Talent Intelligence Tracker

How often does this update?

Collection runs once a day, at 4 PM UTC. The most recent capture was Aug 15, 5:57 PM UTC. Figures on this page are computed on request and cached for five minutes, so a correction or a fresh run appears immediately rather than on a schedule.

What do the evidence labels mean?

Verified means the claim was read from a primary document, usually a regulatory filing or an employer's own statement. Reported means a publication made the claim. Rumoured means the source itself hedged it. A label is capped by the kind of source it came from: a news article cannot become verified however confident it sounds, and nothing is promoted quietly.

Why are layoffs not on this tracker?

They are collected by the AI Layoff Tracker instead, and read from it rather than duplicated here, so there is one source of truth per fact. This tracker is the hiring side: who is adding roles, raising money, changing leadership or moving site.

What does "headcount not stated" mean?

It means the source said nothing about which way headcount is going, which is true of 24,226 of the 26,852 updates in this view. A funding round with no hiring plan and a chief executive succession are both real updates that state no direction. Guessing a direction from them would be our claim rather than the document's.

Why are some updates hidden by default?

3,158 of the 30,010 records we hold are routine officer and director filings: accurate, verified, and in such volume that they bury everything else. The default view sets them aside and the Routine Filings control turns them back on, states both counts, and says what we mean by routine. Nothing is deleted.

What do you know you are missing?

We measure it rather than assert it. Every week we grade the collectors against a fixed set of real events, assembled from public sources without ever looking at our own database. We publish the result, including the countries and document types where we come off badly. The countries scoring zero are the roadmap.

How much of this is automated?

About 99%. Collection, classification, validation, deduplication and publishing all run without a human. Three things stay human: repairing a scraper when a site changes, judging whether a novel source is worth reading, and assembling each new recall test set. The last of those is human by design, because a test set built out of what is easy to find measures memory rather than reach.

Can I reuse the data?

Yes, under CC BY 4.0, citing the Talent Intelligence Tracker. The CSV and JSON links take the current view with its filters applied, and the press page carries the headline figures with a link behind each one. There are 26,852 updates across 15,283 employers and 121 countries in this view right now.