Resource guide · October 14, 2026
Your hiring tool had a bias audit. What did it actually show?
Why group outcome figures matter, and which questions remain after you have an audit report.
← All resourcesRead the result in its scope
First, identify which tool version, roles, candidates, time period, and hiring stage were examined. Was the report based on the people who applied to your organization, data pooled from other users, or another permitted dataset? The answer affects how you interpret the findings in your setting.
A bias audit may calculate and compare selection rates or scoring rates across groups. New York City’s rules specify such calculations for covered automated employment decision tools. That required analysis answers a narrower question than whether every score is job relevant or every candidate can access the process. NYC AEDT rules and FAQ ↗
Imagine a screening tool advances 20% of applicants in each of two groups. Those figures show similar selection rates for those groups in that sample. Now imagine its score relies heavily on how quickly someone answers a timed prompt. If nobody has shown that response speed matters for the job, the equal rates do not resolve whether that score is appropriate for selection. The small example also says nothing about other groups, candidates who could not complete the prompt, or how outcomes change when the tool is used for another role.
Then ask what the score means
What does the tool claim to measure? How was that claim tested against the work people will do? Does the supporting study cover your jobs and your use of the tool? The EEOC advises employers to ensure selection procedures are properly validated for the positions and purposes for which they are used. EEOC guidance on selection procedures ↗
Look for limitations and next steps
- Were sample sizes or missing data too limited to draw a clear conclusion about some groups?
- Could accessibility barriers or accommodation requests change who completes the assessment?
- Are the vendor’s findings applicable to your threshold, roles, and hiring workflow?
- What will be monitored if the model, the candidate pool, or the way the score is used changes?
An audit report should state the questions it answered, the data used, and its limitations. A favorable group comparison in one analysis is useful information; it is not a universal pass or a guarantee that a tool is fair, valid, or legally compliant.