How Reqbeat measures hiring demand

The sample

We track job postings from public job boards and ATS pages we observe directly. Every result is a real posting, tagged with the board it was detected on. We never infer intent. Results cover ATS tenants plus public job boards and aggregators. Each row's boards field names its source.

LinkedIn is not included. You get postings you can link in an email without sending your prospect to a login wall.

What this does not measure

• Employment levels, hires or separations. This is a posting-volume measure, not comparable to BLS figures.
• Vacancies that are never posted, or filled internally.
• Postings behind a login, a paywall, or an uncrawled application flow.

A company with no ATS or board coverage answers with coverage_status: "no_ats_signal". That means "unknown", not "not hiring". Payloads say so explicitly. Do not score absence of coverage as a negative signal.

Known biases

• The sample skews toward tech and knowledge work.
• Employer size skews the sample: companies on a public ATS are over-represented.
• Geographic coverage varies by country.
• Re-posting behaviour differs by employer and ATS. Deduplication cannot collapse repeats that arrive with a different company name, title or country.

What we publish from the sample, and what we hold back, is on data ethics.

Deduplication

Repeat and re-posted advertisements collapse to one row per company, job title and country. That triple is the partition key. The key is country, not city. Within each partition, the most complete row wins: the row that resolves to a named company with industry, headcount and location. Remaining ties break by the most recent posting date.

Freshness, vintage and revisions

The corpus refreshes every 3 hours across 7 regions. A webhook fires when a new matching posting is detected, typically hours after it goes live.

The research corpus surface these pages read retains a rolling 30-day window on posting date. Any metric that needs a longer observation span than 30 days is pre-registered and is not published until the retention exists to support it.

Pre-registered definitions

Metric definitions are frozen and dated before any figure is computed under them. A threshold can never be tuned after we see a result. A metric marked pre-registered has no published figure, and the reason is stated with it. One example: a "Ghost Job" is a posting open or re-listed under the same deduplication key for at least 90 days. The definition is pre-registered, but no figure is published, because the corpus retains only 30 days.

The same rule covers inputs. Studies that count postings by title match their titles against a fixed term list. Studies that rank demand across sectors use a fixed set of industry groups. Both are published in full and dated before use. A list cannot be widened or narrowed after a result is seen without that change appearing in the method changelog.

Dated measurements

Some figures are readings rather than metrics: one measurement of a corpus that keeps moving, taken on a stated day. Nothing recomputes them. Each one is published with the population it describes, how it was derived, when it was taken and when it expires. Past the expiry date, our build fails until the figure is re-measured or the claim that quotes it is withdrawn.

Changes to the method

Every change to the method is versioned, dated and recorded in a changelog. A change that publishes a figure requires a review quorum. A change that publishes no figure does not.

See also

Data ethics states what we hold and what we will not publish. The crawler page states how we collect, how the crawler identifies itself, and how a site owner can stop it. Short answers to common data questions are on the FAQ.