Halo effect

An employee scored 9 out of 10 for communication skills. Three questions later, they are asked to rate that same person's technical skills — and again it is 8 out of 10, even though they actually perform at an average level.

A general positive impression spills over onto every aspect — that is the halo effect. In surveys this effect turns detailed ratings into a repetition of the same overall sympathy, masks the real strengths and weaknesses, and makes the data less useful for decision-making.

Definition

Halo effect is a cognitive bias in which a general impression of a person, brand, or product influences how its individual characteristics are rated. Positive halo: an object perceived as good on one attribute is rated higher on others as well, even unrelated ones. The negative version (reverse halo, horns effect) mirrors this in the opposite direction. In surveys it leads to an artificially inflated correlation between ratings of different aspects of the same object.

How the halo effect shows up

In HR appraisal. A manager rates a subordinate across 10 competencies. If the overall attitude is positive, the scores on most items will be high — even on those where the person is objectively average. This makes 360-degree feedback less informative: a detailed profile collapses into an overall sympathy score.

In product evaluation. A customer loves the brand → gives high ratings on every aspect: speed, design, pricing, support. Even on parameters where the product is genuinely behind competitors. The reverse effect: one negative experience (a delivery delay) colours all subsequent ratings in the same tone.

In rating an employee after a successful project. An employee closes a major deal → the semi-annual appraisal looks excellent across the board, including discipline and paperwork, where there are actually problems.

In comparative surveys. A respondent is asked to rate one product, then another. If they liked the first, the halo carries over: "well, this one is decent too," even without an objective comparison.

The psychological mechanism

The halo effect was described by Edward Thorndike in 1920 while analysing ratings of military officers. He noticed that the ratings of one person across different competencies correlated far more strongly than objective reality would allow. Discipline, intelligence, physical fitness, and loyalty received similar scores — as if the rater did not distinguish between these aspects.

The mechanism is tied to cognitive economy. Rating each parameter independently is hard: it requires recalling specific examples, comparing them with a benchmark, and excluding the overall attitude. The brain simplifies: it takes the general impression and extends it to all the sub-items. The result is a fast but biased rating.

Halo effect vs other biases

The halo effect overlaps with other cognitive biases in surveys:

• Anchoring — being anchored to the first number you see. The halo is broader: the influence of a general impression rather than a specific number.
• Primacy effect — the first piece of information carries more weight. The halo uses the first impression as the basis for all subsequent ratings.
• Acquiescence bias — the tendency to agree. The halo works differently: not "agreeing," but "rating higher" (or lower) because of the overall attitude.

All of these effects are forms of systematic bias, and they often act at the same time, reinforcing one another.

Example: the halo effect in a service evaluation

A restaurant chain runs a survey: guests are asked to rate 8 parameters — service speed, food quality, atmosphere, cleanliness, prices, staff courtesy, the menu, and the restroom. The scale is 1-10.

When analysing the data, they found that all 8 ratings were strongly correlated (an average correlation of 0.72). At first glance, the restaurants really are good or bad "across the board." But on closer inspection: at restaurants with low scores, the "restroom cleanliness" and "service speed" parameters received ratings of 6-7 — clearly higher than the real values (which were measured independently through mystery shopping). Halo effect: a general bad impression lowered all the ratings, but a general good one, conversely, raised them uniformly. The breakdown by parameter turned out to be less useful than expected — it effectively duplicated the overall rating.

How to reduce the halo effect in surveys

Separate the ratings of different objects or blocks. Rate each aspect on its own, separating them with other questions or pages. This gives the respondent a pause and reduces the "inertia" of the general impression.

Use concrete wording with anchors. Instead of "Rate the quality of service 1-10," use "Over the past month, how often did the staff member help you solve a problem — never / sometimes / often / always?" Concrete behavioural indicators weaken the effect of the general impression.

Separate positive and negative wording. If all the questions follow the same tone ("How much do you like..."), the halo is amplified. Alternating positive and negative wording ("What frustrated you?") forces the respondent to stop and think separately about each item.

Randomise the order. Randomising questions and answer options across respondents reduces systematic distortion — the halo is diluted across the sample, though it remains in individual responses.

Ask about the overall impression separately. Instead of letting the halo "seep" into the detailed ratings, pull the overall impression out into a separate question. Then the detailed ratings become more differentiated.

When the halo effect is a useful signal

In some tasks the halo does not get in the way but helps. For example, for measuring brand perception or overall loyalty, the general impression is exactly what you need to measure. Breaking it down into components is not always meaningful.

The halo problem arises when you want a differentiated rating — to understand the specific strengths and weaknesses rather than the overall attitude. In that case the halo masks precisely what you are trying to find.

Halo effect in SurveyNinja

When designing surveys in SurveyNinja, account for the halo effect at the questionnaire-design stage: use randomisation of question order, separate rating blocks with meaningful pauses (a welcome screen between sections, open-ended questions), and alternate the wording. After collecting the data, a high correlation across all items (0.7+) is a signal of the halo effect, not necessarily of real uniformity in the ratings.

The halo effect is when a general impression "paints" all the detailed ratings in a single tone. Sympathy makes all the ratings higher, antipathy makes them lower, regardless of the reality of the specific parameter. The result: a detailed profile turns into the overall rating repeated many times over. Minimise it through randomisation, concrete wording, and alternating the tone of the questions.

Frequently asked questions

How can I tell whether my data contains a halo effect?

Calculate the correlations between the ratings of different parameters of the same object. If the average correlation is above 0.7 and the parameters are substantively different, it is most likely a halo. An additional sign: low differentiation in one respondent's ratings (most parameters receive similar values).

Can the halo effect be eliminated completely?

No. It is a fundamental cognitive bias that arises in most people in most contexts. You can only reduce its influence through proper survey design: separating questions, concrete wording, randomisation, and alternating the tone. Completely "clean" ratings do not exist in surveys.

How does the halo differ from a stereotype?

A stereotype is a prior expectation about a group of people or objects of a certain category. The halo is the spread of an impression from one observed quality to others. A stereotype works before contact, the halo afterwards. In surveys both biases can act at the same time.

Does the halo affect only positive ratings?

No. There is a reverse effect — the horns effect — when a general negative impression spreads to all the ratings. One bad experience, a communication slip, or an unsuccessful packaging design can lower the ratings of all the other parameters, even objectively good ones.

Does the halo effect work only with brands and people?

No, it works with any objects of evaluation: products, services, ideas, cities, politicians, films. Wherever there is a multidimensional rating of a single object, a halo is possible. In UX research the halo is often linked to the impression of the visual design: a beautiful interface is perceived as more usable, even if the usability is actually average.