How we keep results trustworthy

Collected, checked, then shown.

CareerSentry keeps the real RevRag source separate from a small fictional test used to explain self-healing.

Real company source

RevRag output cleanup

Verified
  1. First pass15 raw entries

    The first Bright Data run mixed 10 active roles with five entries marked “Opening soon” and exposed application-form links.

  2. Bright Data repairSame scraper, cleaner output

    An approval-gated Self-Healing cleanup kept active roles and official RevRag detail pages, without changing the Collector ID.

  3. Current result13 verified roles

    The the current stored snapshot contains no error rows or application-form URLs, so those are the records the Jobs page shows.

Technical evidence
Collector
c_mt3ctgtj2rqnwsqm8p
Snapshot
j_mt4woooa2ou55pzj7q
Storage
Neon PostgreSQL

Separate fictional test

What happens when a careers page changes?

Recovery verified

The same three fictional roles are rendered with a different HTML structure. The old scraper returns no rows, which gives CareerSentry a failure to detect and Bright Data a repair to propose.

Before the page change3 rows

The original structure is readable by the collector.

After the page change0 rows

The roles remain, but the old selectors no longer find them.

Recovery status3 rows restored

An inspected same-ID repair returned all three roles after approval, and the schema adapter normalized every row.

Open the fictional test pageIt contains no employer or applicant data.
Technical evidence

Owned test Collector ID: c_mt5mfwuv2910ltr23s. A layout change is intentionally simulated on the stable test URL; one post-approval Layout-B run verified three adapter-valid rows. This experiment is separate from RevRag.

Browsing this page never runs a scraper.

The website reads the trusted snapshot. Collector creation, Bright Data repair, approval, and reruns remain terminal-operated workflows.