Promotion and Pay Action Analyzer
Promotion rates and pay increases by any grouping, raw and like for like, with every gap tested against chance.
Reads .xlsx and .csv inside your browser. Nothing is uploaded.
What this is not
It tests no legal threshold and reaches no conclusion about why a gap exists. A group that stands out is a question about how promotion or pay decisions were made, for the people who can look at those decisions. If you are preparing anything for a regulator or a court, it is a way to understand your data first, not the analysis itself.
The review cycle
Drop a spreadsheet here
One row per person: the group to compare by, and whether they were promoted and what their pay did. Level and rating let groups be compared like for like.
Two gaps, never one
A group promoted at half the rate of everybody else may be a group concentrated in levels where promotion is rare. A group with smaller raises may be a group that was promoted less, since promotions carry their own raise. So every group gets the plain gap beside the gap among people at the same level, with the same rating and, for pay, the same promotion outcome. When the gap disappears like for like, the question moves to how people got into those levels. When it survives, the question is about the decisions themselves.
A gap is tested, not eyeballed
With a handful of groups of a few dozen people, gaps of ten points happen by chance all the time. Each like-for-like gap is tested by shuffling group membership thousands of times within the same levels and ratings, and asking how often chance alone produces a gap as large. With no levels to hold constant, that test is Fisher’s exact test, and it is checked against it. Every group is tested at once, so the results are corrected for that: a group that stands out survived the correction, one worth a look did not. The same file always gives the same answer.
Holding rating constant can hide the gap
A rating is itself a judgment. If one group is rated lower, comparing people with the same rating removes exactly the difference worth finding. Run it with and without the rating: whatever gap appears only without it sits in the ratings, and the Rating Distribution Analyzer is the place to look at those.
Small groups are protected
A group under the minimum size is not shown, and when hiding one group alone would let anyone with the totals work it out by subtraction, a second is hidden too. Hidden groups are left out of the download, and still counted in everybody else’s comparisons. The minimum is yours to raise.
Everybody else includes the outlier
Each group is compared with everybody else. When one group is far below the rest, every other group looks a little above it, and with a large enough file even that small lift can pass a test. Read the largest gap first.
Related tools
For pay levels rather than pay changes, the Pay Equity Gap Analyzer compares raw and like-for-like pay gaps, and the Merit Increase Modeler plans a raise budget before the decisions are made.