Prove It Again: Why Your Competence Gets Re-Litigated and His Gets Assumed
The usual explanation for why your career moves slower than a peer’s is that people underrate your work. The evidence says something more specific, and more useful. They do not underrate your work. They rate it higher than his. What they discount is what your work is taken to predict.
That distinction is the whole game.
The study that names the mechanism
In February 2026, the American Economic Review published “Potential” and the Gender Promotion Gap by Alan Benson, Danielle Li and Kelly Shue. The researchers looked at 29,809 management-track employees at a large retail chain, where every employee received two separate scores: a performance rating for work already delivered, and a potential rating for what they were expected to become.
Women scored higher on performance than men. They scored substantially lower on potential. The gap in potential ratings accounted for roughly half of the entire gender promotion gap at the company.
Then the researchers checked whether the forecasts were any good. They were not. The women who did get promoted went on to outperform the men promoted alongside them, both on average and at the margin. The low potential ratings were not reading something the performance ratings missed. They were wrong.
This is one company, and it deserves to be read as one company. But it is a peer-reviewed finding in a top economics journal, and it converges with a pattern that has been documented qualitatively for over a decade.
The pattern already had a name
Joan C. Williams called it Prove It Again. In What Works for Women at Work, built on interviews with 127 successful professional women, Williams identified it as one of four recurring bias patterns: women reset to zero credibility in each new room while their male colleagues carry a presumption of competence forward.
The Bias Interrupters project at NYU’s Meltzer Center states the mechanism plainly in its performance evaluation toolkit: majority men tend to be judged on their potential, while women, people of color and other groups are judged on their performance. Williams described the pattern. Benson, Li and Shue put a number on it.
What it costs at the first rung
The downstream effect shows up in the pipeline data. Women in the Workplace 2025, the eleventh annual study from LeanIn.Org and McKinsey drawing on more than 120 companies and 9,000 employees, found that for every 100 men promoted to manager, only 93 women were. For Asian women and Latinas the number is 82. For Black women it is 60.
The same report found that only 31% of entry-level women have a sponsor, compared with 45% of men. That combination matters. The first promotion is decided with the thinnest performance record available, which means it leans hardest on the forecast, and the forecast leans hardest on whoever is willing to make it out loud.
Read this through Authority
Authority is evidence of judgment made visible. Not output. Judgment.
Performance ratings measure output, and you are already winning there. Potential ratings measure a forecast of judgment, and that is where the loss happens. The instinct when you feel under-credited is to produce more, better, faster. The research says that is the wrong lever. Output is the thing already working. What is not working is that your judgment is not legible as something that predicts.
You are being evaluated on a question you are not answering: what does she see before anyone else does?
Three moves
1. Put a forecast on the record, and date it. Judgment only reads as potential when it is visible before the outcome, not after. Before the next consequential decision, send a short written call: what you expect to happen, why, and what would change your mind. Two paragraphs is enough. Then reference it once when the result lands. This converts a record of output into a record of prediction, which is exactly what a potential rating is trying to estimate and currently guessing at.
2. Ask for your potential rating separately, in writing. Bias Interrupters recommends organizations assess performance and potential as separate criteria, precisely because collapsing them is where the bias hides. You do not have to wait for your company to adopt that. Ask directly: “You have my performance. What is your read on my ceiling, and what specifically would move it?” A vague answer is useful data about your position. A specific answer is a plan.
3. Get someone else to make the forecast. A potential rating is a bet stated out loud, and bets you place on yourself land differently than bets others place on you. Give one senior person something concrete enough to repeat in a room you are not in. Not “she is strong,” which evaporates. Something closer to “she flagged the retrieval failure six months before we hit it.” Hand them the sentence. That is not immodesty, it is supplying evidence to someone who needs it to advocate well.
You are not short on evidence. You are short on the kind of evidence that reads as a forecast. That is an Authority problem, and Authority problems have moves.
Which dimension is actually constraining you right now? The Impact Architecture diagnostic takes a few minutes and tells you whether Authority, Reach, Access or Reputation is your primary constraint. Take the diagnostic.
Sources
- Benson, A., Li, D., & Shue, K. (2026). “Potential” and the Gender Promotion Gap. American Economic Review, 116(2), 375-417.
- Williams, J. C., & Dempsey, R. (2014). What Works for Women at Work: Four Patterns Working Women Need to Know. NYU Press. See also the Bias Interrupters Performance Evaluations toolkit, Meltzer Center, NYU School of Law.
- LeanIn.Org & McKinsey & Company (2025). Women in the Workplace 2025.
