Search review

The Search Review workbench (/ops/search-review) allows Operations agents to inspect and evaluate recruiter-generated search queries and AI candidate matching lists.

---

Active queue

The left side displays the query logs:

• Query text: In double quotes (e.g. “React developer 5+ yrs exp”).
• Requester role: e.g. Recruiter, Company.
• Timestamp: Time elapsed since the query occurred.
• Filters applied: List of active search filters (e.g. location: London).
• Results count: The size of the matching candidate result list.
• Review status: Displays Reviewed or Pending QA.

---

Evaluating quality

Select a query card to load the evaluation panel:

1. Overall quality rating

Select a rating for the query matching output:

• EXCELLENT — relevant candidate profiles accurately ranked at the top.
• ACCEPTABLE — relevant matches present, but ranking or alignment is sub-optimal.
• POOR — irrelevant matches or incorrect criteria parsing.

2. Issue tags

Identify specific issues in the query output:

• Hallucinated Skill
• Incorrect Experience Filter
• Irrelevant Location Match
• Other

3. Case notes

Input free-text findings (max 5,000 characters).

---

Result-level labels

Under the result list, every returned candidate is displayed with their rank. Assign direct labels for each candidate card:

Action	Saved Status	Description
Good Match	GOOD_MATCH	Candidate matches query profile.
Poor Match	POOR_MATCH	Candidate is not a good fit for the query criteria.
AI Hallucination Flag	POOR_MATCH	Flags the profile as showing hallucinated skills (automatically tags the query with Hallucinated Skill issue).

Click Submit QA review once all fields and labels are complete to save your evaluation.