them.
Use a maintained domain parser for registrable domains; taking the final two
labels fails for some domain suffixes. A consumer mailbox domain is not employer
evidence. For uncertain trial signups, use signup enrichment.
Check the company and profile URL
Compare these observations before accepting the entity:
To find a company profile URL by domain, inspect
company_linkedin_url and its
associated website. If the lookup returns no URL, search for the company name
and domain together, then check the candidate profile’s website field. Preserve
the source URL and connecting evidence. Do not guess a profile slug from the
business name. A snippet or plausible URL cannot settle conflicting candidates;
keep the match unresolved when evidence is insufficient.
Accept fields separately
After accepting the entity, inspect the fields you need, such asheadcount,
industry, hq_country, and total_funding_usd. Null means missing evidence,
not zero. Record the source and observation time and apply freshness and source
precedence per field. The PDL contract
and the selected Play describe their fields; a review policy does not alter
the maintained implementation. Use a custom Play for additional
sources or precedence rules.
An employee measure might be an estimate, a range, or associated profiles.
Keep its definition. Compare entity scope and freshness before reconciling two
sources; averaging parent and subsidiary counts has no useful meaning.
Worked account: accept the profile, hold the segment
A fictional CRM account represents Example Labs atexample.com. A matching
profile reports 51–200 employees. Your routing policy splits accounts at 100:
The midpoint, 125.5, would invent precision. A second result of 1,400 employees
for the parent does not settle the subsidiary’s count. Store identity and field
acceptance separately so the profile can pass while the segment stays in review.
This is an example acceptance policy, not the
company-to-info response schema.
Expand to a file and review field proposals
Preparecompanies-pilot.csv with representative records and a domain or
company_name header. Preserve source IDs and existing CRM account IDs. Inspect
the batch contract and map different headers explicitly:
domain header. Include parent/subsidiary, rebrand,
consumer-domain, and no-match cases. Use the CSV input guide
for file preparation and saved-run results
for inspection and export.
For each proposed field, retain the source ID, CRM account ID, previous value,
candidate value, source, observation time, and acceptance reason. These are
review-file fields you define, not a fixed API schema.
Use Repair CRM records for field ownership,
approval, writes, and read-back. Continue to account scoring
only with accepted evidence.