Quick summary
A performance review that actually changes behavior separates pay from development, and draws on specific examples gathered across the whole period. It sets concrete goals, and gets rehearsed before the room rather than improvised inside it. Most reviews fail before the meeting starts, when a manager walks in with vague impressions instead of specific evidence.
The traditional review does not have a strong track record. Gallup research found that only 14% of employees strongly agree their performance review inspires them to improve. The same research cites Kluger and DeNisi’s 1996 meta-analysis, published in Psychological Bulletin. It found that feedback interventions made performance worse in roughly a third of the cases studied. Gallup also found that managers who give weekly feedback, instead of annual, see a real difference. Their team members are 5.2 times more likely to say they receive meaningful feedback, and 2.7 times more likely to be engaged at work.
Deloitte reached a similar conclusion independently. In a 2015 Harvard Business Review article, Marcus Buckingham and Ashley Goodall described Deloitte’s own performance management overhaul. The company had been spending close to 2 million hours a year on the process. The company also found that a rating said more about the rater’s own tendencies than about the person being rated. Researchers call this pattern an idiosyncratic rater effect. Deloitte’s response included separating pay decisions from ongoing performance conversations entirely, the same principle covered in step two below.
Conducting a good review is a specific, learnable skill, not a personality trait some managers happen to have. The seven steps below cover how to prepare for one, structure the conversation itself, and follow up afterward.
Step 1, gather specific examples across the whole review period, not just the last few weeks
Write down concrete moments as they happen throughout the period, rather than trying to reconstruct a year from memory the week before the review. Recency bias is one of the most common ways reviews go wrong. A single strong or weak week right before the review tends to dominate a manager’s impression, even when it does not represent the broader period. A short running note, even a few lines added after each notable project or interaction, solves this more reliably than memory does.
Step 2, separate the pay conversation from the development conversation
Decide in advance whether this specific review is about compensation, development, or both, and be explicit about which. Gallup’s research found that mixing too many purposes into a single conversation is one of the clearest reasons reviews feel confusing and tense. Deloitte reached the same conclusion, separating pay decisions from its ongoing performance conversations after finding the combination undermined both. An employee who is bracing for a pay outcome will not absorb developmental feedback in the same conversation, regardless of how well it is delivered.
Step 3, lead with the specific behavior and its impact, not a rating
State a specific situation, describe the observable behavior, then explain its impact. This is the same SBI structure that works for any difficult feedback conversation. “Your rating this cycle is a 3 out of 5” tells someone where they landed. “The client escalation you resolved in March prevented us from losing that account, and it is the reason your rating improved this cycle” tells them why. A rating without the specific behavior behind it teaches nothing about what to repeat or change.
Worth remembering
A number without a story behind it is not feedback. It is a grade with no explanation for how it was earned.
Step 4, make it a conversation, not a monologue
Ask questions and leave real space for a response, rather than reading through prepared points from start to finish. A question like “How did that project feel from your side” can surface context a manager did not have. It also signals that the rating reflects a discussion, not a verdict delivered from behind a desk. Gallup’s research points to the same idea, describing the best managers as architects of a genuine dialogue, not administrators of a fixed script.
Step 5, set specific goals for the next period, not general aspirations
End the review with a small number of concrete goals that can be checked at the next one. “Keep improving communication” cannot be verified later. “Send a written project update every Friday, and flag blockers within 24 hours instead of at the next standup” can be checked directly. Reviews that end on a vague aspiration tend to produce the same vague conversation again at the next cycle.
Step 6, rehearse the hardest parts of the conversation beforehand
Practice the specific moments most likely to go sideways before delivering them for real. That might be a lower-than-expected rating, a promotion that did not happen, or a piece of overdue feedback. This is not a new idea. In Gallup’s account, Adobe’s HR team found that moving to more frequent check-ins required direct manager training. That meant literally role-playing what a positive manager-employee conversation looks like before doing it with real employees. EasyCoach, Easygenerator’s AI coaching product, gives managers a private space to run that same kind of rehearsal. It lets a manager practice the exact review conversation with an AI playing the employee before the real one happens. Our guide to preparing for a hard conversation with AI covers the same rehearsal process in more depth.
Step 7, treat the review as one checkpoint in an ongoing conversation, not the whole conversation
Follow up on the specific goals from step five well before the next formal review, rather than letting a year pass in silence. This is the same coaching cadence that separates a genuine, ongoing coaching relationship from an isolated annual event. A review that stands alone, with no check-in until the next one, is asking a single conversation to do the work of an entire year.
Common mistakes that undo an otherwise good review
Even a well-intentioned manager can undercut a review with a handful of recurring mistakes.
Recency bias
Rating based on the most recent weeks instead of the full period is the most common mistake, covered in step one.
Surprising ratings
Giving an employee a rating they did not see coming is another. A review should confirm ongoing feedback, not introduce it for the first time.
One-way monologue
Turning the conversation into a one-way monologue, rather than the two-way exchange covered in step four, is a third mistake.
No checkable goal
Ending the review without any specific, checkable goal all but guarantees the same vague conversation repeats at the next cycle.
Annual-only reviews versus ongoing performance conversations
| Annual-only reviews | Reviews backed by ongoing conversations | |
|---|---|---|
| Frequency of feedback | Once or twice a year | Weekly or biweekly check-ins, plus the formal review |
| Risk of surprise | High, since months pass between conversations | Low, since the rating reflects what was already discussed |
| Basis for the rating | Often skewed toward the last few weeks | Draws on notes and context gathered across the period |
| Employee engagement | Gallup found 2.7x lower among infrequently reviewed employees | Higher, tied to consistent, meaningful feedback |
| What happens after | Often nothing until next year | Specific goals get checked before the next cycle |
A review that actually changes something
None of these seven steps require a different personality or a natural gift for hard conversations. They require preparation before the room, a specific structure inside it, and a follow-up habit afterward. Most performance reviews that fail to change anything do not fail because the manager was unclear or unkind. They fail because the evidence was never gathered, the goals were never specific, or nobody checked back in before the next cycle rolled around. A manager who treats the review as one step in a year-long conversation, rather than the whole conversation itself, is the one whose reviews actually move the needle.
Feedback that actually changes something
None of these seven steps are complicated on their own. Naming a specific behavior, deciding an outcome, structuring the delivery, checking intent, agreeing on a next step, rehearsing it, and following up are all straightforward ideas by themselves. What makes them hard is doing all seven in the same conversation. That gets harder still under time pressure, with a person who might react in a way you did not expect.
Most feedback that fails does not fail because the manager lacked courage. It fails because one of these seven steps got skipped under pressure. Usually that is the specific behavior in step one or the follow-up in step seven. Those two steps take the least visible effort to leave out in the moment. The manager who treats this as a repeatable process, not a single hard moment to survive, is the one whose direct reports actually change their behavior afterward.
