Sample report
Independent bias audit: Northwind Logistics (fictional)
A redacted, made-up example of the report an employer receives. The company, tool and figures are fictional, so you can see the structure and the kind of findings before you book an audit.
1. Scope and tool
- Employer
- Northwind Logistics (fictional)
- Tool audited
- , a resume-screening tool that scores applicants and advances those above a cutoff to interview
- Vendor
- Jobs covered
- Warehouse associate and dispatch roles in New York City
- Data period
- Twelve months of historical applicant data (illustrative)
- Employer contact
The audit covers one automated employment decision tool used to screen candidates, as described under New York City Local Law 144. It reports the figures the law requires. It doesn't certify the tool, and the law sets no pass or fail threshold.
2. Data and method
- Data: the employer's own historical records for 3,518 applicants with known sex and race/ethnicity, of whom 1,000 were advanced. Another 95 applicants (26 advanced) had an unknown category; they are reported here but not in the impact ratios.
- Selection rate: the share of applicants in a category that the tool advanced.
- Impact ratio: a category's selection rate divided by the rate of the most selected category. Calculated by sex, by race/ethnicity and by both together (intersectional categories).
- The four-fifths rule (an impact ratio below 0.80) is used only as a screen to point to results worth a closer look. It is not a legal test of bias.
- Small samples: categories with under 2% of the data are shown in brackets and left out of the comparison, because a few decisions can swing their rates widely.
- Every figure was recalculated from the raw data, not taken from the vendor's reports.
3. Results
| Category | Applicants | Selected | Selection rate | Impact ratio | Note |
|---|---|---|---|---|---|
| Male | 1,792 | 535 | 29.9% | 1.00 | |
| Female | 1,726 | 465 | 26.9% | 0.90 |
| Category | Applicants | Selected | Selection rate | Impact ratio | Note |
|---|---|---|---|---|---|
| Hispanic or Latino | 820 | 234 | 28.5% | 0.92 | |
| White | 1,170 | 363 | 31.0% | 1.00 | |
| Black or African American | 790 | 181 | 22.9% | 0.74 | Below 0.80 screen |
| Native Hawaiian or Pacific Islander | 34 | 9 | 26.5% | (0.85) | Under 2% of data |
| Asian | 560 | 174 | 31.1% | 1.00 | |
| American Indian or Alaska Native | 26 | 7 | 26.9% | (0.87) | Under 2% of data |
| Two or More Races | 118 | 32 | 27.1% | 0.87 |
| Category | Applicants | Selected | Selection rate | Impact ratio | Note |
|---|---|---|---|---|---|
| Male, Hispanic or Latino | 420 | 126 | 30.0% | 0.94 | |
| Male, White | 610 | 195 | 32.0% | 1.00 | |
| Male, Black or African American | 380 | 95 | 25.0% | 0.78 | Below 0.80 screen |
| Male, Native Hawaiian or Pacific Islander | 18 | 5 | 27.8% | (0.87) | Under 2% of data |
| Male, Asian | 290 | 93 | 32.1% | 1.00 | |
| Male, American Indian or Alaska Native | 14 | 4 | 28.6% | (0.89) | Under 2% of data |
| Male, Two or More Races | 60 | 17 | 28.3% | (0.88) | Under 2% of data |
| Female, Hispanic or Latino | 400 | 108 | 27.0% | 0.84 | |
| Female, White | 560 | 168 | 30.0% | 0.94 | |
| Female, Black or African American | 410 | 86 | 21.0% | 0.65 | Below 0.80 screen |
| Female, Native Hawaiian or Pacific Islander | 16 | 4 | 25.0% | (0.78) | Under 2% of data |
| Female, Asian | 270 | 81 | 30.0% | 0.94 | |
| Female, American Indian or Alaska Native | 12 | 3 | 25.0% | (0.78) | Under 2% of data |
| Female, Two or More Races | 58 | 15 | 25.9% | (0.81) | Under 2% of data |
4. Findings in plain words
- The tool advanced female applicants at 0.90 times the rate of the most selected sex.
- The lowest race/ethnicity result was Black or African American applicants, at 0.74 times the rate of the most selected group.
- 3 reported categories fall below the four-fifths screen: Black or African American; Male, Black or African American; Female, Black or African American.
- Several small categories could not be compared reliably; their figures are shown for completeness only.
5. Recommendations
- Review which resume features drive the score for the flagged categories, with the vendor, before the next hiring cycle.
- Check whether the score cutoff is set higher than the job requires.
- Improve the collection of voluntary demographic data to reduce the number of unknown records.
- Publish the summary of results and give candidates the required notices, as the law requires of the employer.
6. Limitations
- Results describe past data only and may change as applicant pools or the tool change.
- Impact ratios show differences in outcomes; they don't explain why the differences occur.
- Demographic data is self-reported and incomplete for some applicants.
- This report is not legal advice.
7. Auditor independence statement
- Audits are performed and signed by BlissPoint Consulting, LLC, a separate company from the one that sells Organization Viewer, under its own agreement and billing.
- BlissPoint never audits a tool it built, sold, resold, configured or recommended.
- No AI Hiring Risk Assessment, remediation advice or other consulting on the same tool in the 12 months before the audit, or during it, so the auditor never reviews their own advice.
- The audit is never bundled with Organization Viewer or other consulting, and never discounted for buying anything else.
- The fee is fixed before work starts and doesn't depend on the results.
- Any other relationship with the employer, including an Organization Viewer subscription, is disclosed in writing before the engagement and stated in the audit report. The auditor declines the engagement if a lawyer advises the relationship could impair independence.
- BlissPoint holds no ownership or other financial interest in the employer or the tool's vendor, and has no employment relationship with either.
- The auditor recalculates every figure from the raw data, whatever software the employer used to prepare it.
Signed: , independent auditor. Date:
Want this for your company?
Prefer to talk first? Book a free 15-minute call
From an I-O psychology consultant. Covers NYC Local Law 144, the EU AI Act and the four-fifths rule.