Score a requisition

What this actually measures

A requisition can be perfectly reasonable and still be unfillable — not because the pay is wrong or the team is unappealing, but because the person it describes barely exists in the place it is looking. This scores that: how many people plausibly meet every hard requirement in the posting, in the market it is hiring in.

Paste the description and it reads out the requirements the way a candidate would — years, level, degree, clearance, the skills that are genuinely gates rather than wish-list items — classifies the role to a standard occupation, and sizes the pool against federal employment data for that occupation and area. Then it narrows: each required skill is priced by how common it actually is across our corpus of live postings, so a stack of five ordinary requirements and a stack with one rare one come out very differently.

What it can and cannot tell you

The pool figure is a model, not a headcount, and it is shown as a range for that reason. It answers "is this stack realistic in this market" well, and "exactly how many people" not at all.

Where the estimate cannot honestly be made, it says so instead of producing a number. A posting with no hard requirements has nothing to narrow by; an occupation and area with no published employment figure has no denominator; a role that is not a technology job gets a plain caveat rather than a fabricated score. A declined estimate is not a small one, and we would rather show you nothing than a zero you would reasonably misread.

It also reads the posting with a language model, which can misread. The report shows the model's own reasoning — which of two occupations a hybrid role was filed under, where an either/or was collapsed, where two qualification paths became one figure — so you can see the judgement calls rather than only their results.

Where the numbers come from

Employment counts and wage benchmarks come from the US Bureau of Labor Statistics. Occupation definitions and their associated technologies come from O*NET. Skill prevalence and competing-opening counts come from our own corpus of live postings. Nothing here is self-reported, and no figure is published without a source.

The full statement — including the floors we publish above and the claims we deliberately refuse to make — is in our methodology. If you want the pay side on its own, that is salaries by occupation and metro.