Headcount and revenue for the private companies no index knows

The problem this solves

Every enrichment run ends with a set of rows nobody could fill. They are usually the same kind of company: recently founded, privately held, outside the markets the big databases cover well, or simply too small to have been collected. On a list of four hundred they might be sixty rows, and those sixty are often the ones a growth team is most curious about.

The information frequently exists in public. A company announces a funding round and names its headcount in the second paragraph. An annual filing carries a revenue figure. A careers page lists forty open roles, which tells you something about size on its own. The reason a database missed it is that nobody collected it, and the reason nobody collected it is that it is scattered across pages rather than sitting in a feed.

What has to hold when you fill those gaps is the standard. A figure read off a press release is not the same as a figure from a filing, and a sheet that mixes the two without saying which is which quietly lowers the trust in every other column.

How the mission runs

1. Work only the rows the indexes left empty. The mission looks at list <name> and selects the rows where headcount or revenue is still missing. The company databases have already had their turn and are cheaper when they know an answer, so research runs last and only on what they could not supply.
2. Read the sources a database would not. For each remaining company the agent reads the site itself, any filings that are public, and press coverage around funding or expansion. These are exactly the places a figure hides when it has never been collected into a structured record.
3. Record the value with its source and confidence. Each answer lands with the page behind it, the wording that supports it, and a confidence grade. A revenue figure taken from a filing and one inferred from a press mention are both usable, and the difference between them stays visible in the row.
4. Leave the genuinely unknown blank. Where nothing credible is published, the cell stays empty. This is the instruction that makes the whole column trustworthy: a blank means nobody could find it, so a filled cell means somebody did.
5. Export with the coverage reported. The result goes to Google Sheets with the new values, their sources and the count of rows filled against rows attempted, so the improvement in coverage is a number rather than an impression.

The prompt

This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.

What comes back

The same list with headcount and revenue filled wherever a credible public source exists, each value carrying its source page and confidence, exported to Google Sheets with a before and after coverage count.

Make it yours

• Ask for founding year, ownership structure or parent company in the same pass, since the sources that answer one usually answer the others.
• Restrict it to a single country when local filings are strong there, which tends to lift the fill rate sharply.
• Run it as the last stage of a larger enrichment so it only ever sees the genuine remainder.
• Keep a minimum confidence and discard anything below it when the list feeds a model rather than a person.

Frequently asked questions

Go deeper

• AI agents for waterfall data enrichment: the complete guide
• Data Infrastructure for Enrichment
• CRM hygiene with agents: the maintenance nobody budgets