The review is due Friday. You open the blank form, and your mind helpfully offers you the last three weeks: the deck she pulled together for the board, the deadline he nearly missed, the thing someone said in standup on Monday. This feels like remembering. It is not. It is the most recent sediment, and you are about to mistake it for the riverbed.
I am not going to explain why memory does this; a year’s worth of work compresses into a handful of vivid recent scenes for reasons that are well understood and not your fault. The interesting question is what a fair manager does about it once they accept it as a fact of their own equipment. Because the damage lands hardest right here, in the one document that follows a person into their next raise, their next role, their next sense of whether the work was seen.
Stop trying to remember. Start trying to reconstruct.
The single most useful move is to stop treating the review as an act of recall and start treating it as an act of reconstruction. Recall asks, what do I remember about Priya this year? Reconstruction asks, what actually happened, month by month, and where can I go to check? Those are different cognitive jobs, and the second one is far more honest.
So before you write a word of assessment, rebuild the timeline. Open the calendar and walk it backward, quarter by quarter. Pull the shipped projects. Skim the channels where her work showed up. Reread the notes from your one-on-ones if you kept them. What you are after is not a highlight reel but a census: the ordinary weeks count too, and they are precisely the ones memory throws away. When I do this properly I am always slightly embarrassed by how much I had forgotten: a migration that ate three weeks in the spring, a quiet save during a hire’s first month, an entire workstream that had simply fallen below the waterline of recent attention.
Never write a review from memory alone
Treat that as a rule, not a preference. A review written purely from your head is a review of the last month with footnotes, no matter how fair-minded you intend to be. The fix is not more concentration. You cannot strain harder and recover information that was never durably stored; effort applied to an empty bucket just produces confident invention.
What recall can’t give you, a contemporaneous record can. A line you wrote in April (handled the vendor escalation calmly, the client specifically asked for her on the next phase) is worth more than an hour of squinting in June, because the April version still has the texture. It knows what actually happened before the ending sanded it smooth. Recollection gives you a verdict; the note gives you the evidence the verdict was supposed to rest on. This is the whole reason to keep any record at all: not so you write less, but so that what you write is true.
Weight the year, don’t average the mood
Reconstruction gives you the raw material. Now you have to judge it, and judgment is exactly where recency sneaks back in wearing a suit. A strong recent month feels like momentum; a strong quiet spring feels like ancient history. They are not actually worth those amounts.
So weigh them deliberately. Lay the whole period out and ask what genuinely moved: the spring stretch where he rebuilt the onboarding flow with no fanfare and cut drop-off by a third is not lesser work because it didn’t happen in the same week as the review. A loud finish is sometimes a real acceleration and sometimes just proximity. The test I use is simple: if this had happened in February instead of last week, would I still rate it this highly? If the answer is no, I am grading the calendar, not the contribution. The reverse matters more. The person who did excellent, unannounced work in March and went quiet in June is the one the year is most likely to cheat, and they are usually the one who least expects you to have noticed.
Calibrate, and let yourself be corrected
Then check your read against something outside your own head. Compare ratings across your team and look for the tell: are your top scores clustered among the people who happened to be loud in the last six weeks? Sit with a peer manager and trade calibrations; if you cannot point to specific reconstructed evidence for a rating, that is a sign the rating is a mood with a number attached. Calibration is not bureaucracy. It is the cheapest available correction for a bias you already know you have.
And when you hit the gap, the moment you realize you genuinely cannot remember what someone did in the second quarter, be honest that the gap is in your input, not in their performance. The instinct is to paper over it with a confident generality. Don’t. Go find the record instead, or admit in the review that a stretch is thinly documented and weight it cautiously rather than inventing a verdict to fill the silence. A fair review is allowed to say I am less sure about this period. It is not allowed to guess and call the guess an assessment.
Fix the input, not the output
All of this points at one uncomfortable conclusion: the quality of a review is decided long before review season, by whether anything got written down while it was still true. You can be the most scrupulous, least biased manager in the building and still produce a thin review, because scrupulousness applied to a faded memory just gives you a carefully reasoned account of the last three weeks.
This is where a quiet habit beats a heroic effort. Keeping a private notebook on the people you manage (a line when something lands, a note when a hard conversation goes well, dropped in from your desk, or over WhatsApp, in the ten seconds before it evaporates) is the unglamorous discipline that makes June honest. Notivo exists for exactly this: a private place to jot what you observed about a person as it happens, so that when the review comes due, the year is already written down and the assistant can hand you back what you noticed in April rather than asking you to invent it. It doesn’t make the judgment for you. It just makes sure the judgment has the whole year to work with.
The form will still be blank on Friday. The difference is whether you fill it from a record or from a feeling. One of those is a review. The other is a guess that the last three weeks happened to write for you.
Frequently asked questions
How do I write a fair review if I did not keep notes all year?
Reconstruct instead of recall. Walk the calendar backward quarter by quarter and pull from real sources: shipped projects, channels where the person's work showed up, and any one-on-one notes you saved. Where a stretch is genuinely thin, say so in the review and weight it cautiously rather than inventing a verdict to fill the silence. Then start a running log now so next cycle is reading, not remembering.
How do I reconstruct a year of someone's work before I rate them?
Open the calendar and treat it as a census, not a highlight reel. Go month by month and write down what actually happened, including the ordinary weeks that memory throws away. Skim shipped work, reread your one-on-one notes, and check the places the person's contributions landed. In Notivo this step is already done: if you tagged them with @Name through the year, their dated timeline is sitting there in order, so reconstruction becomes reading.
How do I check whether a strong recent month is real or just recent?
Use a simple test on each example: if this had happened in February instead of last week, would I still rate it this highly? If the answer is no, you are grading the calendar, not the contribution. The harder direction matters more. Quiet, excellent work done early in the period is the work the year is most likely to cheat, so go looking for it on purpose.
Does calibration with other managers actually reduce recency bias?
It helps, because it checks your read against something outside your own head. Compare ratings across your team and look for the tell: are your top scores clustered among the people who were loud in the last six weeks? Trade calibrations with a peer manager, and if you cannot point to specific dated evidence for a rating, treat that as a sign the number is a mood, not an assessment.