Run

Every two hours

Method

Fetch the employer's canonical ATS board or permitted first-party structured careers page, never an aggregator

Sources

Supported ATS boards and permitted first-party structured careers pages

On change

Write a diff to the change feed and keep the previous state

On failed refresh

Keep the last verified state and record the source failure — never guess

Publish

The same records to people and to agents, at the same time

Hiring signal is a directional 0–100 measure of observed hiring momentum, not a probability or calibrated forecast. It combines open-role volume (0–30), net role growth over 90 days (0–30), funding stage and recency (0–25), and canonical-board freshness (0–15). Evidence confidence is separate: it measures verification recency (0–40), canonical-board coverage (0–30), company-record completeness (0–20), and independent-source corroboration (0–10).

The 30-day change shown beside each company counts roles opened minus roles closed in the change feed over the last 30 days. Where we hold no change events for a company in that window it reads flat, because that is what the evidence supports.

We retain each source's detailed industry label and normalize it into a stable public sector taxonomy for filters, sector totals, and movement views. When a source does not publish an industry, we apply a high-confidence context matcher to the company name and domain. Records without a reliable source or context signal remain Other, while missing source labels are shown as Not published.