A potential rating is a manager's forecast of how far someone can grow, usually recorded on the vertical axis of a nine-box grid next to a performance rating. Because it is a guess about the future rather than a judgement of past work, it is the part of calibration where bias hides most easily. Treat potential as a claim that needs evidence, review it by group, and never let it outweigh a strong performance record without a written reason.
It is written for CPOs and People leaders in UK and Irish firms with 100 to 1,500 employees who are about to calibrate year-end ratings and agree promotions. A study published in the American Economic Review in February 2026 found that women at one large employer were rated higher on performance and lower on potential, and that the potential gap explained about half of the gender promotion gap.
Key takeaways
- Benson, Li and Shue (American Economic Review, February 2026) studied 29,809 management-track employees and found women received higher performance ratings but lower potential ratings than men.
- In that study, differences in potential ratings accounted for about half of the gender promotion gap, and women later outperformed the men they had been ranked below.
- The authors point to two drivers: stereotyping, and managers rating people they feared losing more generously.
- Potential is a forecast with no record behind it, so calibration should test it with written criteria and examples, exactly as it tests performance.
- Before your next calibration, run last cycle's ratings through a bias check by group and decide which promotion inputs you will make visible.
What a potential rating measures, and what it does not
Calibration meetings tend to mix three different questions, and potential sits in the vaguest of them.
- Performance asks what someone delivered this period against the bar for their level. There is a record to check.
- Readiness asks whether they already show the behaviours the next level needs. There should be examples.
- Potential asks how far they might go. It is a prediction, and often nobody writes down what it rests on.
A potential rating is not a measure of ambition, confidence in meetings or how much a manager likes working with someone. When it drifts towards those things, it rewards visibility rather than results. Our guide on why the 9-box grid is dead covers the wider problems with the grid itself.
What the 2026 AER study found about potential ratings
Alan Benson (University of Minnesota), Danielle Li (MIT) and Kelly Shue (Yale) analysed 29,809 management-track employees at a large North American retail chain. Supervisors rated each person on a nine-box grid: performance from 1 to 3 and potential from 1 to 3. The paper appeared in the American Economic Review (volume 116, issue 2, February 2026). The figures in the second half of this table come from the authors' 2022 working paper version, so treat them as indicative; the headline findings are the published ones.
| Finding | Figure | What it means for your calibration | What to change |
|---|---|---|---|
| Women's performance vs potential ratings (published, 2026) | Higher on performance, lower on potential | A strong record can be cancelled out by a forecast nobody has to justify | Ask for evidence behind every high and low potential call |
| Share of the promotion gap explained by potential (published, 2026) | About half | The promotion decision is made in the potential column | Show performance and potential side by side when promotions are discussed |
| What happened next (published, 2026) | Women went on to outperform male colleagues | Low potential calls for women were wrong forecasts, not cautious ones | Compare last year's potential ratings with this year's performance |
| Top potential rating (2022 working paper) | Women 28% less likely to receive it | The gap concentrates at the top of the grid, where promotions come from | Review the "high potential" list by group before it is shared |
| Weight of potential in promotion (2022 working paper) | Moving from 2 to 3 on potential raised the annual promotion rate by 75%; the same move on performance, by 27% | Your system may be paying more for the forecast than for the results | Decide, and publish, how much each input counts |
The authors name two mechanisms. One is stereotyping: in the working paper, women's potential ratings were lower relative to men's in US counties with fewer women in management roles, a result the authors call suggestive. The other is strategic retention: in the working paper, men who were passed over for promotion were 32% more likely to leave, against 12% for women, which suggests managers may rate the people they fear losing more generously. Caveats matter here. This is one employer, in retail, in North America, using one grid. The study shows how potential ratings behaved in that firm, not that every nine-box is biased. Wider data points the same way, though: McKinsey and LeanIn.Org's Women in the Workplace 2025 found that only 93 women were promoted to manager for every 100 men across 124 organisations in corporate America.
6 fixes for potential ratings before calibration
1. Write down what potential means in your company
Define it as observable behaviours expected at the next level, not as a feeling. Why: a criterion people can read is a criterion people can challenge. The talent bar by seniority is a good starting point for that wording.
2. Require one example for every high or low call
A potential rating of 1 or 3 needs a dated example, just like an extreme performance rating. Why: the AER study shows forecasts drift furthest from reality when nobody has to justify them.
3. Look at ratings by group before the meeting, not after
Run last cycle's performance and potential ratings through the free performance review bias checker, then bring the pattern into the room. Why: a gap noticed during calibration can still be fixed, while one noticed in the pay review cannot.
4. Separate "might leave" from "has potential"
Discuss flight risk as its own item, with its own response such as pay, scope or a career conversation. Why: the strategic-retention finding suggests the two get blended, and the people least likely to complain lose out.
5. Make promotion inputs transparent
Publish what a promotion case must contain and how performance, readiness and potential are weighed. Why: GOV.UK guidance for employers (August 2025) rates transparency in promotion, pay and reward processes as an effective action, because managers who know their decisions can be reviewed tend to base them on evidence.
6. Run calibration as an evidence review
Use a fixed agenda: outliers first, then any potential rating that contradicts the performance record. Why: the same GOV.UK guidance lists calibration as a promising action, but only because it creates accountability. Our step-by-step guide to running calibration sets out the agenda.
Risks when you change how potential is rated
- Managers stop rating potential honestly and cluster everyone in the middle. Safeguard: keep the middle allowed, but ask for a development note for every person, whatever their rating.
- Group reviews feel like quotas. Safeguard: the rule is that a gap triggers a second look at the evidence, never an automatic change of rating.
- Small teams make the numbers noisy. Safeguard: look at patterns across departments and two cycles, not one team of six.
- The promotion decision becomes a spreadsheet. Safeguard: the final call and the conversation with the employee stay with the manager and their leader, in person.
If you want a wider view of where bias enters reviews, our pillar page on performance review bias and calibration flags lists the common patterns.
How to tell whether potential ratings are getting fairer
These are playbook suggestions, not research findings. Track them over two cycles.
- The gap between performance and potential ratings, by gender and by department.
- The share of high and low potential calls that carry a written example.
- How often last year's potential rating predicted this year's performance rating.
- Promotion rates by group at the first step up to manager, the rung Women in the Workplace 2025 highlights.
- Regretted leavers among people rated high performance but low potential.
For Irish employers, these numbers also help with the "reasons and measures" part of the gender pay gap report. The Department of Children, Disability and Equality (guidance updated October 2026) says organisations with over 50 employees must report this year by the end of November, including the employer's view of the reasons for any gap.
Our perspective
In a lot of calibration meetings, the heat goes into the performance column and potential gets waved through in the last ten minutes. That is backwards. Performance has a paper trail; potential is the one number a single manager can set on instinct, and it is often the one that decides who gets promoted. My advice to People leaders is simple: if a potential rating could not survive the question "show me an example", it should not survive the meeting. Treat the forecast with more suspicion than the record, not less.
Ash
Where Revolut People fits, and where it doesn't
The AER finding is a problem of evidence: a forecast carried the weight and the record did not. That is where software can help. In Revolut People, every review is built on a three-part scorecard of goals, skills and values, so a promotion discussion starts from what someone delivered and how they did it, measured against a competency framework for their role. Calibration dashboards suggest a grade using several inputs and highlight outliers, including lenient managers and inflated grades, so the meeting spends its time on the exceptions. Ebury, which moved off grading spreadsheets, reports departmental calibration falling from four hours to one in the Ebury calibration story.
The honest limit: no tool can tell you whether a manager's forecast about someone's future is fair. Software can make the evidence visible and the outliers obvious, but defining potential, asking for the example and changing a rating still takes a People team willing to push back. If you want to test this against your own process, a 15-minute discovery call is a fit-check: bring last cycle's nine-box or promotion list with names removed, and you will leave with the two or three places where a forecast is outweighing the record, whether or not Revolut People suits you.
A first 90 days plan for fairer promotion decisions
| When | What to do | Owner |
|---|---|---|
| Days 1 to 30 | Pull last cycle's performance and potential ratings and promotions; compare them by group and department; write a one-page definition of potential for each level | People team |
| Days 31 to 60 | Brief managers on the definition; require an example for every high and low call; add flight risk as a separate item on the calibration agenda | People partners with department heads |
| Days 61 to 90 | Run calibration as an evidence review; publish what a promotion case must contain; record which ratings changed and why | People team and senior leaders |
The calibration meeting template for mid-market tech gives you the agenda and the outlier sheet for the final month.
Bottom line
The February 2026 American Economic Review study is a clear warning about forecasts: at one large employer, women out-performed men on the record and lost out on potential, and that one subjective rating explained about half the gap in promotions. Most companies of 100 to 1,500 people use some version of a potential judgement, even without a formal grid. Define it, demand an example for it and check it by group before calibration, and promotion decisions become easier to defend to employees, to the board and, in Ireland, in this month's pay gap report.
FAQ
What did the 2026 American Economic Review study find about potential ratings?
Alan Benson, Danielle Li and Kelly Shue studied 29,809 management-track employees at a large retail chain that used a nine-box grid, in a paper published in the American Economic Review in February 2026. Women received higher performance ratings but lower potential ratings than men, the potential gap accounted for about half of the gender promotion gap, and women went on to outperform male colleagues.
What is the difference between a performance rating and a potential rating?
A performance rating looks back: it judges what someone delivered against the bar for their role over the review period. A potential rating looks forward: it is a manager's forecast of how far and how fast someone could grow. Because a forecast has no record to check against, it is far easier for impressions, stereotypes and retention worries to shape it without anyone noticing.
Should we stop using the nine-box grid?
Not necessarily, but stop letting the potential axis decide promotions on its own. Playbook guidance: rate potential only against written, observable criteria for the next level, require an example for every high or low call, and review potential ratings by group in calibration. If you cannot define what high potential looks like in your company, the axis is adding noise rather than information.
Does calibration reduce bias in performance and promotion decisions?
It can help, if it asks for evidence. GOV.UK guidance for employers published in August 2025 lists calibration as a promising action: when managers know others will review their decisions, they seek more information and make less biased calls. The same guidance rates transparency in promotion, pay and reward processes as an effective action. A calibration meeting that only debates impressions will not deliver either benefit.
Do Irish employers need to explain their gender pay gap in 2026?
Yes, if they are in scope. Guidance from Ireland's Department of Children, Disability and Equality, updated in October 2026, says organisations with over 50 employees must report in 2026, based on a June snapshot date, by the end of November, through the Gender Pay Gap Portal and their website. The report must include the employer's view of the reasons for any gap and the measures being taken.
Sources
- American Economic Review: Alan Benson, Danielle Li and Kelly Shue, "'Potential' and the Gender Promotion Gap", vol. 116, no. 2, pp. 375 to 417 (2026)
- Benson, Li and Shue: "'Potential' and the Gender Promotion Gap", working paper version (2022)
- McKinsey & Company and LeanIn.Org: Women in the Workplace 2025 (2025)
- GOV.UK, Office for Equality and Opportunity: How to improve gender equality in the workplace: actions for employers (2025)
- Government of Ireland, Department of Children, Disability and Equality: What is the Gender Pay Gap Information Act 2021? (updated 2026)