Article & InsightBehaviour Intelligence

Halo and Recency: How One Impression Can Shape a Whole Review

Halo and recency effects can distort a performance review. How each works, and four checks to run against a dated record before you submit.

A woman in glasses and a navy blazer rests her chin on her hand and looks upwards, with a glowing halo above her head and a row of star-rating cards rising towards a sunrise over misty mountains.
OE
Opptimise Editorial Team
Behaviour Intelligence Research
November 2026
•
7 min read

It's the second week of December, and a manager is drafting year-end reviews. This is an illustrative scenario. The first review is for Priya, an analyst.

The manager's clearest memories of Priya are a strong analysis in January and two missed deadlines in the last two months. In between, the year blurs. Ratings start to fill in, and most of them lean the same way.

However, that lean is worth examining. A review summarises twelve months, but it's written at one moment, from one memory.

The short version: A review is a summary of a year written at one moment. Two habits of judgement can distort it. A halo spreads one strong impression across unrelated ratings. Recency gives the latest weeks more weight than the earlier months. The order you see things in can tilt the rating either way. Rate each dimension separately, and check your draft against a dated record.

What are halo and recency?

The halo effect is the tendency for one overall impression of a person, good or bad, to colour how you judge their separate qualities.

Recency is the tendency to give the most recent events more weight than earlier ones when you sum up a period.

Both can work as shortcuts, and in the halo study viewers didn't notice theirs. They also sit alongside the fundamental attribution error, which concerns whether you explain a result through the person or the situation. This article is about something different: how you weigh and spread what you saw.

Our guide to Behaviour Intelligence makes the wider case. A manager's reading of behaviour isn't a neutral record, so it's worth checking.

What does the research show?

Nisbett and Wilson showed viewers two videotaped interviews with the same college instructor. In one he was warm and friendly. In the other he was cold and distant. Viewers of the warm version rated his appearance, mannerisms and accent favourably. Viewers of the cold version rated the same features unfavourably. The viewers didn't realise their overall liking had affected those ratings (Nisbett & Wilson, 1977).

For recency, the evidence is less tidy. Highhouse and Gallo had 117 trained student raters watch a candidate in two role-play exercises. They predicted that the later exercise would weigh more, and reported results consistent with that (Highhouse & Gallo, 1997). Rosales and colleagues asked 146 students to rate a fictitious office worker from five written work samples, one of them poor. The poor sample lowered the task ratings more when it came first than when it came last (Rosales, Díaz-Cabrera & Hernández-Fernaud, 2021).

So the order in which you see someone's work can change the rating. The direction isn't fixed: sometimes the latest events weigh most, sometimes the earliest. All three studies used staged material, and the two order studies used student raters. None involved managers rating their own staff. None tells us how often this happens in a real review, or how large the effect is. That's why the checks below use a dated record of the whole year, not only the last month.

How they show up in one review

Return to Priya, whose review shows how both can work at once. Illustrative scenario. The manager's draft says "Strong year. Sharp analysis." Here's the record the manager could rebuild from calendar entries and old messages.

PeriodWhat the record showsWhat the draft says
January to MarchLed a client analysis that the team still reuses.Carries the whole "strong year" line (a possible halo), including the ratings it has no link to.
April to SeptemberDelivered on time most months. Two reports needed a second pass.Not mentioned.
October to NovemberTwo deadlines missed during a system migration that changed her inputs.Rated "reliability" down to "mixed".

Two things are happening here. First, the January analysis may be lifting the other ratings, including those it has no link to. Second, the October and November misses may be outweighing six months of steady delivery.

As a result, the review says "strong" and "mixed" about one person, and neither reflects the year. We can't say which distortion is larger. That's the point: you can't tell from the inside.

Why it feels like a fair review

From the inside, nothing feels wrong. Each line feels backed by a real memory, and that feels like evidence. The October and November misses are vivid, and vivid events feel important.

That is why these habits are hard to catch by effort alone. The manager isn't careless. The manager is working from the memories that come most easily, and those aren't a fair sample of the year.

A review also carries weight beyond the page. It can shape a bonus conversation, a development plan or a promotion case. Therefore a distortion at this point can travel a long way.

Four checks before you submit

These are our working checks, not a validated method.

  1. Rate one dimension at a time. Cover your overall view, then rate quality, reliability and collaboration separately, citing a dated event for each. If every rating moves together, ask why.
  2. Read the record in both directions. Start with January once, then November. Notice what changes.
  3. Remove the last month. What would you write if it hadn't happened? If the answer differs sharply, the last month is carrying the review.
  4. Run it for someone you rate highly. Would the same late deadline have gone into their review? If not, a halo may be working on one of you.

None of these makes you neutral. They make the distortion easier to see.

Here is how the illustrative review might change. "Strong year" becomes "Strong analysis in Q1, steady delivery through September." "Reliability: mixed" becomes "Two missed deadlines in October and November, after the migration changed her inputs. Worth discussing together." The new versions are longer, but they give Priya something to respond to.

Keep the record through the year

All four checks need something to check against. Memory won't do, because the distortion acts on memory.

So a short dated line, written when something happens, is enough. "12 March: presented analysis to the client. Questions on method, answered." Also add context where you know it, for example: "October: inputs changed after the migration."

Observation vs Interpretation explains how to keep those lines to what you saw. Performance Is a Signal. Not a Diagnosis. shows what to do when the latest signal is a dip.

Common questions about halo and recency

Is recency always a bias?

Not always. Recent behaviour can show where someone is now, and that matters for a development plan. However, it becomes a problem when it stands in for the whole period. The same is true of the first months: one study found that a poor start weighed more than a poor finish. A review should say which period it is describing.

Does this make reviews pointless?

No. It makes the method matter. A review built on a dated record, rated by dimension, is easier to defend and easier to discuss with the person.

What about calibration meetings?

They add a second pair of eyes, which helps. They can also pass a halo along if each manager arrives with only a headline. Bring the dated record, not only the rating.

Where this leaves you

You can't remove these habits by knowing about them. You can make them visible by changing what the review rests on: a record built through the year, and ratings given one at a time.

A review that reflects the whole year treats the person fairly, and it's a better basis for the next decision.

Opptimise helps people leaders look at the whole pattern of a person's year, so an overall impression doesn't stand in for the evidence.

Rate the year, not the last month. Better understanding comes before better action.

Sources

Topics:#halo effect#recency bias#performance reviews#manager judgement
OE

Opptimise Editorial Team

Leading research and enterprise implementation in organizational psychology, executive development, and continuous Behaviour Intelligence.

Put Research Into Action

Ready to Embed Behaviour Intelligence in Your Organisation?

Explore how Opptimise turns behavioural science into actionable daily coaching for your leaders and teams.