Performance reviews: why they change nothing, and how to split them in two

A performance review changes nothing because one conversation is asked to do two incompatible jobs: to judge last year, with consequences attached to pay, and to open the subject of what someone cannot yet do. Judgement always wins, because judgement has consequences. Development becomes decoration. The fix is not to drop the review, but to separate it.

The meeting runs an hour. In the first half, a manager walks through a form and lands on a rating that will feed the compensation round. In the second, the same manager asks, in good faith, "so where do you feel you need to grow?"

Nobody answers that honestly. Not because people are evasive, but because they have just watched, live, what happens to things said in this room: they become a number. Someone handed a rating at minute 20 will not describe their gaps at minute 40. They will describe the gaps they can afford to name.

The usual conclusion is that managers are bad at the conversation, so the organization buys feedback training. Sometimes that helps. More often not, because the problem is not how the conversation is held. It is that it has been given two jobs that cancel each other out.

TL;DR

• An evaluation asks someone to defend their work; a development conversation asks them to expose what they cannot do. Put both in the same hour and the first drives out the second, because only one affects pay.
• Feedback is not automatically useful. In Avraham Kluger and Angelo DeNisi's meta-analysis of 607 effect sizes across 23,663 observations, interventions improved performance on average, but over a third reduced it.
• The two clocks: evaluation runs on an annual or semi-annual clock, looks backward, and carries consequences. Development runs on a fortnightly clock, looks forward, and carries none.
• The non-contamination rule: what is said on the development clock never becomes evidence on the evaluation clock. Write it down, tell the team, hold to it. Broken once, the development conversation converts permanently into another review.
• Only 14% of employees strongly agree that the reviews they receive inspire them to improve, and 48% say they are reviewed once a year (Wigert and Mann, Gallup, international data).
• Whether to scrap ratings altogether is still contested inside industrial-organizational psychology itself, with serious arguments on both sides (Adler and colleagues, 2016). Separating the two functions does not require you to pick a side.

Why do performance reviews fail to change anything?

Because they fuse a judicial function with a growth function. Evaluation decides retrospectively how the year went and carries consequences. Development asks someone to name what they cannot yet do. The second cannot happen honestly in the same hour as the first, however well the meeting is run.

The two conversations make opposite demands.

An evaluation looks backward, compares people against one another, produces a document and carries administrative consequences. The rational posture for an employee facing one is to defend their work, because that is precisely what will happen to the conclusions: they will be used.

A development conversation looks forward, compares a person only against themselves, produces no document, and requires them to name things they have not mastered. The rational posture is to expose.

There is no way to run a meeting that extracts both defence and exposure from the same person, in the same hour, about the same facts. Ask for both and you get the protective behaviour: the employee frames the year favourably and names a gap small enough to sound self-aware without costing anything. The form gets completed, and nothing developable has been said out loud.

This is also why feedback training produces mixed results. It improves how the message is delivered but does not touch the structure: a well-trained manager holding both conversations in one hour still gets defence, delivered more gracefully. It is the architectural error described in why training programs fail.

The practice data supports the diagnosis. Ben Wigert and Annamarie Mann report, drawing on Gallup research, that only 14% of employees strongly agree the reviews they receive inspire them to improve, and 48% say they are reviewed once a year (Wigert and Mann, "Give Performance Reviews That Actually Inspire Employees", Gallup, 2017, updated 2026). These are international figures rather than Romanian ones, but they describe an instrument the people receiving it do not find motivating.

What does the research on feedback actually show?

That feedback is not automatically beneficial. Kluger and DeNisi analysed 607 effect sizes across 23,663 observations and found an average improvement in performance, but over a third of feedback interventions reduced it. The explanation they proposed turns on where feedback sends the recipient's attention.

The silent assumption inside any appraisal system is that people improve when told how they are doing. It has been tested, and the result is less comfortable than you would expect.

Avraham Kluger and Angelo DeNisi published a meta-analysis of feedback interventions in 1996, covering 607 effect sizes across 23,663 observations. The average effect is positive. But over a third of the interventions reduced the performance of the people receiving feedback, and that result was not explained by sampling error or by feedback sign, meaning whether the feedback was positive or negative (Kluger and DeNisi, Psychological Bulletin, 119(2), 254-284, 1996).

The theory they built to account for that spread, Feedback Intervention Theory, proposes that what matters is where feedback moves attention. When it stays on the task, effects tend to be good; when it moves to the self, effects degrade. This is a theory offered to explain those data rather than a law, and the research base is largely North American.

The practical consequence is hard to ignore. A rating attached to pay is built to move attention to the self: it is comparative, durable and administratively visible. An appraisal system that ties every conversation about growth to a moment like that has bought exactly the condition Kluger and DeNisi associate with weak effects.

It is worth being explicit about what this study does not support. It does not show that performance reviews reduce performance. Kluger and DeNisi studied feedback interventions in general, not annual appraisal meetings. The defensible conclusion is narrower: "I told them where they were going wrong" is not evidence that anyone will improve, and the design of the conversation decides more than its existence does.

The two clocks: how to separate evaluation from development

Evaluation stays on its annual or semi-annual rhythm, with the single job of being fair and documented. Development gets a clock of its own, fortnightly, with no administrative consequences, about the work immediately ahead. Between them sits one rule: what is said on the development clock is never used in the review.

We call it the two clocks, because the difference is not content but rhythm and consequence.

The evaluation clock. Once or twice a year. Looks backward. Carries consequences: pay, promotion, role. It has one job, to be fair and accurate, and Wigert and Mann show that current practice struggles even there, with 29% of employees strongly agreeing their review is fair. What this clock may not do is pretend to develop anybody.

The development clock. Every two weeks, twenty minutes, with the direct manager, about a piece of work in the next fortnight's calendar. It is not a career conversation and produces no document. We choose that interval as a design decision rather than a demonstrated effect: the conversation has to reach the work before the work is over, and an annual cycle reaches it afterwards.

The non-contamination rule. What is said on the development clock does not become evidence on the evaluation clock. The manager does not quote into the appraisal form what someone admitted in a one-to-one, and does not build a rating on those conversations. The rule gets written down, communicated to the team, and visibly honoured.

The third element cannot be improvised, and in the rollouts we see it is the one missing. It explains why so many continuous-feedback programmes produce nothing: they added frequency without separation, so they multiplied the occasions on which someone calculates their answer.

The rule has an unpleasant property: it breaks on first violation, and for the whole team. A single quote from a one-to-one surfacing in an appraisal form gets around, and converts the development clock into another review.

The failure pattern recurs often, and what follows is a composite of it rather than any single client. A company introduces monthly one-to-ones, with good participation at first. In the fourth quarter, the appraisal system asks managers to support their ratings with concrete examples. Managers take them from the one-to-ones, where the substantial conversations happened. The following year participation holds and the content empties out. Nobody broke a rule, because no rule had been written.

For the structure of the development conversation, the development contract sets out the four lines that make it checkable. Raising a correction early, without waiting for the form, is the Performance Correction cluster in Accountability.

Do you have to get rid of ratings?

Not necessarily, and it is not a settled question. In 2016, seven industrial-organizational psychologists published a structured debate on exactly this, with strong arguments on both sides. Separating evaluation from development does not depend on the answer: it works with ratings and without them.

The wave of large companies dropping ratings created an impression that the question had closed. It has not. Seymour Adler, Michael Campion, Alan Colquitt, Amy Grubb, Kevin Murphy, Rob Ollander-Krane and Elaine Pulakos published a debate in which half the authors argue for eliminating performance ratings and half for retaining them, growing out of a session held at the 2015 SIOP conference (Adler and colleagues, Industrial and Organizational Psychology, 9(2), 219-252, 2016).

The case for elimination rests on measurement error, process cost and demotivating effects. The case for retention rests on the need for defensible personnel decisions and on the observation that, without an explicit rating, differentiation does not disappear; it becomes invisible. Both sets are serious.

Our position, stated as a position rather than as a research finding, is that the ratings question is secondary. An organization that keeps ratings and separates the two clocks gets usable development conversations. One that removes ratings without separating the clocks will find that evaluation has not disappeared, it has moved into informal conversations, where it can no longer be contested or corrected.

What to change before your next review cycle

Do not rebuild the system. Do one thing before the next cycle.

Take the appraisal form you use today and cut every section asking for future plans, development needs or growth objectives. Keep only what looks backward. What you cut is the material that was not working there.

Then put a twenty-minute conversation every two weeks into each manager's calendar, with each person, about the work ahead. And write one sentence for the team: what is discussed there does not enter the review.

If you want to see where performance is lost at the level of the system, a capability diagnostic locates the real constraint before you change the process. And if you work in HR or L&D and recognise the meeting from the opening, what we ask of HR teams sets out the alternative.

So we will leave you with a question. If you asked five people tomorrow what happens to the things they say in one-to-ones with their manager, how many would know, and how many would tell you something you would rather not hear?

FAQ

What is a performance review?

It is the process by which an organization establishes retrospectively how well someone performed their role over a given period, usually ending in a rating with administrative consequences. It is not development, which looks forward and should carry no consequences. Confusing the two functions is why the instrument changes nothing.

How often should performance reviews happen?

The evaluation can stay annual or semi-annual, because its purpose is to be fair and documented rather than frequent. The development conversation needs a different rhythm, every two weeks, because it concerns work in progress. Gallup recommends progress reviews at least twice a year.

Why does feedback in an annual review not work?

Because it arrives attached to a consequence, which moves attention from the task to the self. Kluger and DeNisi found, across 607 effect sizes, that over a third of feedback interventions reduced performance, and the theory they proposed links weak results to feedback directing attention toward the self.

What is the non-contamination rule?

It means that what an employee says in a development conversation is never used as evidence in their evaluation. The manager does not quote a one-to-one admission into the appraisal form. The rule has to be written and communicated, because its value lies in people believing it.

Do we have to drop ratings for this to work?

No. Whether to eliminate performance ratings remains contested among industrial-organizational psychologists themselves, with serious arguments on both sides (Adler and colleagues, 2016). Separation works with ratings and without. Removing ratings without separating the functions moves evaluation into informal conversations, where it becomes harder to contest.

Sources

• Kluger, A. N. and DeNisi, A. (1996). The Effects of Feedback Interventions on Performance. Psychological Bulletin, 119(2), 254-284. Meta-analysis of 607 effect sizes across 23,663 observations: feedback interventions improved performance on average, but over a third of them reduced it, a result not explained by sampling error or feedback sign. Feedback Intervention Theory is proposed to account for the spread. Laboratory and field research, largely North American.
• Wigert, B. and Mann, A. (2017, updated 2026). Give Performance Reviews That Actually Inspire Employees. Gallup. 14% of employees strongly agree reviews inspire them to improve, 29% that they are fair, 26% that they are accurate; 48% are reviewed annually. Gallup recommends progress reviews at least every six months. Gallup's international database; the source does not state a sample size.
• Adler, S., Campion, M., Colquitt, A., Grubb, A., Murphy, K., Ollander-Krane, R. and Pulakos, E. D. (2016). Getting Rid of Performance Ratings: Genius or Folly? A Debate. Industrial and Organizational Psychology, 9(2), 219-252. Debate among seven industrial-organizational psychologists, arising from the 2015 SIOP conference, with arguments for and against eliminating performance ratings. US and international.

By Szilárd Kacsó