“Watch your tone” is not a performance goal. In one study of 248 reviews, personality criticism showed up in 71 of 94 critical reviews of women and 2 of 83 of men. 5 ways to keep personality out of your ratings.

A Thursday question for People leaders, CPOs and COOs in Dublin, London, Amsterdam, Stockholm and Tallinn: if you read last cycle’s written reviews side by side, how much of the critical feedback is about the work, and how much is about the person’s personality?

It comes up again and again when managers ask for help online: how do you tell someone their manner is a problem without the review turning into a verdict on who they are? It’s a fair question, and the honest answer starts with the data. One of the clearest pieces of evidence is still Kieran Snyder’s study in Fortune, “The abrasiveness trap”. It started with an engineering manager preparing promotions: one person was “really talented” but “abrasive”, the other was “smart and great to work with” and just needed “to be a little more patient”.

1. What the data actually says

• The sample. 248 reviews from 180 people, 105 men and 75 women, at 28 companies ranging from large technology corporations to mid-size and smaller firms. Snyder saw no effect of company size and no pattern unique to one company.
• Most reviews carried criticism. 177 of the 248 reviews, about 71%, contained critical feedback, even though almost all of them were positive overall.
• It wasn’t spread evenly. 58.9% of the reviews received by men contained critical feedback, against 87.9% of the reviews received by women.
• The type of criticism differed most. Men’s critical feedback was mostly suggestions for skills to develop. Negative personality criticism appeared in 2 of the 83 critical reviews received by men and in 71 of the 94 received by women.
• The same words kept coming back. “Abrasive” was used 17 times to describe 13 different women. Of the words Snyder tracked, only “aggressive” appeared in men’s reviews, three times, and twice as encouragement to be more of it.
• It wasn’t about who wrote the review. Women wrote 62 of the 248 reviews, about 25%, and accounted for just over 23% of the negative critical feedback.

Honest caveats. This study is from 2014, covers US technology reviews and relies on reviews people chose to share, mostly from strong performers. Snyder herself says she can’t tell whether women were simply more willing to share reviews with critical language. It isn’t a benchmark for your company. Her own conclusion is the useful part: most HR teams already check rating scores for systematic bias, and the same rigour should apply to the review language itself.

2. The mid-market hurdle: managers write the words, nobody checks them

In a company of 100 to 1,500 people, most managers write reviews with a free-text box and a rating, often for the first time, often in the last week before the deadline. Calibration then looks at the ratings. It rarely looks at the sentences behind them. So a phrase like “comes on too strong” can sit in someone’s record, shape a promotion conversation and never be tested against the work.

That is a problem for a CPO and for a COO for different reasons. For the CPO it is a fairness and retention risk: people remember the one line in a review that described them rather than their work, and they don’t know what to change. For the COO it is a quality problem: if feedback can’t be tied to an outcome, it doesn’t help anyone deliver more next cycle. A strong performer who hears “be less abrasive” has nothing to act on. One who hears “in the last two design reviews you closed the discussion before the two newer engineers had spoken, and we missed a risk they had spotted” knows exactly what to do.

3. 5 ways to keep personality out of your ratings

1. Rate against a written scorecard, not an impression. Before the cycle opens, write down what is being judged: goals delivered, skills for the person’s level and how they work with others, described as behaviours. “Shares context early so others can act” can be checked. “Has a good attitude” can’t. The rating scales guide has examples.
2. Use a simple test on every critical sentence. Does it name what happened, the effect on the work and what good would look like? If it only says what the person is (abrasive, too quiet, not a team player), send it back to the manager to rewrite before it is submitted.
3. Read the language, not only the scores, in calibration. Snyder’s point was that review text deserves the same scrutiny as ratings. Take a sample of written feedback by team and by group before the calibration meeting and count how much describes personality. The review bias checker and the guide to reducing rating bias help.
4. Collect the evidence during the year, in 1:1s. Personality labels fill the gap when a manager has no examples to hand. A short note after each 1:1 with what went well and what to change gives the review real examples, and nothing in the review should be the first time someone hears about a concern.
5. Give managers two examples and ten minutes of training. Show one personality comment and the same point rewritten as behaviour and impact. Most managers aren’t biased on purpose. They just haven’t been shown the difference, and first-time managers least of all.

For the PDF, use your work email. The call is a free 15-minute discovery call: bring your review template and a few anonymised review comments.

4. Where Revolut People fits, and where it doesn’t

Revolut People started as the system Revolut built to run performance as it scaled to 20,000 people across 60+ countries. The parts that map to this problem are the ones that keep reviews tied to the work. Reviews use one scorecard (goals, skills by seniority and values), so every manager rates against the same written bar, and each review opens with the person’s goals and previous-cycle data already filled in. Goals cascade from company to team to person, so feedback can point at what was actually delivered. 1:1s come with AI-powered summaries, so examples and agreed actions are on record long before review time. Calibration brings performance data into one place, so a committee can compare ratings across teams and see the evidence behind them, not just a single number. Core HRIS records, including reporting lines, sit on the same platform.

To be straight about the limits: software won’t stop a manager writing “abrasive”. Someone still has to read the comments, push back on personality labels and coach managers on better feedback. What one shared record does is put the scorecard, the goals and the 1:1 evidence next to every rating, so the conversation is about the work. Related reads: when feedback crosses the line, PwC on trust in managers and year-end reviews, and feedback that never sticks.

Bottom Line

In Snyder’s study of 248 reviews, personality criticism appeared in 71 of 94 critical reviews received by women and 2 of 83 received by men, whoever wrote them. The data is old and small, but the lesson travels: a rating is only fair if the feedback behind it is about the work. For a mid-market People team or COO in the UK, Ireland, Benelux, the Nordics or the Baltics, that means a written scorecard, a behaviour-and-impact test on every critical sentence, review language checked in calibration and evidence collected in 1:1s. Try it with one department this cycle.

FAQ

What did the review study find?

Kieran Snyder collected 248 performance reviews from 180 people (105 men and 75 women) at 28 technology companies of different sizes, and published the results in Fortune on 26 August 2014. 177 of the 248 reviews, about 71%, contained critical feedback. 58.9% of the reviews received by men contained critical feedback, against 87.9% of those received by women. Negative personality criticism, such as being told to watch your tone or step back, appeared in 2 of the 83 critical reviews received by men and 71 of the 94 received by women.

Did it matter whether the manager was a man or a woman?

No. In the study, women wrote 62 of the 248 reviews, about 25%, and accounted for just over 23% of the negative critical feedback, close to what you would expect from their share. The pattern came from the reviews, not from one group of managers.

What is the difference between personality feedback and performance feedback?

Performance feedback names an outcome or an observable behaviour and its effect on the work, for example that a design review ran over because two decisions were reopened. Personality feedback describes who someone is, for example abrasive, too strong or judgmental. The first gives a person something specific to change and can be checked in calibration. The second can’t be checked, and it tends to land unevenly across groups.

How reliable is this study for a European mid-market company?

Treat it as a strong directional signal, not a benchmark. It is from 2014, it is US technology reviews, the reviews were volunteered rather than sampled, and most came from strong performers. Snyder herself says she can’t rule out that women were more willing to share reviews with critical language. The useful test is your own review text: read a sample of last cycle’s written feedback and count how much of it describes personality rather than work.

Is Revolut People available in the UK, Ireland, Benelux, the Nordics and the Baltics?

Yes. Revolut People is available to businesses with a registered entity in the UK, the EEA or the US, which covers Ireland, the Benelux, Nordic and Baltic countries.