Every performance rating is a judgment made by a person, and people are not measuring instruments. A manager who rates ten employees in one afternoon is recalling months of work from memory, weighing it against a scale that may be vague, and doing it while tired, rushed and aware of what the rating will mean for pay. Under those conditions, predictable errors creep in.
That matters because trust in reviews is already low. Deloitte's 2025 Global Human Capital Trends research found that 72% of workers and 61% of managers could not say they trust their organization's performance management process. Gallup reports that only 14% of employees strongly agree their performance reviews inspire them to improve. A review that feels unfair does more than fail to motivate: it can affect who gets promoted, who gets a raise and who decides to leave.
This guide covers the nine rating errors that show up most often, how to recognize each one in your own reviews, and what managers and HR teams can do about them. A note on method: we checked every source in November 2026 and link each figure to the page it comes from. The sources are university HR guidance, government sites and published research from Gallup, Deloitte, Stanford and Textio. This article does not quote any software prices. The practical checks and examples are our own.
What is performance review bias?
Performance review bias is any systematic error that makes a performance rating reflect something other than the employee's actual work during the review period, such as the rater's habits, first impressions, recent events or assumptions about the person.
Two kinds are worth separating:
- Rating errors are patterns in how a rater uses the scale or weighs evidence. Rating everyone a 3, or letting one great project lift every score, are rating errors. They often affect everyone a manager rates.
- Identity-based bias is when ratings or written feedback differ by gender, race, age or other characteristics in ways the work doesn't explain. This kind can also create legal risk, which we cover below.
Both are usually unintentional. Most managers who make these errors believe they are being fair. That is why the fixes below focus on process and evidence rather than on asking people to "be less biased."
The 9 most common rating errors
Dartmouth College's HR team publishes a clear set of definitions for common rater errors, and we use its wording where it helps. For each error, we add a quick self-check you can run on a draft review.
1. Halo effect
Dartmouth defines the halo effect as "the tendency to make inappropriate generalizations from one aspect of a person's job performance." An employee who closed a major deal gets high marks for teamwork, documentation and attendance too, even when the evidence for those is thin.
Self-check: For each high rating, can you name a specific example for that competency, not just for the person's best skill? If the same example supports three different ratings, halo is likely.
2. Horns effect
The horns effect is the reverse of halo: one weakness or one bad incident drags down ratings across the board. A missed deadline in March colors how a manager sees an employee's communication, judgment and initiative for the rest of the year.
Self-check: Look at your lowest ratings. Is the evidence for each one different, or does one event explain all of them?
3. Leniency
Leniency is "the tendency to evaluate all people as outstanding and to give inflated ratings," per Dartmouth. It often comes from wanting to avoid a hard conversation. The cost is that real high performers can't be told apart from everyone else, and weak performance goes unaddressed until it becomes a bigger problem.
Self-check: What share of your team is rated above "meets expectations"? If it's most of them, can you defend that to a peer who doesn't know your team?
4. Strictness (severity)
Strictness is the opposite: Dartmouth describes it as rating "all people at the low end of the scale" and being "overly critical of performance." Some managers believe nobody deserves the top rating, or hold their team to a higher bar than other managers use. Their people are then penalized in pay and promotion compared with peers doing similar work elsewhere.
Self-check: Compare your average rating with other managers' averages. If yours is clearly lower, check whether your team is genuinely weaker or your standard is different.
5. Central tendency
Central tendency is "the tendency to evaluate every person as average regardless of differences in performance." It's the safe choice: a middle rating rarely needs a long justification. It often signals that a manager doesn't know an employee's work well enough.
Self-check: Do your ratings use the full scale? If almost everyone sits in the middle, check whether you have enough evidence to separate them.
6. Recency bias
Recency bias is giving too much weight to the last few weeks before the review. A strong final quarter can hide a weak year, and a rough final month can bury eleven good ones. Annual cycles are the most exposed, because memory fades over a year. Gallup found that 47% of employees received feedback from their manager "a few times or less" in the past year, which suggests many managers have little documented evidence to draw on when review time comes.
Self-check: Do your examples come from the whole review period? Sort them by month. If most are from the last six weeks, go back to your notes, project records and one-on-one agendas.
7. First impression (primacy) error
Dartmouth describes this as a manager making "an initial favorable or unfavorable judgment about someone" and then ignoring later information that doesn't fit. It is common with new hires and with people who joined a team through a reorganization: the label formed in month one sticks.
Self-check: Would someone reading only the last six months of evidence reach the same rating?
8. Similar-to-me (affinity) bias
The similar-to-me effect is "the tendency to more favorably judge those people perceived as similar to the leader." Similarity can mean background, interests, working style, communication style or even the hours someone keeps. Because it feels like genuine rapport, it is one of the hardest errors to notice in yourself. It is also one of the ways identity-based bias enters ratings.
Self-check: Which employees do you find easiest to talk to? Are they also your highest rated? If so, test whether the work evidence alone supports the gap.
9. Contrast effect
The contrast effect is "the tendency for a rater to evaluate a person relative to other individuals rather than on-the-job requirements." A solid performer looks weak next to a star, or an average performer looks strong after a manager has just written up someone struggling. The order in which you write reviews can change the outcome.
Self-check: Rate each person against the written standard for their role and level, not against teammates. Writing reviews in a different order, or in two sittings, can expose contrast effects.
Quick reference table
| Rating error | What it looks like | Fastest fix |
|---|---|---|
| Halo | One strength lifts every rating | Require a separate example per competency |
| Horns | One weakness drags down every rating | Same as above |
| Leniency | Most of the team rated high | Compare distributions in calibration |
| Strictness | Most of the team rated low | Compare distributions in calibration |
| Central tendency | Everyone rated "meets expectations" | Use behavioral anchors; gather more evidence |
| Recency | Evidence comes from the last few weeks | Keep notes all year; review the full period |
| First impression | Early label never revisited | Weight recent evidence explicitly |
| Similar-to-me | People like the manager rated higher | Evidence-only justification; second reviewer |
| Contrast | Ratings shift with comparison or order | Rate against role standards, not peers |
Bias in the written feedback, not only the score
Ratings get most of the attention, but the comments that sit alongside them can carry bias too, and they shape how an employee is seen in promotion discussions later.
Researchers at Stanford, led by sociologist Shelley Correll, analyzed performance reviews at a Fortune 500 technology company, including both the language and the numeric ratings. Stanford Graduate School of Business summarizes their findings: the biggest biases appeared in evaluations of personality, and women's critical feedback "tends to be more vague, which turns out to be less helpful," while men's criticism was more often framed as developmental.
Textio, which makes writing software, published research in 2022 based on performance feedback for more than 25,000 people at 250 organizations. It reported that women received 22% more feedback about their personality than men, that people under 40 were described as "ambitious" 2.5 times as often as people 40 and older, and that Black people, Latinx people, workers over 40 and women received significantly less actionable feedback than their coworkers. Textio sells tools in this area, so treat it as vendor research, but the pattern matches the academic work.
Practical checks for written comments:
- Personality vs. behavior. Words like "abrasive," "nice," "aggressive" or "ambitious" describe a person, not their work. Replace them with what the person did and what effect it had.
- Vague vs. specific. "Needs to be more strategic" gives the employee nothing to act on. "In the Q3 planning document, the options weren't tied to budget impact; next time, include cost and risk for each" does.
- Length and effort. If some reviews are a paragraph and others are a page, ask why. Uneven effort is itself a form of unequal treatment.
- Same standard, same words. Would you use the same words for the same behavior if the employee were a different gender, age or background?
Our guide on how to give constructive feedback has more examples of turning vague comments into specific, actionable ones.
Why it matters: fairness, retention and legal risk
Bias in reviews compounds. Ratings feed into pay, bonuses and promotion lists, so a small, consistent skew for one group, repeated every cycle, becomes a large gap over several years.
There is also legal exposure. The U.S. Equal Employment Opportunity Commission states that the law "forbids discrimination in every aspect of employment," including job assignments, promotions and pay, based on protected characteristics such as race, color, religion, sex, national origin, age (40 or older), disability and genetic information. In the UK, the government lists nine protected characteristics, including age, disability, race, sex and sexual orientation, under the Equality Act. Performance ratings are often the paper trail behind pay and promotion decisions, so a consistent, documented, evidence-based process protects both employees and the organization. This is general information, not legal advice; check with employment counsel for your jurisdiction.
If ratings drive pay in your organization, a regular pay equity audit is the natural companion to a review-bias check: it shows whether rating differences are turning into pay differences.
How to reduce performance review bias: 8 steps
No process removes bias completely. The aim is to make errors less likely and easier to catch. These steps are ordered from the ones any manager can do alone to the ones that need HR and leadership support.
Step 1: Define what each rating means
Vague scales invite every error above. Replace labels like "good" or "strong" with behavioral anchors: short descriptions of what performance at each level looks like for a given role and competency. A competency-based review makes this much easier, because the standard is written down before anyone is rated.
Step 2: Collect evidence all year
The best defense against recency and first-impression errors is a running record. Keep brief notes after one-on-ones, project milestones and feedback conversations. A regular one-on-one meeting rhythm gives you a natural place to capture them. Many teams move to more frequent check-ins for this reason; our comparison of continuous feedback vs. annual reviews covers the trade-offs.
Step 3: Write the evidence before the rating
Ask managers to list specific examples for each competency first, then choose the rating the evidence supports. Deciding the number first and then finding examples to justify it is how halo and similar-to-me errors get locked in. A structured form helps: our performance review builder prompts for examples against each section so the evidence comes before the score.
Step 4: Add more perspectives
A single rater's blind spots go unchecked. Input from peers, cross-functional partners or direct reports adds evidence the manager doesn't see. A 360-degree feedback process does this formally; even one or two structured peer comments can balance a manager's view.
Step 5: Review your own drafts for patterns
Before submitting, run the self-checks above on your full set of reviews, not one at a time. Look at the spread of ratings, the time period your examples come from, and whether review length or tone differs across people.
Step 6: Calibrate across managers
Calibration is a meeting where managers compare proposed ratings and the evidence behind them, so that "exceeds expectations" means the same thing on every team. It is the most direct check on leniency, strictness and contrast effects, because outliers become visible. Our performance calibration guide explains how to set one up, and our calibration meetings guide covers facilitation in detail. Two cautions: calibration can itself push ratings toward the middle if no one challenges the clustering, and it can amplify bias if the loudest voice wins. A facilitator who asks "what's the evidence?" for every change keeps it honest.
There is evidence that clearer criteria help. The Stanford Clayman Institute for Gender Research reported early results from a mid-sized tech company that reduced ambiguity and set clearer evaluation criteria: the company saw "a positive net change for women in evaluations of their leadership skills" and "more balanced gender rankings of men and women as top performers" in calibration meetings.
Step 7: Analyze outcomes by group
HR should look at rating distributions by team, manager, gender, race, age band, part-time status and location, cycle after cycle. Differences don't prove bias, but they show where to look more closely. Pair the numbers with a sample read of written comments for personality language and vagueness. If you use a nine-box talent review, check the "potential" axis especially closely: the Stanford research found the biggest biases in judgments of personality and potential, which are harder to tie to evidence than results are.
Step 8: Train raters, then reinforce
A short session before each cycle that walks through these nine errors, with real (anonymized) examples from your own past reviews, is more useful than a generic annual course. Follow it with the tools that make good habits easy: anchored rating scales, evidence prompts in the review form and a calibration step. If you are introducing AI writing assistance into reviews, our guide to AI in performance reviews covers how to keep a human accountable for every rating.
A pre-submission checklist for managers
Copy this into your review process and run it before every cycle closes:
- Every rating has at least one specific example for that competency.
- Examples come from across the whole review period, not just the last quarter.
- My ratings use the scale as defined, and my distribution is defensible to a peer.
- No single event explains several different ratings.
- Written comments describe behavior and impact, not personality.
- Developmental feedback says what to do differently, not just what was wrong.
- Reviews are similar in length and effort across my team.
- I've rated each person against the role standard, not against teammates.
- I would use the same words for the same work from anyone on my team.
The bottom line
Performance review bias is mostly the product of process: vague scales, thin evidence, one rater and no comparison. Managers can catch many of the nine rating errors by gathering evidence all year, writing examples before choosing a score and checking their own patterns. HR can catch the rest with clear rating definitions, calibration and regular analysis of outcomes by group. For where reviews fit in the wider process, see our overview of the performance management cycle.
Frequently asked questions
What is the most common bias in performance reviews?
There is no single agreed ranking, but recency bias, leniency and central tendency are among the most frequently cited because they come from how annual reviews are run: evidence is recalled from memory at the end of the year, and middle or generous ratings need the least justification.
Can performance review bias be eliminated?
No. It can be reduced and made easier to catch. Clear rating definitions, year-round evidence, multiple perspectives and calibration across managers each remove some of the conditions that let errors through.
What is the difference between halo and horns effect?
Both let one aspect of performance spread to unrelated ratings. With the halo effect, one strength raises other ratings; with the horns effect, one weakness or incident lowers them.
Does calibration remove bias?
It helps, especially with leniency, strictness and contrast effects between managers. But calibration can introduce its own problems, such as pressure toward middle ratings or deference to senior voices. A facilitator, a shared rating standard and an "evidence for every change" rule make it more reliable.