Leadership
June 25, 2025
6
Min
Confirmation Bias in Performance Reviews: How to Avoid Rating Errors
Leadership
|
That was the opening remark from a divisional VP in a performance calibration session. We were reviewing a director who had launched two major initiatives that had both gone over budget and behind schedule. When someone raised concerns, the VP brushed them off: “Let’s not forget she saved our Q3 numbers last year. She’s earned our trust.”
The room nodded. No one challenged it further. The rating stayed high.
And just like that, confirmation bias won again.
Performance reviews are supposed to reflect reality. But confirmation bias - our tendency to seek, interpret, and remember information in ways that confirm pre-existing beliefs - quietly undermines this goal. Once we label someone as high-performing (or not), our brains begin to filter evidence to fit that narrative.
This matters because inaccurate reviews corrode accountability, distort development plans, and quietly demotivate top contributors.
In a meta-analysis published in Personnel Psychology, researchers found that even trained managers exhibited strong confirmation bias during evaluations, often ignoring disconfirming data once an initial impression was formed. The cost isn’t just reputational - it’s organisational. Promotions, raises, and retention all hinge on flawed signals.
We’ve seen it repeatedly: a high-potential leader leaves, frustrated that despite delivering results, they were “still seen as the guy from that project two years ago.” Or a struggling team member continues to coast on old praise while their performance flatlines.
It’s not about bad intent. It’s about blind spots.
So how do we fix this?
Let’s be clear - we’re not aiming for utopia. No system will be completely bias-free. But we can make it fairer, more disciplined, and anchored in evidence. Here’s the model we use with executive teams:
Before reviews begin, we ask: “What might we be wrong about?” This question pauses the brain’s auto-pilot.
This matters most for long-tenured team members, rising stars, and chronic underperformers. The longer the history, the stronger the story we’ve built.
Reflection Prompt:
Who on your team might be getting too much benefit of the doubt - or too little?
Micro-action: Ask one skip-level stakeholder for input before forming your rating. It disrupts your mental loop.
Comments like “She’s not strategic” or “He lacks initiative” are judgments, not data. Instead, we ask:
This subtle shift forces reviewers to cite observable facts rather than character traits. It also arms reviewees with concrete feedback they can actually act on.
Mini-exercise:
Take a piece of feedback you’ve written this cycle. Is it behaviour-based or perception-based? Rewrite it with specifics.
Micro-action: In your review form, add a column: “Observation or interpretation?” Tag each comment.
No one sees the full picture. Managers see effort; peers see collaboration; customers see outcomes. Use all three.
By triangulating, you weaken the influence of any one viewpoint - especially your own.
Pro Tip: Ask each reviewer to cite at least two non-managerial data points.
Micro-action: Build a “feedback dossier” for each person, updated quarterly. It beats a last-minute memory scrape.
If everyone agrees too easily in calibration meetings, it’s not consensus - it’s avoidance. The goal isn’t harmony. It’s intellectual honesty.
Great calibration sessions sound like this:
Psychological safety plays a role here. But so does structured dissent.
Micro-action: Assign a “bias challenger” role in each calibration meeting. Their only job: poke holes.
The model works best when it’s operationalised, not just aspirational. Here’s how teams can start:
Pro Tip: Create “rating guides” with clear definitions and examples. It aligns expectations and reduces subjectivity.
Fix: Build a bias radar - a short reminder slide before any review process begins. Keep it visible.
Prompt 1: When was the last time someone surprised you in a review - positively or negatively? What allowed that to happen?
Prompt 2: How might your past labels be distorting present assessments? Take 5 minutes to jot down three people and reassess.
Getting this right pays off. Teams see:
Importantly, it tells your people: “We see you clearly. And we’re willing to adjust when the data tells us to.”
Because the opposite - sticking to outdated stories - costs more than we admit.
This week, choose one high-visibility review and put it through the 4-part model. Don’t aim for perfection. Just test the discipline.
And if you’re up for a deeper dive, we’re happy to share templates or workshop this with your leadership team.
Team SHIFT
“She’s a star, always has been.”
That was the opening remark from a divisional VP in a performance calibration session. We were reviewing a director who had launched two major initiatives that had both gone over budget and behind schedule. When someone raised concerns, the VP brushed them off: “Let’s not forget she saved our Q3 numbers last year. She’s earned our trust.”
The room nodded. No one challenged it further. The rating stayed high.
And just like that, confirmation bias won again.
Performance reviews are supposed to reflect reality. But confirmation bias - our tendency to seek, interpret, and remember information in ways that confirm pre-existing beliefs - quietly undermines this goal. Once we label someone as high-performing (or not), our brains begin to filter evidence to fit that narrative.
This matters because inaccurate reviews corrode accountability, distort development plans, and quietly demotivate top contributors.
In a meta-analysis published in Personnel Psychology, researchers found that even trained managers exhibited strong confirmation bias during evaluations, often ignoring disconfirming data once an initial impression was formed. The cost isn’t just reputational - it’s organisational. Promotions, raises, and retention all hinge on flawed signals.
We’ve seen it repeatedly: a high-potential leader leaves, frustrated that despite delivering results, they were “still seen as the guy from that project two years ago.” Or a struggling team member continues to coast on old praise while their performance flatlines.
It’s not about bad intent. It’s about blind spots.
So how do we fix this?
Let’s be clear - we’re not aiming for utopia. No system will be completely bias-free. But we can make it fairer, more disciplined, and anchored in evidence. Here’s the model we use with executive teams:
Before reviews begin, we ask: “What might we be wrong about?” This question pauses the brain’s auto-pilot.
This matters most for long-tenured team members, rising stars, and chronic underperformers. The longer the history, the stronger the story we’ve built.
Reflection Prompt:
Who on your team might be getting too much benefit of the doubt - or too little?
Micro-action: Ask one skip-level stakeholder for input before forming your rating. It disrupts your mental loop.
Comments like “She’s not strategic” or “He lacks initiative” are judgments, not data. Instead, we ask:
This subtle shift forces reviewers to cite observable facts rather than character traits. It also arms reviewees with concrete feedback they can actually act on.
Mini-exercise:
Take a piece of feedback you’ve written this cycle. Is it behaviour-based or perception-based? Rewrite it with specifics.
Micro-action: In your review form, add a column: “Observation or interpretation?” Tag each comment.
No one sees the full picture. Managers see effort; peers see collaboration; customers see outcomes. Use all three.
By triangulating, you weaken the influence of any one viewpoint - especially your own.
Pro Tip: Ask each reviewer to cite at least two non-managerial data points.
Micro-action: Build a “feedback dossier” for each person, updated quarterly. It beats a last-minute memory scrape.
If everyone agrees too easily in calibration meetings, it’s not consensus - it’s avoidance. The goal isn’t harmony. It’s intellectual honesty.
Great calibration sessions sound like this:
Psychological safety plays a role here. But so does structured dissent.
Micro-action: Assign a “bias challenger” role in each calibration meeting. Their only job: poke holes.
The model works best when it’s operationalised, not just aspirational. Here’s how teams can start:
Pro Tip: Create “rating guides” with clear definitions and examples. It aligns expectations and reduces subjectivity.
Fix: Build a bias radar - a short reminder slide before any review process begins. Keep it visible.
Prompt 1: When was the last time someone surprised you in a review - positively or negatively? What allowed that to happen?
Prompt 2: How might your past labels be distorting present assessments? Take 5 minutes to jot down three people and reassess.
Getting this right pays off. Teams see:
Importantly, it tells your people: “We see you clearly. And we’re willing to adjust when the data tells us to.”
Because the opposite - sticking to outdated stories - costs more than we admit.
This week, choose one high-visibility review and put it through the 4-part model. Don’t aim for perfection. Just test the discipline.
And if you’re up for a deeper dive, we’re happy to share templates or workshop this with your leadership team.
Team SHIFT
“She’s a star, always has been.”
That was the opening remark from a divisional VP in a performance calibration session. We were reviewing a director who had launched two major initiatives that had both gone over budget and behind schedule. When someone raised concerns, the VP brushed them off: “Let’s not forget she saved our Q3 numbers last year. She’s earned our trust.”
The room nodded. No one challenged it further. The rating stayed high.
And just like that, confirmation bias won again.
Performance reviews are supposed to reflect reality. But confirmation bias - our tendency to seek, interpret, and remember information in ways that confirm pre-existing beliefs - quietly undermines this goal. Once we label someone as high-performing (or not), our brains begin to filter evidence to fit that narrative.
This matters because inaccurate reviews corrode accountability, distort development plans, and quietly demotivate top contributors.
In a meta-analysis published in Personnel Psychology, researchers found that even trained managers exhibited strong confirmation bias during evaluations, often ignoring disconfirming data once an initial impression was formed. The cost isn’t just reputational - it’s organisational. Promotions, raises, and retention all hinge on flawed signals.
We’ve seen it repeatedly: a high-potential leader leaves, frustrated that despite delivering results, they were “still seen as the guy from that project two years ago.” Or a struggling team member continues to coast on old praise while their performance flatlines.
It’s not about bad intent. It’s about blind spots.
So how do we fix this?
Let’s be clear - we’re not aiming for utopia. No system will be completely bias-free. But we can make it fairer, more disciplined, and anchored in evidence. Here’s the model we use with executive teams:
Before reviews begin, we ask: “What might we be wrong about?” This question pauses the brain’s auto-pilot.
This matters most for long-tenured team members, rising stars, and chronic underperformers. The longer the history, the stronger the story we’ve built.
Reflection Prompt:
Who on your team might be getting too much benefit of the doubt - or too little?
Micro-action: Ask one skip-level stakeholder for input before forming your rating. It disrupts your mental loop.
Comments like “She’s not strategic” or “He lacks initiative” are judgments, not data. Instead, we ask:
This subtle shift forces reviewers to cite observable facts rather than character traits. It also arms reviewees with concrete feedback they can actually act on.
Mini-exercise:
Take a piece of feedback you’ve written this cycle. Is it behaviour-based or perception-based? Rewrite it with specifics.
Micro-action: In your review form, add a column: “Observation or interpretation?” Tag each comment.
No one sees the full picture. Managers see effort; peers see collaboration; customers see outcomes. Use all three.
By triangulating, you weaken the influence of any one viewpoint - especially your own.
Pro Tip: Ask each reviewer to cite at least two non-managerial data points.
Micro-action: Build a “feedback dossier” for each person, updated quarterly. It beats a last-minute memory scrape.
If everyone agrees too easily in calibration meetings, it’s not consensus - it’s avoidance. The goal isn’t harmony. It’s intellectual honesty.
Great calibration sessions sound like this:
Psychological safety plays a role here. But so does structured dissent.
Micro-action: Assign a “bias challenger” role in each calibration meeting. Their only job: poke holes.
The model works best when it’s operationalised, not just aspirational. Here’s how teams can start:
Pro Tip: Create “rating guides” with clear definitions and examples. It aligns expectations and reduces subjectivity.
Fix: Build a bias radar - a short reminder slide before any review process begins. Keep it visible.
Prompt 1: When was the last time someone surprised you in a review - positively or negatively? What allowed that to happen?
Prompt 2: How might your past labels be distorting present assessments? Take 5 minutes to jot down three people and reassess.
Getting this right pays off. Teams see:
Importantly, it tells your people: “We see you clearly. And we’re willing to adjust when the data tells us to.”
Because the opposite - sticking to outdated stories - costs more than we admit.
This week, choose one high-visibility review and put it through the 4-part model. Don’t aim for perfection. Just test the discipline.
And if you’re up for a deeper dive, we’re happy to share templates or workshop this with your leadership team.
Team SHIFT