What is the minimum sample size for a bias audit?
There is no legally mandated minimum sample size, but statistical reliability requires sufficient data:
- **Minimum recommended:** 30 candidates per demographic group per pipeline stage
- **Ideal:** 100+ candidates per group for high-confidence results
- **Below 30:** Results are flagged as low-confidence in OnHirely reports
- **Below 5:** Insufficient for meaningful statistical analysis
For the 4/5ths rule to be meaningful, each comparison group needs enough candidates to produce a reliable selection rate. Statistical significance tests (chi-squared, Fisher's exact) also require adequate sample sizes.
If your dataset is small, OnHirely uses Fisher's exact test (designed for small samples) instead of chi-squared, and clearly flags confidence levels in the audit report.
For organizations without sufficient historical data, the NYC LL144 allows the use of test data — synthetic candidate data designed to test the AI tool's behavior across demographic groups.
Related Questions
Still Have Questions?
Start a free audit and see how OnHirely makes AI hiring compliance simple.
Start Free Audit