Are Performance Reviews Effective? What the Evidence Says in 2026.

It is review season. Somewhere this week a manager is writing up a year of someone's work in an afternoon, from memory and a few saved emails, and both of them half expect the form to change nothing except the size of the raise.

That suspicion deserves a proper test, because whether performance reviews are effective usually gets a shrug for an answer. Ask the people who run them and they work. Ask the person being reviewed and they are a ritual.

Both are right. A review is asked to do three different jobs at once, and it does them very unevenly. Grade the jobs one at a time and the question of whether performance reviews work gets a clear answer.

What a performance review is supposed to do.

The purpose of performance reviews, once you strip out the HR language, comes down to three jobs. Most arguments about whether reviews work talk past each other because each side is grading a different one.

1. Decide. Settle pay, bonuses, promotions and sometimes exits, on a basis the organization can explain and defend.
2. Develop. Help the person do better work next year than this year.
3. Measure. Leave an accurate record of who performs and how, which talent reviews and succession plans run on long after the meeting ends.

These are the benefits of performance reviews when they work. The review itself is one event in the wider performance management cycle, which runs all year: plan, monitor, develop, rate, reward.

The review is the rating step. The trouble starts when it is asked to carry the other four as well.

Are performance reviews effective? The evidence, job by job.

Here is how each job scores against evidence published this year. The verdicts differ, which is why performance appraisal effectiveness has been so hard to pin down as a single answer.

01Deciding pay and promotion: mostly yes

This is the job reviews do best. In the Talent Strategy Group's 2026 benchmark of more than 250 organizations, those with a formal review rated their process 41% more effective at paying for performance (Talent Strategy Group, February 2026).

The same organizations rated it only 16% more effective at assessing performance accurately. Read the two numbers together: a review is good at producing a decision, and much less good at producing a correct one.

The decision is only as good as the ratings underneath it, and ratings drift upward. The US federal government gave the starkest example this year: in its own appraisal data, about 0.1% of covered employees receive an Unacceptable rating.

In a final rule published in July 2026, the Office of Personnel Management concluded that rating inflation had "so egregiously undermined the credibility and accountability of performance appraisal systems" that it lifted the ban on capping top ratings.

Supervisors, the rule says, inflate ratings "to avoid confrontation." The cap itself is contested, and commenters cited research against forced distributions. You do not have to like the cap to accept the diagnosis.

A scale on which almost nobody ever lands at the bottom is not deciding very much. It is recording a decision someone made somewhere else.

02Improving performance: mostly no, as usually run

This is the job reviews are sold on, and the one they do worst. As of August 2026, 18% of US employees said their performance is managed in a way that motivates them to do outstanding work (Gallup, August 2026).

Only 19% said they had received meaningful feedback in the last week. Four in five went the whole week without any.

Are performance reviews effective for the people being reviewed? What US employees say

Say their performance is managed in a way that motivates them 18%

Received meaningful feedback in the last week 19%

Have a clear definition of exceptional performance in their role 22%

The research explains why. A systematic review of 173 studies, led by Emma Heine of Macquarie University and published in the Journal of Organizational Behavior in February 2026, found a clear split in what feedback does (Macquarie University, January 2026).

Positive feedback reliably improves performance. Negative feedback works only when there is a strong, trusting relationship between the manager and the employee.

Now look at the format. The annual review saves up a year of criticism and delivers it in one meeting, with pay on the table, whether or not the relationship can carry it.

Heine's verdict: "Annual reviews, especially when tied to bonuses or promotions, are often too charged to genuinely support development and performance improvement."

03Measuring performance accurately: the weakest job

The same review found that ratings reflect more than performance. Managers tend to overrate in formal evaluations, not out of generosity, but because the formal setting raises the risk of conflict. Gender stereotypes also shape how people are judged.

There is a simpler problem too. Only 22% of US employees say they have a clear definition of what exceptional performance looks like in their role, on Gallup's August 2026 figures. Nobody can rate accurately against a standard that was never written down.

And in 2026 the standard is moving. In a March 2026 survey by Betterworks, which sells performance software, 90% of HR leaders said AI has redefined high performance, yet only 42% of organizations reflect AI expectations in their goals.

Most reviews this year will grade last year's job. That matters beyond the meeting, because the ratings outlive it: talent reviews and succession decisions built on those ratings inherit every error.

Why annual performance reviews fail.

Most of the case against annual performance reviews is a case against the format, not against reviewing. Four things go wrong when a year is squeezed into one meeting.

1. A year gets compressed into an afternoon. Work now happens continuously, not episodically, as I put it in Rewired Leadership. A review written from memory in November mostly remembers October.
2. Three jobs share one meeting, and money wins. Once the raise is on the table, the development conversation is something the person sits through, not something they hear.
3. Criticism is saved up and delivered at once. That is the condition the research says works worst, in the meeting where defenses are highest.
4. Goals are set once, in a year the job moved. In the Talent Strategy Group benchmark, 56.3% of organizations still review once a year, 84.8% set goals once a year, and only 46.7% actively encourage adjusting goals along the way.

So are performance reviews outdated? The annual decision is not, and it can stay annual if that suits your pay cycle. Feedback saved up for a year is what is outdated, and AI has made it more so, because the work now changes faster than the goals written for it.

Should you get rid of performance reviews?

For most organizations, no. The decision job does not go away when the form does. Pay still gets set and promotions still get made, only now in a room the employee never sees, by people who never wrote down why.

That is less fair, not more. Very few organizations have taken the plunge: in the Talent Strategy Group benchmark, 92.4% of organizations that review performance use ratings, and just 7.6% run without them.

The more useful finding is about redesign itself. More than half of the organizations (55.4%) redesigned their process in the last two years, and how recently they redesigned it made no material difference to how effective they rate it.

Changing the format, in other words, is not the fix. What did go with higher ratings of effectiveness was less glamorous: mandatory training on how to review, which only 19.2% require, and one process across the whole organization instead of several.

These are correlations in self-reported data, so treat them as direction, not proof. But they agree with the research: the people running the review matter more than its format.

The question to ask, then, is not whether performance reviews are necessary. It is whether yours does the one job it is good at, and whether the other two have somewhere better to live.

What works instead: keep the decision, move the development.

Making performance appraisal more effective starts with that split. Keep the review for the decision it is good at, and move development out of it into the weeks in between. These are the six changes, in the order I make them when I redesign a review process with a CHRO.

01Hold the pay conversation apart from the feedback conversation

The rating feeds pay, and it should. But hold the conversation about money weeks apart from the conversation about the work, so the second one gets heard.

Asked whether performance reviews should be tied to compensation, my answer is yes for the decision and no for the feedback.

02Write down what exceptional looks like before the year starts

One or two sentences per role, agreed with the person at the start of the year, not discovered by them at the end. Write it for the roles you will need, not the ones on last year's org chart.

In 2026 that definition has to say what the person does with the work AI hands back: the calls they make, not the volume they produce. In Rewired Leadership I make this the first mandate for the CHRO: measure the quality of human judgment, not the volume of human activity.

03Make feedback weekly and mostly positive, then earn the hard part

Continuous feedback works when it is specific and mostly about what is going well, because that is the kind the research shows reliably lifts performance. A week is a better unit than a year.

Criticism still has its place. "Negative feedback should not be avoided when it is warranted," Heine says. "But to achieve performance improvement, a high-quality relationship is critical."

Where that relationship is not there yet, she adds, clear goals and a credible manager help. Building the trust that lets hard feedback land is the manager's real job here, and it happens week by week, not in November.

04Debrief the work right after it happens

The best development tool I know is not a review at all. It is the after-action review, borrowed from the military and set out in my book: a short, candid debrief straight after any significant piece of work, built on four questions.

• What was supposed to happen?
• What happened?
• Why was there a difference?
• What will we do differently next time?

Learning must be built into the workflow, not bolted on as a quarterly review.

When something has gone wrong, ask which guardrail failed before you ask who made the error. That habit, practiced by a manager who coaches in the moment, does more for next year's performance than any form.

05Calibrate the ratings and train the people who give them

Calibration means managers compare their ratings across teams before they are final, and it is the practical counter to inflation. In the Talent Strategy Group benchmark, 84.7% of organizations already hold calibration meetings.

Training is the gap: fewer than one in five require reviewers to be trained. Train for the conversation and the judgment, not for filling in the form.

06Let AI remember the year, never let it rate

A model that reads a year of project notes, check-ins and customer feedback fixes the recency problem better than any reminder can. That alone makes a review fairer.

The rating stays with a manager who can defend every line of it in person, and people are told when AI helped draft their review. A score nobody is accountable for is not a decision.

How to measure performance review effectiveness.

Performance review effectiveness is rarely measured, which is how weak processes survive for decades. These six tests will tell you more about yours than another engagement survey.

1. Look at the distribution. What share of people sit in the top two rating levels, and how many below "meets"? If almost nobody does, the rating is not deciding anything.
2. Ask ten people what exceptional looks like in their role. If their answers do not match their managers' answers, the standard does not exist yet.
3. Count the conversations. Compare the feedback conversations that happened with what the policy asks for. In the benchmark, 68.5% of organizations report just one or two per cycle.
4. Check whether ratings predict anything. Do this year's ratings line up with next year's results, with how promoted people perform, with who you regret losing?
5. Look for rater patterns. If ratings cluster by manager rather than by team results, you are measuring the managers.
6. Read the form against the work. Does it still score tasks that AI now does?

If three or more of those fail, your review is producing decisions on noise. An outside read of how your review process really runs is where I would start, because it shows which of the three jobs is breaking before you redesign anything.

Frequently asked questions.

The short version.

Performance reviews are good at producing a decision and poor at producing better work or an accurate measure. Keep the review for the decision, calibrate it and train the people who give it.

Move development into weekly conversations and debriefs, and rewrite the standard for the work people do now that AI does part of it. That is the heart of redesigning the review around the work people now do, and it starts with knowing which of the three jobs yours is failing.