recency effect
Also called: recency bias
The tendency to give the latest events more weight than earlier ones, so the last weeks before a review outweigh months of steady work. A fact log kept through the whole period is the usual cure.
The recency effect is the tendency to give the latest events more weight than earlier ones simply because they are easier to remember. In a review it means the last few weeks before the review outweigh months of steady work, whether those weeks were unusually good or unusually bad. Memory works this way for everyone: recent, vivid and emotional events come to mind first.
It makes reviews unfair in predictable ways. Work varies from week to week, so one recent week is a small and noisy sample; a person with a late peak gets rewarded and a person whose strong months were in the spring disappears. A recent week often also rests on few observations: a satisfaction rate built on 25 survey answers moves 4 points with every answer. And once people notice what counts, effort starts to bunch up right before the review.
The cure is not to ignore recent events but to put them in proportion. Keep a fact log through the whole period: dated, observable facts and their impact, good and bad, for everyone. Compare a striking week with the same person's own baseline and with how much of the period it covers. Check the size of the base behind any rate. And hold regular one-to-ones through the year, so nothing in the review is a surprise.
Example
Klasna's review season covers January–July 2027. Taras's most memorable week is the last one he worked before the reviews, 12 July, during a payment-page fault:
The week is 1 of 30 in the period, 152 / 2,833 = 5.37% of his tickets and 2.96% of his survey answers. Counted in, it moves his seven-month reopen rate from 9.51% to 9.07%, while his reopen rate was above 8% in every month from January to June, and his other July week looks like the rest of the year. Its satisfaction rate rests on 25 answers, so a single answer moves it by 4 points.
A July-only review would show Taras at 4.49% reopened and 87.04% satisfied, close to Bohdan's 3.42% and 87.88%, and rate him far higher than seven months of facts support. The week goes into the review as one dated fact, not as the review.
Common mistakes
- Reviewing from memory. Without a fact log, the last weeks fill the whole picture.
- Reading a standout week as a new normal. Compare it with the person's own months before it.
- Ignoring a small base. A rate on 25 answers can swing by several points from one answer.
- Looking only at the last month of data. Use the whole review period, with the same window for everyone.
- Dropping the recent event altogether. It is still a fact; give it its share, not zero.
Learn it in the course
- Evaluating people fairly · Keep a fact log, spot halo, recency and similar-to-me bias, use the same criteria for everyone and ask for a second opinion.
- Case: the review season · Seven months of facts, four coordinators, one great July week: write fair evaluations and a next-year plan for each person.