UEQ calculator: User Experience Questionnaire analysis with the benchmark

The User Experience Questionnaire was built in Germany in 2005 by Bettina Laugwitz, Martin Schrepp and Theo Held, from 229 candidate items reduced by expert review and factor analysis to 26. Each item is a pair of opposite terms on a seven-point line; half of them print the positive term on the left, so that respondents have to read rather than tick down one column. The 26 items form six scales:

Attractiveness is pure valence. Perspicuity, efficiency and dependability are the pragmatic, goal-directed side; stimulation and novelty are the hedonic side. The handbook’s plain reading of a scale mean is that −0.8 to +0.8 is neutral, above +0.8 is positive and below −0.8 is negative — and that real studies almost never go above +2 or below −2, because people with different answer styles get averaged together.

The short version, UEQ-S

Eight of the 26 items — supportive, easy, efficient, clear for pragmatic quality; exciting, interesting, inventive, leading edge for hedonic — with the negative term always on the left. It gives two scales and an overall mean, not the six detailed ones, and its authors recommend it only where the full questionnaire will not fit: an exit survey, a slot inside a longer product survey, or several variants rated in one session. Paste eight answers per row and the page switches to it.

A mean of +1.2 says nothing until you know what other products get. The UEQ team publishes a benchmark built from hundreds of product evaluations and places each scale mean in one of five bands:

• Excellent — in the top 10 % of the benchmark;
• Good — 10 % of products are better, 75 % worse;
• Above average — 25 % better, 50 % worse;
• Below average — 50 % better, 25 % worse;
• Bad — in the bottom 25 %.

The borders below are the ones published in the UEQ Handbook (version 11) and used by the official Excel tool. A product must exceed a border to move into the band above it.

Two specialised benchmarks exist, and the page offers both: business software (158 evaluations) and web sites and web services (85). They matter — web perspicuity needs 1.84 to be Good where business software needs 1.48, and web novelty is Above average from 0.49. For the UEQ-S the published borders are pragmatic 0.72 / 1.17 / 1.55 / 1.74, hedonic 0.35 / 0.85 / 1.20 / 1.59 and overall 0.59 / 0.98 / 1.31 / 1.58.

Use the benchmark as the handbook does: per scale, as a statement about where the product sits among typical products. There is no single UEQ score to grade, and averaging the six scales into one — the thing most dashboards want — is exactly what the instrument’s authors say not to do without importance weights from your own users.

Online, a share of respondents click through. The UEQ team published two rules for spotting them, and the official Excel applies both; so does this page, and it lists who it removed so you can disagree.

• Items in a scale disagree. All items of a scale measure the same thing, so the best and worst answer inside one scale should be close. A spread over 3 points in one scale can be a slip; in three or more of the six scales it marks the response as suspicious. The strict setting removes at two.
• Too many identical answers. More than 15 of the 26 answers in the same circle — usually the middle one — is not a considered response.
• UEQ-S. Both scales over the spread limit, or all eight answers in the middle.

The identical-answers rule counts the circle actually ticked, not the recoded value, which is why somebody who loves the product and ticks the positive end everywhere is not removed: because the questionnaire alternates sides, their answers are split between 1 and 7.

• One number hides the finding. The UEQ is designed to give six values, and the interesting result is usually the pattern — high attractiveness and low perspicuity is a product people like and cannot learn. Report the profile.
• Overlapping intervals are not “no difference”. If two products’ intervals do not overlap, the difference is significant at the chosen level; if they do overlap, it still may be. Compare two versions with a proper two-sample test, not by eye.
• A weak alpha has two usual causes. An item that means something else in your context — the handbook’s example is secure / not secure in a social network, read as “is my data safe?” — or a scale that simply does not apply to the product. Below 40 respondents alpha is itself noisy, so look at the item means first.
• Stakeholders balk at the scale names. “Stimulation” and “Novelty” sound odd for an accounting system. The handbook explicitly allows renaming them in a report (“fun of use”, “interest”) as long as the meaning is kept.
• Edited items break the benchmark. Change or drop a single item and the benchmark no longer applies. Drop a whole scale if you must.
• Ask after the discussion and the answers drift. Collect the questionnaire right after the tasks, before the debrief.
• Launch-week data is about change, not quality. Right after a redesign people rate the disruption. Wait until they are used to the new version.

The UEQ suits software well because it separates two things product teams often blur: whether people like the product and whether they can get on with it. A few practical notes from running it on web apps:

• Perspicuity is the onboarding scale. Its items — understandable, easy to learn, easy, clear — are exactly what a first session either delivers or does not. A product whose attractiveness is Good and perspicuity Below average is telling you the first-run experience is the problem, not the product.
• Survey new users and established users separately. Perspicuity rises with familiarity, so a mixed sample flatters it. Split by tenure and compare.
• Pair it with behaviour. UEQ tells you how it felt; time to value, the activation funnel and feature adoption tell you whether that turned into use.
• Know its siblings. The System Usability Scale is one number for usability with its own grade curve; NASA-TLX measures how heavy a single task felt. The UEQ is broader than SUS and coarser than TLX: use it when you need the hedonic side too.
• Use the UEQ-S in-product. Eight items fit in a short survey after a key task; the full version belongs in usability tests and periodic studies.

1. Laugwitz B., Schrepp M., Held T. “Construction and evaluation of a user experience questionnaire.” In Holzinger A. (ed.), USAB 2008, LNCS 5298, 63–76 — the instrument, the six scales and their validation.
2. Schrepp M., Hinderks A., Thomaschewski J. “Construction of a benchmark for the User Experience Questionnaire (UEQ).” International Journal of Interactive Multimedia and Artificial Intelligence 4(4), 2017, 40–44 — how the benchmark bands are built.
3. Schrepp M., Hinderks A., Thomaschewski J. “Design and evaluation of a short version of the User Experience Questionnaire (UEQ-S).” IJIMAI 4(6), 2017, 103–108.
4. Schrepp M. User Experience Questionnaire Handbook, version 11, 2023 — the reading of scale means, the general, business-software and web benchmark borders, the agreement bands for standard deviations and the cleaning heuristics.
5. Schrepp M. “Datenqualität bei Online-Fragebögen sicherstellen.” Mensch und Computer 2016 – Usability Professionals; and “Enhancing the UEQ heuristic for data cleansing by a threshold for the number of identical responses,” research report, 2023.
6. The UEQ, its translations and analysis tools are provided free of charge by the UEQ team at ueq-online.org. The scoring, recoding, cleaning rules and benchmark borders on this page were checked against the official English questionnaire and the version 15 Excel analysis tools, and reproduce their scale means and alphas on the tools’ own example data. This page is not affiliated with or endorsed by the UEQ team.