UX research: usability testing for websites and apps

Questions this report answers

Can users complete key tasks such as finding a product, signing up or paying, and how long does each take?
Where exactly do they hesitate, make errors or give up?
Do our navigation and labels match the way users think about the content?
How does our usability compare with a competitor or with the previous version?
Which problems should developers fix first?
Does a new prototype perform better than the live site before we build it?
Are there barriers for older users or people using assistive technology?

When you need it

UX research is most valuable at three points: before a redesign, to set a baseline; during design, to test prototypes while changes are cheap; and after launch, to confirm the new version works. Other triggers are a step in your analytics where users drop out, support tickets that keep mentioning the same feature, low customer satisfaction scores for the app, or a move into a new market.

Audiences differ in digital experience. According to Eurostat, 89.5% of EU individuals used the internet daily in 2025, against 82.7% in Bulgaria and 98.4% in the Netherlands. A service tested only on frequent users may fail for the rest, so recruit to match the real user base.

Accessibility now carries legal weight as well: the European Accessibility Act has applied to many consumer-facing digital services, including e-commerce and banking, since June 2025.

To choose between two live variants on real traffic, run an A/B test and check the result with our A/B test significance calculator. Whether people want the product at all is a question for concept testing, and problems spanning several channels call for customer journey research.

Sample table of contents

Executive summary
the top issues, overall verdict and quick wins
Objectives and tasks
what was tested and how success was defined
Participants
recruitment criteria, screener and profiles
Method
moderated or unmoderated, devices, prototype or live site
Task results
success rate, time on task and errors per task
Usability issues
each with severity, frequency and evidence
Information architecture
card sort and tree test results
SUS and ease scores
against the benchmark or the previous version
Accessibility observations
barriers found during sessions
Recommendations
changes ranked by severity and effort
Highlight reel
short clips of the most telling moments

Methodology and sources

Moderated usability tests

Participants think aloud while working through realistic tasks with a moderator, remotely or in a test room. For finding problems, five to eight participants per distinct user group and round are usually enough. Several small rounds with fixes in between find more than one large round. Mobile and desktop are tested separately when behaviour differs.

Unmoderated remote tests

Larger groups complete tasks on their own devices, which suits benchmarks of task success and time. Small samples give wide intervals: with 20 users, a 50% success rate carries a margin of error of about ±22 points at 95% confidence, falling to about ±10 points with 100 users.

Card sorting and tree testing

Open card sorting reveals how users group content and what they call it; closed sorting checks a proposed structure. Tree testing measures whether people can find items in a menu stripped of visual design, and it needs larger samples than moderated tests.

Standard measures

The SUS has ten statements on a five-point scale and produces a score from 0 to 100; 68 is the widely used average reference. A single ease question after each task shows where difficulty concentrates.

Recruitment and ethics

A screener selects real or target users, never colleagues. Sessions are recorded only with informed consent, stored securely and deleted after an agreed period, in line with GDPR.

What you receive

Usability report with every issue rated for severity
Highlight video clips, shared with participants' consent
Task metrics in Excel: success, time and errors
Card sort similarity matrix and tree test results
SUS score with benchmark comparison
Annotated screens with design recommendations
Debrief workshop with design and development teams

How the project runs

Kick-off, goals and task definition — 2–3 days
Recruitment and screening — 1 week
Test script and a pilot session — 2–3 days
Test sessions — 2–5 days
Analysis and severity rating — 3–5 days
Debrief and iteration plan — 1 day

Hypothetical example for illustration

Worked example

Example: a hypothetical SaaS company that sells invoicing software to small businesses notices that many trial users never send their first invoice. Product analytics show where they stop, but not why.

The company runs moderated usability tests with eight bookkeepers and eight small-business owners on desktop, plus a remote tree test of a proposed new menu. The sessions reveal three blockers: the VAT set-up screen uses accounting jargon, users confuse "Clients" with "Contacts", and the product asks for bank details before the first invoice can be created. Business owners give a SUS score below the 68 reference, while bookkeepers find the product easy.

The team moves the bank-details step to after the first invoice, merges the two menu items and rewrites the VAT screen in plain language. In a second round on the revised prototype, all six business owners complete the first invoice; the changes then ship behind an A/B test on trial activation.

Common mistakes and how to avoid them

Testing with colleagues or friends. They know the product and the jargon; recruit real users with a screener.
Leading tasks. "Click on Products" gives the answer away; describe the goal and let users find the path.
Testing once, at the end. Problems found after development are costly to fix; test early prototypes and iterate.
Reporting opinions instead of behaviour. "I like the colours" is not usability evidence; watch what users do.
Quoting percentages from five users. Small qualitative rounds find problems; they do not measure their prevalence.

Who it suits

Suitable for	Why
E-commerce and retail websites	Search, product pages and checkout decide conversion
SaaS and fintech products	Onboarding and first use determine whether trials convert
Banks, insurers and telecoms	Self-service portals that work reduce calls to the service centre
Public-sector digital services	Services must work for everyone, including users with disabilities
Teams planning a redesign	Baseline and prototype tests cut the risk of a failed relaunch

Frequently asked questions

How many users do we need for a usability test?

For finding problems, five to eight participants per user group and round usually suffice. For benchmarks that report success rates or SUS scores with confidence intervals, plan several dozen participants or more. Our margin of error calculator shows how precision changes with sample size.

How long does a UX research project take?

A single round of moderated tests typically takes two to three weeks from kick-off to debrief, with recruitment as the longest step. Projects combining card sorting, tree testing and two test rounds take four to five weeks. Hard-to-reach users, such as clinicians or finance directors, need extra recruitment time.

What is the difference between UX research and A/B testing?

UX research explains why users struggle, using a small number of observed sessions. A/B testing measures which of two live versions performs better on real traffic, but it does not explain the result. Used together, research suggests the improvements and an A/B test confirms their effect.

What affects the cost of UX research?

The number of user groups and rounds, the difficulty of recruiting the right participants, moderated versus unmoderated sessions, the number of devices and languages, and whether card sorting or tree testing is added. Outline your product and request a quote.

How do you handle recordings and personal data?

Participants receive clear information and sign a consent form before recording starts. Recordings and notes are stored on secure systems, shared only with your named team and deleted after the agreed retention period. We act in line with GDPR and the ICC/ESOMAR International Code.

Can you test a prototype before anything is built?

Yes, and it is the cheapest moment to find problems. We test clickable prototypes from design tools, and even paper or static mock-ups for early navigation questions. Prototype tests also let you compare two design directions side by side before development starts.

Sources

Updated: 5 Nov 2026

UX research: usability testing, card sorting and SUS benchmarks for websites and apps