Recency bias in performance reviews
Most performance reviews are quietly a memory test, and memory brings a bias to it. You sit down to summarize six or twelve months, and the last few weeks arrive in high resolution while the earlier ones show up as a blur you have to squint at. Whatever's recent feels representative, though it usually isn't. That's recency bias, and in a review it does real damage: a strong quarter undone by a rough final month, or a weak stretch rescued by a good sprint that happened to land in the window just before you wrote.
It's worth knowing how much of this is you and how much is just how recall works. Mostly the latter.
The problem is well documented
The plain version of the effect is one of the oldest findings in memory research. When people try to recall a list, the items at the end come back most easily; Murdock mapped that recency curve in 1962, and it has held up across decades of replication. A review period is a list of sorts, and the end of it is what your memory hands you first. Decay does the rest: Ebbinghaus's forgetting curve, measured in 1885 and replicated with modern controls in 2015, is steepest in the first days, keeps sliding for months, and grants no exemption to the things that felt memorable at the time.
The effect shows up specifically in performance evaluation, not just word lists. Steiner and Rain had raters evaluate a series of performances and found that a recent good or poor showing pulled the overall rating toward itself, more so when the performances were spread over days rather than seen in one sitting, which is the real-world case, where you watch someone over months. The thing you saw last weighs more than its share.
The deeper problem is worse than recency
Set recency aside, and a rating from memory has a second flaw that's larger and stranger. In a study of thousands of managers rating their reports, Scullen, Mount, and Goff found that most of the variance in the ratings, more than half, came not from the person being rated but from idiosyncrasies of the rater doing the rating. As Marcus Buckingham and Ashley Goodall summarized that work in HBR, the ratings "reveal more about the rater than they do about the ratee." Your review of someone is, to a large degree, a reflection of you.
This is the honest backdrop to any tool that promises to fix reviews, including mine: the raw material is noisy at the source. No note-taking habit makes your judgment objective, and nothing here claims to.
What actually helps, and what it doesn't prove
What a running record does is narrower and real: it takes the recency part out. If you wrote down what you noticed the week it happened, then at review time you're reading dated observations from across the whole period, instead of asking a recency-biased memory to reconstruct them. The good sprint from May is on the page at the same resolution as the rough one in October. You can still weigh them badly, but you're weighing the whole arc rather than whichever slice recall surfaced.
It's also the one fix the appraisal literature actually tested. In DeNisi, Robbins and Cafferty's lab studies, raters who kept structured diaries organized what they'd seen better in memory and recalled it more accurately. The finding then survived contact with the field: across two organizations, raters trained in diary keeping remembered more of what their people had done and produced ratings that were less inflated and better differentiated.
Candor about the limits. Those studies trained raters in a structured practice; nobody has run the controlled trial where managers keep private notes of their own accord and the reviews come out measurably fairer. The mechanism is established, the diary evidence sits close to it, and the last step is still an inference, though one I think is worth acting on. There's also a selection problem in the wild: the managers who keep a record are already the ones who care about getting this right. If someone runs the private-notes study, I'll link it here, whichever way it comes out.
On that evidence, the move is cheap and the downside is small: write the read down as you go (the template has the fields), so the review reads from a record instead of a reconstruction. It won't make you objective. It will stop the last three weeks from doing all the talking.
Full disclosure: I make Continuum, a Mac app built on this. You note what you see when you see it, and older notes fade unless you notice the same thing again, so what's in front of you at review time is the whole live picture rather than the last month crowding it out. It keeps confidence and status in words and deliberately never scores a person, precisely because a score would dress up exactly the noise the research above describes. Nothing leaves your Mac; it's free for up to three people, and Pro is a one-time purchase if you'd rather not subscribe.