Performance reviews are a moment when reviewers feel like a traffic police officer.
They get this warrant to be authoritative. To be a counselor. Someone worthy of giving guidance. The process grants them a feeling of heightened self-esteem. They feel a dopamine rush the same way a police officer feels giving you a ticket.
The officer isn’t thinking about road safety. Not really. The act of writing the citation, the slow walk back to your window, the handing over of the slip of paper with your name on it. That’s authority made physical. That’s power confirming itself.
Performance reviews operate on the same circuitry.
607 Studies and a Coin Flip
I was curious whether my gut instinct had any empirical backing. So I looked at it.
The most comprehensive study ever conducted on feedback interventions was Kluger and DeNisi’s 1996 meta-analysis. They examined 607 effect sizes across 23,663 observations. The finding that should have ended the entire performance review industry: over one-third of feedback interventions actually decreased performance. Not maintained it. Not had no effect. Actively made things worse.
Consider what that means in practice. No pharmaceutical would receive FDA approval with a 33% adverse effect rate. No surgeon would perform an elective procedure that made patients sicker one out of every three times. No airline would fly planes that crashed on a third of their routes.
But organizations conduct millions of performance reviews annually, knowing that statistically, a third of them are doing damage.
The remaining two-thirds doesn’t mean “working” either. The researchers calculated something called an effect size, labeled “d,” which measures how much difference an intervention actually makes. A d of 0.8 is large enough that anyone would notice. A d of 0.5 is medium, visible if you’re paying attention. A d of 0.2 is small, barely detectable.
Performance reviews scored a d of 0.41. Small to medium. Squint and you might see it.
To put that in tangible terms: imagine two groups of employees. One group receives performance reviews. The other receives nothing. At the end of the year, you line them all up by productivity and try to guess who got the reviews. With a d of 0.41, you’d guess correctly about 61% of the time. Eleven points better than flipping a coin.
And remember, that 0.41 is an average. It includes the one-third that got worse, which drags the number down. It includes a large middle chunk where nothing measurable happened at all. The modest gains of a few high performers are carrying the entire statistic. Most employees who receive reviews look indistinguishable from employees who don’t.
You also can’t predict which third will be harmed. No diagnostic tells you in advance whether this particular review, with this particular manager and this particular employee, will help or hurt. Every review is a pull on a slot machine where one-third of the outcomes cost you performance instead of improving it.
The researchers couldn’t explain the negative effects with sampling error. They couldn’t explain it with whether feedback was positive or negative. Existing theories offered nothing. After examining all that data, they concluded that something fundamental about the feedback intervention itself was causing harm in ways that existing models couldn’t account for.
If a training program made employees worse at their jobs a third of the time, it would be canceled. If a software implementation degraded productivity for a third of users, it would be rolled back. Performance reviews get a pass because we’ve categorized them as necessary infrastructure rather than as an intervention with measurable outcomes.
They’re not infrastructure. They’re a bet. And the house edge is worse than most organizations realize.
One of the largest studies on performance ratings found that more than half of the variance in ratings had more to do with the quirks of the person giving the rating than the person being rated. Rater bias was the biggest predictor. It held more weight than actual performance, the dimension being measured, or even measurement error.
The review tells you more about the reviewer than the reviewee.
Only 14% of employees strongly agree that their performance reviews inspire them to improve. Deloitte calculated their process consumed 1.8 million hours across the firm each year.
That’s a tax. A performance tax. Organizations pay it because the ritual feels like management, even when the evidence says otherwise.
The Brain Doing What Brains Do
Lisa Feldman Barrett is a neuroscientist at Northeastern University who has spent decades studying how humans construct emotions. Her research on affective realism explains something important about performance reviews.
Affective realism is the phenomenon where your internal state becomes evidence about the external world. You feel negative, so you conclude something bad is happening. You feel threatened, so the person across the table must be threatening you. Your feelings become facts. I wrote extensively on Affective Realism in another article:
Barrett’s research demonstrated that affect changes how participants literally see neutral stimuli. In experiments using interocular suppression, researchers showed that neutral faces were perceived as more smiling when paired with affectively positive images the participants couldn’t consciously see. The brain constructs perception from prediction, and those predictions are saturated with internal state.
Now place a manager in a conference room. They had a bad commute. Their kid was difficult that morning. They’re worried about their own upcoming review. They sit down across from an employee whose performance they’re about to “objectively” evaluate.
The manager’s interoceptive state is constructing what they perceive. Their affect is not separate from their assessment. It is their assessment. The employee’s responses, body language, tone of voice are all being filtered through a predictive engine that has already decided how to feel.
This is the architecture of human cognition. The brain runs on predictions to save energy and time. Affective realism lets us act quickly without verifying every sensory input against reality. The price we pay is that our feelings masquerade as objectivity.
Ink on Wet Paper
Performance reviews are a festival of cognitive errors.
The halo effect: higher ability on one performance dimension leads raters to score higher on completely unrelated dimensions. The employee is charming, so they must also be competent. They’re punctual, so their strategic thinking must be strong. One trait bleeds across the entire evaluation like ink on wet paper.
The horn effect: the inverse. One negative quality shadows everything else. A single missed deadline becomes evidence of broader character failure.
Recency bias: what happened last month overwhelms what happened for the previous eleven. The review is technically annual but functionally measures a few weeks.
Similar-to-me bias: managers rate employees who resemble them more favorably. Same background, same communication style, same hobbies. Yale researchers found that when you first agree to the criteria used in an assessment and then make the evaluation, you’re less likely to rely on stereotypes. Most organizations don’t structure reviews this way.
Idiosyncratic rater bias: managers evaluate skills they’re not good at highly and rate others lower for skills they excel at. The review becomes a projection of the manager’s own psychological landscape.
Stack these biases together. Add affective realism. Then ask whether the resulting document tells you anything reliable about an employee’s actual contribution to the organization.
Power, Money, Theater
Three forces keep the performance review alive.
The first is psychological. Feeling powerful leads to higher self-esteem. Performance reviews grant temporary power. For thirty minutes or an hour, the manager holds someone’s career in their hands. They get to pronounce judgment. They get to decide what gets written into the permanent record. The dopamine hit is real even when the performance improvement is not.
The second force is economic. Performance reviews have evolved into cost control machinery.
Jack Welch’s “vitality curve” at GE made this explicit. Employees were forced into a 20-70-10 distribution: top 20% rewarded, middle 70% maintained, bottom 10% fired. The system guaranteed that no matter how well a team performed, someone had to be labeled a failure. The curve didn’t measure performance. It manufactured scarcity.
Stack ranking, forced distribution, rank and yank. The names vary. The function is identical. These systems exist because they control budget. You always know exactly how much money is being given out in raises when only a fixed percentage can receive top ratings. The review becomes a rationing mechanism disguised as an evaluation.
Someone always has to be a 5 out of 5. Someone always has to be a 1. The bell curve demands it. An exceptionally talented team still has to identify a bottom 10%, even when that 10% would be top performers anywhere else. Microsoft employees avoided working with other talented developers because being on a strong team meant a higher chance of being ranked low relative to peers. The smart play was to surround yourself with mediocrity.
Calibration sessions make the cost control function even more visible. Managers sit in rooms arguing over which employees get slotted into which boxes. They come armed with paper trails to defend their people and incriminate members of competing groups. The process resembles a court case more than a development conversation. The outcome is predetermined by the curve, not by actual performance.
When a company needs to reduce headcount, performance reviews provide legal cover. The documentation exists. The ratings exist. The terminations look objective even when they’re not.
The third force is theatrical. Performance reviews look like management. They feel like leadership. Organizations keep them because discontinuing them feels irresponsible, like removing a load-bearing wall. The absence of a formal process seems more dangerous than a formal process that doesn’t work.
But the wall isn’t load-bearing. Adobe removed it and voluntary attrition dropped 30%. Microsoft removed stack ranking in 2013 and innovation returned. GE itself abandoned the vitality curve in the mid-2010s. The companies that pioneered these systems eventually recognized what they had built: not a performance engine, but a cost control mechanism that happened to destroy collaboration and trust along the way.
The ritual persists because it serves managers psychologically, serves finance departments economically, and serves organizations theatrically. What it doesn’t serve is actual performance improvement. That was never the point.
Ego Instead of Task
Kluger and DeNisi proposed a theory to explain why feedback so often fails. They argued that feedback effectiveness decreases as attention moves from the task toward the self.
Three levels: task learning (how to do the work better), task motivation (why to do the work), and meta-tasks (including self-related processes like protecting ego and managing image).
When feedback pushes attention up the hierarchy toward the self and away from the task, performance drops. Annual reviews are designed to make people think about themselves. Am I good enough? Am I valued? Will I be promoted? Will I be fired? The conversation structurally directs attention to exactly the wrong place.
Complex tasks suffer most. When work is difficult or unfamiliar, feedback that makes people focus on image protection drains the cognitive resources needed for actual improvement. The employee spends their mental energy worrying about how they look instead of how to get better.
The review creates the conditions for its own failure.
“But You Need Measurement”
The objection arrives on schedule. You can’t run an organization without accountability. Not everything can be chaos. You need to measure progress. You need to manage systems at scale.
The objection misunderstands the critique.
The problem isn’t measurement. The problem is when the measure becomes the goal.
Charles Goodhart, a British economist, articulated this in 1975: “When a measure becomes a target, it ceases to be a good measure.” Donald Campbell, a psychologist, arrived at the same insight independently: “The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.”
Campbell was specifically writing about educational testing. His observation: “Achievement tests may well be valuable indicators of general school achievement under conditions of normal teaching aimed at general competence. But when test scores become the goal of the teaching process, they both lose their value as indicators of educational status and distort the educational process in undesirable ways.”
Replace “test scores” with “performance ratings” and the sentence describes exactly what happens in corporate performance management.
The annual review transforms a diagnostic into a target. Employees optimize for the rating, not the work. Managers optimize for the process, not development. The measure corrupts the thing it was supposed to measure.
This is the difference between measurement as means and measurement as end. A thermometer tells you the temperature. It doesn’t make the room warmer. Performance reviews pretend to do both, and accomplish neither.
The question isn’t whether to measure. The question is whether your measurement system is actually producing information about performance, or whether it’s producing information about people’s ability to navigate the measurement system itself.
What Happened When They Stopped
In 2012, Adobe abandoned annual performance reviews entirely. Their HR leader, Donna Morris, had grown frustrated watching voluntary attrition spike every February as employees reacted to disappointing reviews by leaving. The company was spending 80,000 manager hours annually on a process that created, in her words, “no discernible value.”
They replaced it with something called Check-in. Lightweight. Flexible. No ratings. No rankings. No forced distribution curves. Quarterly conversations about goals, feedback, and development. Sessions called by employees, not mandated by HR. Compensation decisions decoupled from the feedback conversation entirely.
The results: voluntary attrition dropped by 30%. The company saved 80,000 hours in the first year alone.
Cisco saw a 10% increase in employee satisfaction after eliminating annual reviews. Cargill reported that 70% of employees felt valued after shifting to ongoing conversations, compared to the wasteland of their previous system. IBM reported a 50% improvement in engagement scores after moving to continuous check-ins.
The pattern across all these cases: feedback that happens close to the work, focused on the task, decoupled from judgment about the person’s worth.
Frequent feedback works because it keeps attention on the task level of the hierarchy Kluger and DeNisi identified. When feedback arrives 72 hours after an event instead of 11 months later, it’s still about the work. The employee can actually adjust. The manager can reinforce or correct in real time. Neither party has to reconstruct a year’s worth of memory to have the conversation.
The research supports this. Companies that review goals monthly are more likely to be in the top financial performance quartile. Organizations that manage objectives quarterly generate higher returns than those addressing them annually. Frequent feedback has been shown to boost performance by up to 39%.
Real-time information allows real-time adjustment. Delayed information allows only retrospective judgment.
The One-on-One
At this point, someone suggests the solution is more one-on-ones. Weekly check-ins. Regular facetime. The research shows engagement improves when managers meet frequently with their reports.
Partly true. Mostly misleading.
Nearly half of direct reports rate their one-on-ones as suboptimal according to Harvard Business Review research. Organizations seldom provide managers with training for these conversations. And the telling detail: leaders consistently believe these meetings are successful while their direct reports hold less favorable views.
The gap between manager perception and employee experience is itself a symptom of the same disease. The manager feels good about the meeting because the meeting serves the manager’s needs. The employee walks away wondering what the point was.
One-on-ones fail for the same reasons performance reviews fail. They center the manager’s agenda. They become status updates disguised as development conversations. They create the appearance of engagement without the substance.
When managers skip one-on-ones entirely, only 15% of employees stay engaged. But the inverse isn’t automatically true. Holding the meeting doesn’t guarantee engagement. The meeting has to actually serve the employee’s needs, not just check a box on the manager’s calendar.
Nearly 60% of managers responsible for one or two employees have no formal training whatsoever. The same is true for over 40% of managers overseeing three to five employees. We ask untrained people to have skilled conversations, then wonder why the conversations don’t produce results.
The one-on-one becomes another ritual. Another performance of management. Another opportunity for the traffic police effect to take hold.
Five Principles
The organizations that have successfully replaced traditional reviews share several principles.
First, they separate feedback from judgment. The conversation about how to improve the work is different from the conversation about compensation and advancement. Mixing them poisons both. When an employee knows the feedback session determines their bonus, they stop listening for development and start managing their image.
Second, they make feedback proximate. The closer feedback arrives to the event, the more useful it is. Seventy-two hours is the outer bound of effectiveness. Eleven months is theater.
Third, they shift ownership. Adobe’s Check-in sessions are called by employees, not mandated by managers. This inverts the power dynamic. The employee identifies what they need. The manager provides support. The agenda belongs to the person who has to do the work.
Fourth, they emphasize narrative over numbers. Quantitative ratings are easier to game and easier to bias. Narrative assessment requires multiple observations, resists reduction to a single score, and captures nuance that numbers cannot.
Fifth, they train managers. Most organizations assume that promoting someone to a management role automatically confers the skills to develop other people. It doesn’t. The skills are learnable, but they have to be taught. Adobe built an entire Employee Resource Center with templates, videos, and training to help managers learn how to give useful feedback. Most companies just throw people into the role and hope.
The alternative isn’t no measurement. It’s measurement that actually measures something, feedback that actually helps, and conversations that actually serve the person doing the work.
Organizations treat performance reviews as infrastructure. Something necessary. Something that obviously works because everyone does it.
The science says otherwise. One-third of feedback interventions make performance worse. More than half of what shows up in ratings reflects the rater, not the ratee. The process costs millions and consumes millions of hours. The ritual satisfies the psychological needs of managers, the budgetary needs of finance departments, and the theatrical needs of organizations. What it doesn’t satisfy is actual performance improvement.
Goodhart’s Law explains why: the measure became the target, and once that happened, it stopped measuring anything real. Campbell’s Law explains the consequence: the more important the target became, the more it corrupted the behavior it was supposed to evaluate.
Performance reviews are a tax on organizational energy. We pay it because stopping feels irresponsible, even when continuing produces no returns.
Every manager sitting down to write a review is a person constructing reality from their own internal state, stacking biases they can’t see, directing attention toward ego instead of improvement. They feel like a traffic police. They get the dopamine hit. The employee gets a document that tells them almost nothing true about their actual work.
That’s the performance tax. We’ve been paying it for decades.
The receipt just says “management.”