A missing row needs an explanation
A record can disappear because of an expired credential, a changed source export, an ingest failure, a filter change or an identity correction. None alone proves a company closed. A publishable rate needs pinned snapshots and a reproducible calculation.
Compare like with like
Hold the authority, credential types, status filters and geography constant. Save collection dates and source-file hashes. Match on stable authority-specific identifiers before using business names. Count new, missing and changed records separately, then investigate unmatched records against the official source.
Report the denominator and limitations
State whether the unit is a credential, business or location. Give the period, initial population, number of missing records and number independently confirmed as status changes. Report source outages and uncertain matches separately. Do not extrapolate one renewal cycle to an annual business-closure rate.
Use changes as a review queue
For a vendor roster, recheck the affected credential with its issuer before changing an approval decision. For outreach, apply current suppression and customer exclusions separately from license status. Our state guides identify the relevant verification distinctions and sources.
Use a change log before calculating a rate
Create separate groups for unchanged records, new identifiers, missing identifiers, changed status, and unresolved identity matches. Keep the same source scope and collection period in both snapshots. Note an interrupted export or a change in covered trades before interpreting the result.
Review missing records against the source. A renamed holder with the same credential can be a name change rather than a new business; an absent row after a source outage should remain unresolved. Report the unresolved group alongside confirmed changes instead of quietly discarding it.
Example calculation with an explicit limit
Hypothetical numbers, not ContractorRoster findings: a starting file has 1,000 credential records and 40 identifiers are absent from a later comparable file. The observed disappearance proportion is 40 / 1,000 = 4%. If source checks explain only 10 as expired credentials, the result is not “4% of contractors closed.” There are 10 confirmed credential expirations and 30 unresolved disappearances.
Report the unit, interval and evidence for each category. Do not multiply that snapshot proportion into an annual rate without a justified model, and do not infer company closure from expiration alone.