How ReJob Measures Hiring Activity

Collection and deduplication

The repository includes JobTech stream and snapshot ingestion and source-specific collectors for career and recruitment pages. Collection schedules are configured per source; a configured source is not a guarantee of current coverage.

Where a source supplies an external identifier, records are deduplicated by source and identifier. JobTech ads from organizations with a dedicated collector can be excluded to reduce overlapping coverage. Reposts and cross-source duplicates can still affect counts.

Employers and classifications

The analysis distinguishes direct advertisers, agencies with a named client and advertisements with an unidentified end employer. Some views use organization identifiers or normalized names; the public stats endpoint instead counts distinct non-empty company strings.

ReJob derives structured roles, skill requirements and categories from posting content. Classification and extraction can be incomplete or incorrect. Skill-domain counts exclude language and soft-skill categories and generic skills in the public breakdown.

Public market intelligence tables

The market-report dataset uses one database statement for consistent current counts. Identified direct employers use normalized organization numbers when available. Named agency clients and name-only records resolve to a direct organization only when the normalized name identifies one organization; otherwise names remain separate keys. Unidentified agency clients are excluded from employer counts.

Technology-related advertisements mention a non-generic skill in the selected technical taxonomy categories. This population is broader than IT occupations. Role and city tables on the tech page use that same population. Skill context lists the most frequent roles, workplace cities and co-occurring technical skills; these associations do not imply causation.

Retained discovery history

The history table groups retained current and archived advertisements by UTC record creation date, covering at most thirty previous calendar days plus the incomplete current day. The displayed first observation may be later if retention or coverage is shorter. Advertiser names in this series are not normalized end employers.

This is collection activity, not an active-vacancy time series. Backfills, deletions and source changes affect observations. Archived advertisements do not retain their job-skill relationships, so current skill counts cannot reconstruct historical skill demand. Job-posting activity measures hiring demand and is not the same as official employment statistics.

Location and active status

The public city breakdown uses normalized job-city records with a successful status. It counts city entries rather than guaranteed unique vacancies; a posting with multiple locations can contribute to more than one place.

Public stats include non-archived parsed records. An advertisement can remain non-archived after a stated deadline until processing updates its status. JobTech deletion signals are handled by the collection pipeline; source behaviour varies.

Time, updates and limitations

The public endpoints cache responses for five minutes. The data update timestamp is the latest record update returned by the stats endpoint, rather than proof that every source was refreshed at that moment. Static public snapshots show when they were retrieved.

Seven-day and thirty-day counts use the record creation time in ReJob. They are discovery counts, not necessarily original publication counts. Historical comparisons require stable source coverage and comparable definitions.

Online advertisements omit hiring through networks and other channels. One ad may cover several positions; several ads may cover the same campaign. Coverage, parsing errors and incomplete locations limit comparisons.