People Analytics for Pay Equity
A regression-based pay equity review controls for legitimate drivers - tenure, role, location, performance - and examines what remains unexplained. The method matters less than the discipline: defined variables up front, consistent tolerance, and documented outcomes for every flag.
The model in plain terms
Regress pay on legitimate explanatory factors, then examine residuals - the unexplained remainder - grouped by protected characteristics. A statistically significant pattern in residuals signals structure worth investigating, distinct from individual outliers which usually have stories.
Variables to include
- Role/level (survey job code, not title)
- Tenure in company and in role
- Location differential
- Performance ratings where defensibly calibrated
- Relevant prior experience at hire
Exclude anything contaminated by bias itself - "negotiated aggressively" imports the gap you are testing for.
Cadence and thresholds
| Decision | Common practice |
|---|---|
| Frequency | Annual full run; spot checks at merit cycles |
| Flag threshold | Statistical significance plus practical materiality |
| Remediation budget | Ringed off before merit increases, not scavenged from them |
| Documentation | Every flag resolved, justified or corrected, with date |
Privilege and communication
Work under legal privilege where possible and communicate outcomes carefully - publishing methodology without individual data builds trust; publishing nothing invites assumption. This analysis feeds directly into gap-closing programs, and needs clean inputs from payroll - see what time and labor data enables.
Next
Frequently asked
Do we need data scientists?
No - standard statistical tools handle the core regression. Judgment about variables and thresholds matters more than tooling sophistication.
Can results be used against us legally?
An unremediated documented gap is exposure; a remediated one with records is defense. Run it to act, not to file away.
Get matched with up to three PEOs
Answer six questions about your headcount, states and timeline. We shortlist providers that can actually serve you, and you choose which ones may contact you.
No obligation. You pick which providers get your details. We never sell your information to a data broker.