Policy
Adverse-Impact Ratio Calculator
People ops teams computing selection rates and impact ratios per group at one step of a hiring process.
A blank instrument for one step of one process. Enter the counts you already hold, group by group, and it returns a selection rate per group and the ratio of each rate to the highest. Below its stated floor it prints the counts and the floor instead of a ratio, because one selection moved across the line changes a small reading. Arithmetic runs in the browser. Nothing is stored, and no input names a person.
Use this on one step of one process, with counts you already hold. It returns a selection rate per group and the ratio of each rate to the highest one. It takes no candidate record, computes nothing about any individual, and stores nothing.
Counts to gather
Three numbers per group, for one step and one period:
- Considered. How many in that group reached the step.
- Selected. How many in that group came through it.
- Unrecorded. How many reached the step with no category recorded.
Take them at the step being measured rather than at the offer. A step that removes people is measurable on its own terms, and a balanced result at the end of a process describes the end rather than the step.
The group table
One row per group. Copy this into your own file and fill it in.
| Group | Considered | Selected | Selection rate | Impact ratio |
|---|---|---|---|---|
| — | — | — | — | — |
| — | — | — | — | — |
| — | — | — | — | — |
| Unrecorded category | — | not applied | not applied | not applied |
Build the table once per category set you record, and keep those tables separate 3.
The arithmetic
One division per row, then one more.
- Selection rate: selected divided by considered, inside that group.
- Impact ratio: that rate divided by the highest rate in the table.
The group holding the highest rate carries a ratio of 1.00 by construction. Print every rate beside the two counts it came from and the dates those counts cover, in the same line, so a reader can redo the division.
The floor
A ratio computed off a handful of selections is noise, so this instrument refuses to print one.
The floor: a ratio prints only where moving one selection across the line, in either direction, leaves it on the same side of the figure the inlined rule carries. Where either move crosses that figure, or where the highest rate is zero, the counts print and the ratio does not.
What prints in its place is the floor and the counts sitting under it. That is the result, not a missing one.
The one-move check
Recompute twice before anything leaves the room: once with one selection moved from not selected to selected, once the other way 2. It takes about two minutes, and it settles the question the rest of the analysis would argue about. Where a single move changes the reading, the number is a property of this period's applicants rather than of the step.
Worked on invented counts, so nothing here carries a measurement date. Twenty in one group reached the step and six were selected, a rate of 30.0%. Fifteen in another reached it and three were selected, a rate of 20.0%. The ratio is 0.67. Now move one selection: four of fifteen is 26.7%, the ratio is 0.89, and the reading has moved with nothing about the process changing.
The rule's own framing
Print this line under any ratio you report, so the number and its framing travel together.
Treat a selection rate below four-fifths of the highest group's rate as evidence the enforcement agencies weigh, not as a legal threshold you pass or fail. 29 CFR 1607.4(D). 1
Unrecorded categories
Count them and print them on their own line 3. Never distribute them across the other groups and never drop them. A ratio computed after dropping them describes a subset, and the write-up says which subset.
The write-up
State the counts first, then what they support. A sentence a reviewer can check names the population, the period, the number of selections and what follows from them.
Where the floor is met:
Across (period), (count) reached (step) and (count) were selected. Selection rates and impact ratios by group are in the table above, each printed with the two counts behind it.
Where it is not:
Across (period), (count) reached (step) and (count) were selected. Counts by group are in the table above. No impact ratio is reported: moving one selection across the line changes the reading at these counts, so a ratio would describe this pool rather than the step.
Limits
- This returns a property of one step. Whether what the step measures matters for the work is a separate question resting on separate evidence.
- It applies no law to your facts and it is not advice. Your own counsel edits anything you send.
- It assembles no document from your answers. The write-up above is a sentence you write.
- The smallest unit of input is a group. There is no name field, no row standing for one person, and no individual result.
- Olive's own bias audit has not been performed. No result here describes any Olive product, and Olive computes, stores and transmits no number about any person.
- Nothing is stored. The arithmetic runs in your browser, no input leaves the page, and reset is a reload.
Run it
Counts, one row per group
Group 1
Group 2
Group 3
Group 4
Group 5
Unrecorded
Nothing here is stored and nothing is sent. The arithmetic runs in your browser, and reset is a reload.
Take it
The file and the credit
The publishing entity legal name and postal address are not filled in yet, and both sit inside the disclaimer every packaged format renders. No file is emitted until they are.
Credit line, to paste beside anything you quote from this document.
<!-- Olive template. Licence: https://olive.is/licence/template/1-0 -->
<p><a href="https://olive.is/answers/tools/adverse-impact-ratio-calculator/" rel="nofollow">Olive</a>, Adverse-Impact Ratio Calculator, version 1.0.0, checked 2026-08-26.</p>Important notes
Not a substitute for the advice of an attorney. This is a starting document for your own attorney to edit. It applies no law to your facts. No attorney has reviewed it for your state or your facts, and Olive is not your lawyer.
Published by Olive Independent Study, Inc., [DELAWARE INCORPORATING ADDRESS], United States. Contact hello@olive.is. A person reads every complaint and answers within ten working days. All concerns that Olive has engaged in the unauthorized practice of law are referred to the North Carolina State Bar, wherever the complaint came from.
Checked 2026-08-26 against the sources listed in this file. Version 1.0.0.
This text disclaims no warranty, caps no liability, waives no remedy, and names no court or state for a dispute. Those absences are deliberate.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.
Packaged files
- The publishing entity legal name and postal address are not filled in yet, and both sit inside the disclaimer every packaged format renders. No file is emitted until they are.
Checks
- Sources last re-opened August 26, 2026.
- Next review due February 26, 2027.
Where it came from
Read the argument behind it
- How Do You Run an Adverse Impact Audit When the Vendor Holds the Data? Run the four-fifths ratio on your own ATS data. Under 0.80 is evidence of adverse impact, not proof, and occupation-specific validity evidence decides if the tool stays.
- Too Few Hires for a Ratio: Measure the Process Instead At single-digit selection counts one candidate flips the verdict. Run the flip test, report counts, and measure the decisions instead, which you have enough of.
References
- 29 CFR 1607.4 - Information on impact (Uniform Guidelines on Employee Selection Procedures, 1978) Code of Federal Regulations, via Cornell Legal Information Institute, 1978. law.cornell.edu The address of the paragraph the fact marker inlines, so a reader can open the wording the printed line came from rather than take it from this file.
- Questions and Answers to Clarify and Provide a Common Interpretation of the Uniform Guidelines on Employee Selection Procedures (Q.11, Q.19) U.S. Equal Employment Opportunity Commission (joint EEOC-DOJ-OPM-DOL-Treasury document, OLC Control Number EEOC-NVTA-1979-1), 1979. eeoc.gov Where the one-move check comes from: the agencies' own small-numbers answer, that selecting one different person for one job reversing the result is the sign the counts are too small to act on.
- Notice of Adoption of Final Rule: Automated Employment Decision Tools (6 RCNY 5-300 to 5-304) NYC Department of Consumer and Worker Protection, 2023. rules.cityofnewyork.us Where the unrecorded line and the separate tables come from: the unknown-category count, and computing sex, race and ethnicity, and intersectional categories as separate tables.
3 sources, numbered by first appearance. How Olive sources claims