Sarah opens her laptop on a Tuesday in mid-October. Annual reviews are due in three weeks. She has six direct reports. She has eleven months of work to evaluate. She has, realistically, four weeks of memories she can actually access without straining.
So she writes what she remembers. Daniel crushed the Q3 launch. Rachel missed a deadline last Thursday. Maya has been quiet in standups lately. By the time the review meetings happen, six people will get rated on six weeks of behavior dressed up as a year of judgment.
This is recency bias. And it is the single most common reason performance reviews feel unfair to the people receiving them.
What is recency bias?
Recency bias is the tendency to weigh recent events more heavily than older ones when forming judgments. In performance reviews, it means a manager’s rating reflects the last month of work far more than the ten months before it.
It is a close cousin of availability bias, the broader heuristic where we judge things by what comes to mind easily. Recent events come to mind first, so they shape the rating. The science here is settled. What is not settled, in most companies, is what to do about it.
The diagnosis most managers reach for is willpower: “I will just try harder to remember the full year.” That fix does not work, because the problem is not effort. Human memory was never built to hold eleven months of granular performance data per direct report. Without a system to externalize that memory, recency bias is structurally guaranteed.
Five places recency bias quietly distorts decisions
Recency bias is not confined to annual reviews. It shows up across the management decision stack:
- Performance reviews. A strong September erases a shaky April. A rough last sprint erases two excellent quarters. The rating reflects the manager’s recall window, not the employee’s actual year.
- Promotions. The person who shipped a visible project last month gets named ahead of the person who has been steadily compounding for two years.
- Hiring. The candidate from last week’s interview feels sharper than the one from three weeks ago, even when the older notes say otherwise.
- Project assignments. A high-stakes project goes to whoever just finished a win, instead of whoever is actually the best fit.
- Disciplinary action. A recent mistake gets weighted as a pattern. A pattern of strong work gets discounted as old news.
In every case, the manager is not being lazy. They are using the only data their memory hands them. The fix is to stop relying on memory.
Why “use technology” is not actually the answer
The standard advice on recency bias goes something like: be aware of it, take notes, use technology, get peer input. This is true and almost useless, because it does not tell a manager what to do on Monday morning.
The version that works is more specific. Recency bias is a process problem. Process problems get solved with cadence, not consciousness. Here is the cadence that actually moves the needle.
1. The five-minute weekly note per direct
Every Friday, before logging off, spend five minutes per direct report writing three lines:
- One thing they did well this week (with specifics).
- One thing they struggled with (with specifics).
- One thing to watch next week.
Five minutes times six directs is half an hour. By the end of a year you have roughly fifty data points per person. That is no longer a memory problem. That is a documentation system.
The specificity matters. “Rachel did great work” is useless six months later. “Rachel rewrote the onboarding doc after the Patrick incident, cut new-hire ramp from 14 to 9 days” is evidence you can cite in a review.
2. The quarterly evidence review
Every quarter, spend an hour per direct report consolidating the weekly notes into themes. What patterns are showing up? Where is the trajectory? What would a fair rating look like if the review were today?
This does two things. It forces you to make judgments four times a year instead of once, which dilutes recency bias by definition. It also gives the employee a quarterly artifact you can share, which makes the annual review a confirmation of known information instead of a surprise.
3. Peer input, structured
Peer input is undervalued because it is usually unstructured. “Anyone have feedback on Maya?” produces nothing useful. Structured peer input asks three questions: What has Maya done well this quarter that I might not have seen? Where could she be stretching more? What is one thing she should keep doing?
Two or three peers, three questions each, ten minutes per peer. You get cross-functional evidence that bypasses your own recall window entirely.
4. Clear expectations set at the start
Recency bias gets worse when the goalposts are fuzzy, because the manager fills the gaps with whatever is fresh. Clear quarterly expectations, written down at the start of the quarter, give you a fixed rubric to evaluate against. Eleanor’s Q2 was strong or weak relative to what was agreed in April, not relative to what you remember from last week. See our guide to setting expectations at work for the mechanics.
5. Shorter review cycles
Annual reviews are recency bias machines because they ask managers to compress a year into a memory. Quarterly reviews shrink the window. Monthly check-ins shrink it further. The shorter the cycle, the less work memory has to do, and the more accurate the assessment.
Companies do not need to abolish the annual review. They need to make sure the annual review is a summary of four quarterly artifacts, not a panicked October recall exercise.
How Merlin closes the bias gap in real time
The hardest part of all this is the discipline. Most managers know they should document weekly. Few do, because it competes with everything else on the calendar.
This is where Risely’s AI coach Merlin helps. Merlin lives natively inside Slack and Microsoft Teams, where managers already work. Three things happen there:
- Pre-review evidence surfacing. Before a review conversation, Merlin pulls forward documented moments from across the period, so the manager walks in with twelve data points instead of two.
- Bias pattern checks. Merlin reviews the language a manager is using across ratings. When the same person is described as “promising” in Q1 and “underperforming” in Q4 with no documented trajectory in between, that is a flag worth pausing on.
- In-the-moment coaching nudges. When a manager asks Merlin to help draft review feedback, Merlin asks for evidence first. If the evidence is thin or all from the last month, Merlin says so.
This is what bias-aware coaching looks like in practice. Not a training module on cognitive biases that managers forget the next day. A coaching layer that catches bias in the workflow where it actually shows up.
For a wider view of how cognitive shortcuts distort manager judgment, see our breakdown of the top ten manager biases and our guide to addressing bias in performance reviews.
What this looks like when it works
Managers who run a weekly documentation cadence and use coaching support consistently rate their directs differently than managers who do not. The gap shows up most clearly in two places: the spread of ratings (less compression toward whatever happened recently) and the quality of the evidence cited in calibration meetings.
Across 5,000+ users coached in 40+ organizations, Risely sees managers improve target manager skills by an average of 26% in 12 weeks, with 73% high engagement on daily nudges. Decision-making and feedback skills are the two that move fastest, because they have the clearest behavioral signals to coach against.
Recency bias is not a personal failing. It is what happens when you ask a human brain to do what a documentation system should do. Build the system, and the bias starts to shrink on its own.
Stop running performance reviews on memory
If you are heading into a review cycle right now, skip the training module on bias. Run a fifteen-minute conversation with your team about what evidence you will document going forward, and put a Friday calendar block in place that makes weekly notes non-negotiable.
Try Merlin free for 14 days and let the in-flow nudges do the discipline work for you. Your next review cycle should be a confirmation of evidence you have already collected, not a panicked October recall.