Build vs buy: license verification

When you should build

If you verify one profession in one or two states and the boards publish clean open-data files, build it. A Socrata endpoint and a nightly job is a day of work and you will not regret owning it. Buying an API for that is overkill, and anyone who tells you otherwise is selling.

The math changes with breadth, not volume. Cost scales with the number of sources you maintain, and sources are not interchangeable.

What the work actually is

• Not every source is a download. Some boards publish open data. Others release the roster only to credentialed accounts, through a portal that has to be navigated rather than fetched - a login, a report you configure, a file generated on request. Those are integrations that break when the portal changes, not URLs you can put in a cron job, and each one carries an account relationship to keep in good standing.
• Formats do not converge. Pipe-delimited fixed-width, Socrata JSON, Excel workbooks with dates stored as serial numbers, CSVs with genuinely malformed quoting that a document-level parser silently swallows thousands of rows after. Each source is its own column map and status vocabulary.
• Status is not a boolean. Boards distinguish active, expired, suspended, revoked, delinquent, voluntarily inactive, retired-but-practicing and probationary - in their own words. Mapping those into something a caller can branch on is per-source judgment.
• Silent failures are the real cost. A truncated download that parses to 200 rows instead of 200,000 looks like a successful run. Without a sanity floor it wipes your roster, and every lookup afterwards returns a confident not found about real license holders.
• It never finishes. Boards re-publish, rename columns, move URLs and change portals without notice. This is a maintenance commitment, not a project.

What you are buying

22,405,527 records across 31 states and 149 professions, refreshed on each source's own cadence, with the source board, source URL and data_as_of attached to every response so a verification can be filed as evidence rather than trusted.

Plus the parts that are separate projects of their own: a mirror of the federal NPPES registry with license claims cross-checked against board rosters, OIG/LEIE exclusion screening, and Ed25519-signed certificates that a third party can verify without an account.

The honest caveat

Buying does not remove the coverage problem, it moves it. If we do not hold a state you need, you are in the same position as if you had built it - except you can see exactly where the gap is, because we return state_not_covered rather than a guess. Check the coverage matrix against your actual states before deciding either way.