Find the Ideal Candidate among 25,000 CVs
A tech consultancy receives 25,000 CVs a month for 200 vacancies. The keyword filter discards good candidates who describe their experience in other words ('I orchestrated the migration to the cloud' instead of 'AWS') and lets through people who stuff their CV with trendy terms. Recruiters take 23 days to present finalists.
Where the money comes from
What to do now
- 1Measure the cost of a vacancy open per day (lost billing).
- 2Semantic search over the candidate base with a fairness audit.
- 3Human review of the top 20 and explanation of each recommendation.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Hiring faster people who resemble past hires can raise turnover or reduce diversity. Measure hiring quality at one year, not just speed.
KPI for the committee
Days to hire and first-year turnover.
Detail for your data team
Proposed technique: Embeddings + Vector Search
Why: The problem is one of meaning: equivalent experiences described with different words. Embeddings represent CVs and vacancies in the same semantic space, and vector search finds the closest ones among thousands in milliseconds. It is a prioritisation tool with a human at the end.
Cost of getting it wrong: A false negative is a great candidate who is never seen; a biased ranking can mean discrimination and regulatory penalties.
Data needed: CVs in free text (PDF) and professional profiles.; Vacancy descriptions and profiles of employees who have succeeded in each role.; Process history: who went to interview, who was hired and their one-year evaluation.