Hiring Pulseby Agilityy Research

Methodology

Sources

Definitions

Verification: how we know the data is complete and accurate

Nobody eyeballs rows. The collection measures itself at the end of every run and writes a quality record for the day; a monitoring job turns the thresholds below into alerts, and a run that fails them is repaired or re-run before the next report.

Check How Threshold
Fetched-OK rate Boards that answered with a job list, over all boards in the universe alert below 97%
Open-roles stability Total open roles against the previous day, overall and per vendor alert on a move of more than ±15% in a day
Careers-page agreement Every run, 25 boards (a fixed sample for the date) are re-counted on the vendor’s human careers page, HTML not JSON; agreement means within ±5% alert below 90% agreement
Closure confirmation Every run, 10 roles the diff just marked closed are re-fetched at their job URL; confirmed means gone, redirected or no longer naming the role alert below 90% confirmed
Freshness A day without a quality record alert after 2 days
Suspicious empties Boards returning zero roles after having some (previous snapshot kept) reviewed when they cluster on one vendor

A healthy day looks like: fetched OK about 99%, open roles within a few percent of yesterday, careers-page agreement 22 or more of 25, closures confirmed 10 of 10. The first full-universe day, 2026-11-22, measured 99.5% of boards fetched, 21 of 21 verifiable boards in agreement and 10 of 10 closures confirmed.

Known limits

Licence and citation

Reports are released under CC BY 4.0. Cite as “Agilityy Research, Hiring Pulse, week of YYYY-MM-DD” with a link to the report. Corrections and questions: open an issue in the repository.