In a study of nearly 4,000 adults, women who scored exactly as well as men on a math and science test rated their own performance about 13 points lower on a 100-point scale. Same scores, same test, and a different story about it. If you’re staring at a blank year-end self-review form this week, that study is the most useful thing I can hand you, because learning how to write a self-assessment is mostly about noticing which of those two stories you’re about to tell.
The self-review is a pay document, even when nobody says so
Most advice on how to write a self-assessment treats it as a reflection exercise. Ask a Manager’s classic column tells you to focus on outcomes, rate yourself honestly, and name your weak spots before your boss does. Lattice’s guide says to measure yourself against your goals and write “as an advocate.” That’s all good advice. What none of it mentions is that this form usually feeds into the rating, and the rating usually feeds into your raise.
The money attached is not subtle. According to Pave’s 2026 merit cycle data, employees rated “Above Expectations” who weren’t promoted got a median raise of 4.7%, while “Meets Expectations” got 3.3%. And budgets aren’t getting bigger. Mercer’s August 2026 survey has U.S. employers planning an average 3.2% merit increase for 2027. When the pot is that small, the difference between one rating box and the next is most of the raise.
I’m not saying your self-review decides your rating. Your manager does that, often after a calibration meeting you’ll never see. But your self-review is the only document in that process you write yourself. You don’t want to be the one arguing against yourself in it.
Women underrate their work even when no one is watching
The study I opened with is “The Gender Gap in Self-Promotion” by economists Christine Exley and Judd Kessler, published in the Quarterly Journal of Economics in 2022 (the working paper is free through NBER). They gave 3,892 online participants a 20-question test, then asked them to rate how well they did. In the version where an employer would see the answer and use it to decide on hiring and pay, women’s ratings of their own performance came in 12.68 points lower than men’s who scored the same, a 24% drop relative to the average answer.
Here’s the part that changed how I think about it. When the researchers told people exactly how they’d scored, the gap shrank to about 7 points but didn’t disappear. When they removed the employer entirely, so the rating was private and there was nothing to gain, the gap was still 13.46 points. This wasn’t women strategically playing it modest. It’s how we describe ourselves by default.
They found the same pattern in more than 10,000 school-age kids, which tells you this habit starts long before your first performance review. So if your instinct is to soften every sentence on that form, it’s worth treating that instinct as a known bias, not as accurate information.
How to write a self-assessment that matches your real results
The fix in the research is the same fix I’d use on the form: replace adjectives with facts. In the study, knowing the actual score cut the gap nearly in half. You can give yourself that same information before you write a word.
Start by pulling the evidence, not your memory of the year. Open your sent folder, your project tracker, and any notes from one-on-ones, and write down what you shipped, fixed, saved, or grew, with a number on each one where you can. “Supported the client onboarding project” becomes “ran onboarding for 14 new accounts and cut setup time from three weeks to eight days.” If you’ve kept a kudos folder, this is its moment.
Then write the rating question last. If your form asks you to pick a box, read your evidence list first and ask what you’d give a coworker who handed you that list. Ask a Manager suggests imagining how a trusted colleague would rate you, and that’s a useful way to step around your own discount. If the answer is “exceeds,” write “exceeds” and let the facts carry it.
For weak spots, name one or two, pair each with what you’re already doing about it, and stop. A self-review that lists six flaws doesn’t read as humble. It reads as a manager’s to-do list.
Finally, run an edit pass for the small hedges. Words like “helped with,” “tried to,” “was somewhat involved in,” and “I think” quietly shave points off work you led. If you led it, say led.
The $820 math on one rating box
Here’s what the gap between “meets” and “above” is worth on a real paycheck. The Bureau of Labor Statistics reported that the median full-time woman earned $1,127 a week in the third quarter of 2026. Multiply by 52 and that’s $58,604 a year.
At Pave’s 3.3% median for “Meets Expectations,” next year’s raise on that salary is $1,933.93. At 4.7% for “Above Expectations,” it’s $2,754.39. That’s $820.46 more in the first year from one box on a form.
The difference compounds, because next year’s raise is calculated on this year’s salary. If those two paths repeated for five years, the “above” track reaches $73,732.79 while the “meets” track sits at $68,933.27. That’s a $4,799.52 gap in annual pay, and $13,680.56 more in total earnings across the five years. Nobody hands you that money because your self-review was well written, of course. But a review that undersells you makes the higher box harder for your manager to defend in calibration, and that’s the conversation where those raises get decided.
The review is one piece of a bigger pay picture. I’ve written about the gender pay gap and raises and about why unused PTO is a pay cut, and both are worth a read before your year-end conversation.
Write it like the evidence is about someone you’re proud of
The self-review isn’t the place for false confidence, and it isn’t the place for apologies either. It’s a record. Exley and Kessler showed that women describe the same performance less generously than men do, even with nothing at stake. You can’t fix that bias in the whole workplace this week, but you can fix it in one document.
So block an hour, gather the numbers, write like you’re describing a colleague you respect, and choose the rating your evidence supports. If you’re figuring out how to write a self-assessment for the first time, that’s the whole method. The 13 points you stop leaving on the table are yours.