Performance reviews are supposed to measure output. In practice they are written quickly, in language managers reach for without thinking, and researchers who have collected thousands of them keep finding the same pattern. Certain words appear far more often in reviews of women and employees of color than in reviews of white men doing comparable work, and those words share a trait. They describe who someone is rather than what someone did. That distinction sounds academic until raise season, when a review full of personality commentary and short on measurable results is the one that loses the argument for promotion.
The first word is abrasive. Analyses of technology industry reviews have found it and its neighbors, harsh, difficult, grating, concentrated overwhelmingly in feedback given to women, often about the same behavior that earns a man a note about high standards. The second is aggressive, which follows Black employees across industries and shifts a description of directness into a description of threat. The third is emotional, a word that almost never appears with a specific incident attached and almost always appears as a summary judgment. None of the three tells a reader what happened, and none of them can be improved against. Ask any employee who has received one what they are supposed to do differently on Monday and watch the silence. That silence is the tell, because real feedback always has an action attached to it.
The fourth word is ambitious, which sounds like praise until you notice how it is used. Applied to a man it usually appears in a sentence about trajectory. Applied to a woman or to a younger employee of color it often appears with a qualifier, ambitious but needs to be patient, ambitious but should focus on the current role. The fifth is helpful. It is a genuine compliment and it is also the most common way a review damns with faint praise, describing someone as supportive, dependable, a great teammate, while the promotion case gets built on words like strategic, decisive, and drives results. Reviews weighted toward warmth and light on competence rarely survive a calibration meeting.
The sixth is potential, and it is the most consequential of the group. Promotion decisions frequently rest on whether someone is described as having demonstrated something or as having potential for it. Research on how the two are distributed has found that white men are more often promoted on projected potential while women and employees of color are more often required to show a completed track record first. When a review says a person has potential to lead, it sounds encouraging. What it usually means in practice is that leadership is being deferred to a later cycle for reasons that never get written down.
There is a seventh pattern that is not a single word, and it may be the most damaging. Vague feedback. Studies of manager behavior have found that some managers, worried about seeming biased or causing offense, soften criticism when the employee is from a different background, giving feedback that is less specific and less actionable. The employee leaves the meeting believing things are fine, gets no chance to correct anything, and hears the real assessment a year later when a promotion goes to someone else. Withheld specificity feels kind in the room and functions as sabotage over a career. Managers who do this are rarely acting in bad faith, which is exactly why it persists without anyone naming it. The remedy is to hold the same standard of directness for everyone rather than deciding privately who can handle it.
If you write reviews, the fix is a rule rather than a vocabulary list. Every critical statement gets a behavior, a date or a project, and an impact. Instead of noting that someone comes across as abrasive, describe the meeting, what was said, who reacted, and what it cost. If you cannot produce those three parts, the observation is not ready to be written down and probably reflects your reaction more than their performance. Then check the balance. If a review spends more lines on how a person communicates than on what they delivered, that is a signal worth sitting with before you send it.
If you receive one of these words, do not argue about the label, because the label is unfalsifiable and arguing confirms it in the reader's mind. Ask for the example instead. Which meeting, which decision, what would have been better, and what specifically would look different in six months. Ask in writing and summarize the answer in writing afterward, sent back to the manager for confirmation. That record does two things. It converts a vague impression into a concrete standard you can actually meet, and it creates documentation if the same unsupported language shows up again next cycle. Managers who meant well usually sharpen up considerably once they know the answer is being written down.
