Eleven capability areas, described from the point of view of the person using them. Each has its own page covering what you can do with it, why we built it that way, and what changed recently.
14,246 companies from four data sources, retrievable in a single shape — from the full statement down to a flat metric series.
A quality mark on each individual document instead of one aggregate statistic. Including the difference between a real error and a mere definitional difference.
Two regulators, one data model. Plus an increment feed where late filings and corrections actually arrive.
Every date carries a confidence value from cross-source agreement. Estimated and confirmed are told apart.
An equity ratio of 30 % means something different for a bank than for a software vendor. Hence percentiles.
A semiconductor maker for automotive electronics and one for data centres share an industry code. In the taxonomy they do not.
You do not have to know what kind of identifier you are holding. One call resolves them all.
To verify a figure you need the original. That path is part of the API, not a link into the void.
The same series the scores rest on. No second dataset quietly drifting apart.
A data source that fails silently is more dangerous than one that fails loudly.
A job whose process died used to sit on “running” forever. Not any more.