A sales leader who closed one major deal in the final fortnight of the review period can receive a stronger rating than a colleague who delivered reliably for 11 months. Both managers may believe they are being fair. This is why performance review bias examples matter: bias is rarely announced or intentional. It enters through incomplete evidence, inconsistent standards and review workflows that allow opinion to become the rating.
For HR and people leaders, the issue is not simply whether a manager has bias. Everyone brings assumptions to judgement. The operational question is whether the review process exposes those assumptions, anchors assessments to the actual role, and makes unsupported ratings difficult to defend.
12 performance review bias examples that distort ratings
1. Halo effect
The halo effect occurs when one positive quality shapes the entire assessment. A confident presenter, well-liked colleague or high-profile project lead may be rated strongly across unrelated responsibilities, including areas where their delivery has been uneven.
Managers can counter this by rating each responsibility separately and requiring examples for material strengths. Strong communication should not automatically translate into a strong rating for planning, technical quality or people leadership.
2. Horns effect
The horns effect is the reverse. One mistake, conflict or weak capability lowers a manager's view of the employee's overall contribution. A missed deadline might become evidence that the employee is unreliable, even where their broader delivery record says otherwise.
This bias is particularly damaging when a manager records only exceptions. Review evidence should include the full period, the work context and the employee's response to setbacks.
3. Recency bias
Recency bias gives disproportionate weight to what happened most recently. It is common in annual reviews because managers often rely on memory rather than a contemporaneous record of work.
A late surge in output, or a visible mistake shortly before the review, should be assessed in proportion to the year. Regular check-ins, task-level evidence and defined review windows reduce the chance that December overrides February through October.
4. Primacy bias
Primacy bias gives too much influence to first impressions. An employee who struggled during onboarding may carry that label long after their capability has improved. Conversely, someone who made an excellent early impression can retain the benefit of it despite later gaps.
The control is simple but disciplined: ask managers to distinguish the employee's current performance from their first few months, and assess progress against the responsibilities that applied during the relevant period.
5. Similarity bias
Managers can rate people more favourably when they share their communication style, career background, working habits or approach to problem-solving. The employee who speaks up in meetings, works the same hours or has followed a familiar career path can feel more capable than a quieter peer with equivalent results.
Similarity is not evidence of performance. Role expectations should specify what was delivered, how quality is measured and which behaviours are genuinely required for the role.
6. Affinity bias
Affinity bias is preference driven by personal connection. It may arise from shared interests, schools, communities, family circumstances or social rapport. It can be subtle: a manager may offer more context, benefit of the doubt or developmental opportunities to people they know better.
This is one of the performance review bias examples that structured evidence can surface. If ratings differ sharply between comparable employees, HR should test whether the evidence and standards differ too.
7. Leniency and severity bias
Some managers rate nearly everyone highly. Others reserve top ratings so tightly that strong performers are compressed into the middle. Neither pattern necessarily reflects the team’s actual performance.
Forced distribution is not always the answer, particularly in small teams or unusual business periods. Calibration should instead compare rating patterns with role scope, outcomes, evidence quality and the standards used by peer managers. The aim is not identical distributions. It is consistent judgement.
8. Centrality bias
Centrality bias occurs when managers avoid decisive ratings and place most employees in the middle. It can look safe, but it obscures meaningful differences in performance and gives employees vague feedback.
Often the cause is not reluctance alone. Managers may lack confidence in the criteria, fear a difficult conversation or have no evidence framework. Clear rating definitions and prompted discussion points make a specific assessment more manageable.
9. Contrast bias
Contrast bias happens when an employee is assessed against the person reviewed immediately before them rather than against their own role requirements. A capable employee can appear average after an exceptional peer, while an average employee can appear stronger after a weak comparison.
Independent assessments reduce this risk. Each review should begin with the position description, agreed responsibilities and evidence of delivery, not an informal ranking in the manager’s head.
10. Attribution bias
Managers may explain their own team’s success through skill and effort, while attributing another employee’s success to luck, easy conditions or help from others. The same pattern can work against employees after setbacks: an external dependency becomes a personal failing.
Good reviews separate outcome from circumstance without excusing poor performance. Ask what was within the employee’s control, what constraints applied, what decisions they made and what evidence supports the conclusion.
11. Confirmation bias
Confirmation bias leads managers to look for information that supports an existing belief. If someone is labelled as not strategic, every operational detail can be read as proof. Evidence that challenges the label receives less attention.
Require managers to identify both supporting and disconfirming evidence for significant ratings. This is especially useful where a rating has changed markedly from the previous cycle or where manager and employee assessments diverge.
12. Visibility bias
Visibility bias rewards work that is easy to see. Employees in customer-facing, office-based or highly vocal roles can receive more recognition than colleagues whose work prevents issues, maintains systems or supports delivery behind the scenes.
Visibility is not the same as value. Managers should review the actual outputs and accountabilities of each position, including risk reduction, quality control, documentation, mentoring and operational continuity.
Build controls into the review workflow
Training managers to recognise bias helps, but training alone asks people to remember a long list of cognitive traps under time pressure. Better results come from process design.
Start with role-specific criteria. Generic competency statements leave too much room for interpretation. Position descriptions should be converted into clear responsibilities, expected outputs and observable standards. This gives employees and managers a shared reference point before the assessment begins.
Next, collect assessments independently. When an employee sees a manager’s comments before submitting their own view, they may adjust their account to match perceived expectations. When a manager sees the self-assessment first, it can anchor their judgement. A locked, blind submission process preserves two independent evidence sets and makes the gap between perspectives visible for discussion.
Then require evidence where ratings matter. A rating without an example is a preference, not a useful assessment. Evidence does not need to become burdensome documentation, but it should show what was delivered, the standard achieved, the timeframe and relevant constraints.
Calibration sessions should focus on outliers and reasoning, not just rating changes. HR can ask whether managers applied the same standard to comparable roles, whether the evidence supports the rating, and whether a pattern of generosity, severity or centrality is emerging. This preserves manager accountability while giving leaders a defensible basis for intervention.
CALIBRATE applies these controls through AI-assisted extraction of role responsibilities, structured self and manager assessments, and blind side-by-side comparison once both submissions are complete. The value is not automation for its own sake. It is a clearer record of how a rating was reached and where a manager conversation needs more evidence.
Fairness requires better evidence, not softer standards
Removing bias does not mean making every review positive or avoiding difficult ratings. It means ensuring a low rating is tied to relevant, documented performance expectations, and a high rating reflects sustained contribution rather than confidence, proximity or a recent win.
The most credible review process makes it easier for good managers to exercise judgement and harder for unsupported judgement to pass as fact. When employees can see that ratings follow the work, not the relationship, performance conversations become more precise and considerably easier to act on.