UX-Check 360° · Expert review, double AI-validated
Check first,
then test.
Before you put budget into a user test: I find the weaknesses in your website, app or booking process that can already be seen today. Every finding is prioritised and backed by a source. Your user test then only deals with the open questions.Before you put budget into a user test: I find the weaknesses you can already see today. Prioritised and backed by sources.
- Fixed price
- Results in 7 working days
- No participant recruitment needed
Sample finding from the “Family holiday in South Tyrol” case (schematic)
Selected filters are lost as soon as the travel dates are changed
Price and amenity filters have to be set again after every date change. Anyone planning flexibly and comparing several periods has to start over each time.All filters must be reset after every date change.
- UX expert
- AI review A
- AI review B
Don’t burn budget
If every participant gets stuck on the same broken filter, you’ve paid a lot for an insight that was obvious beforehand.If everyone gets stuck on the same error, you pay for an obvious insight.
Deeper test results
Major errors mask the subtle ones. Anyone who drops out early never reaches the points where a user test reveals something genuinely new.Major errors mask the subtle ones, and with them what a test could reveal.
More targeted testing
The check delivers the hypotheses too. The test tasks target exactly where findings are disputed or their weight is unclear.The check delivers the hypotheses. Testing focuses on what’s disputed.
Whatever is already proven doesn’t need testing
Many UX questions were answered long ago by studies from Nielsen Norman Group, the Baymard Institute or the GOV.UK Design System. For example, whether a submit button may be disabled while required fields are missing, whether a date of birth should be entered via a calendar picker, or whether the desktop navigation may disappear behind a hamburger menu. The answer in each case: rather not. I check such points directly against the evidence. Your user test stays free for the questions only your users can answer.Much is already proven by studies from NN/g, Baymard or GOV.UK. I check it directly. Your test stays free for questions only your users can answer.
The UX-Check is the starting point, not a replacement
- Stage 1 UX-Check 360° Find, prioritise and substantiate weaknesses.
- Stage 2 Optimise Implement quick wins, with a clickable prototype of the critical screens if you like.
- Stage 3 User test Check with real users what is genuinely open, on the optimised version, not the old one.Real users check what’s open, on the optimised version.
Expert eyes and AI-assisted analysis complement each other: fewer blind spots, faster results, at a fixed price.
- 01KickoffWe clarify goals, users, process and constraints, and define the test tasks.
- 02AnalysisI review your process step by step from your users’ perspective, supported by AI as an independent second opinion.Step by step from the user’s view, with AI as a second opinion.
- 03ValidationI check, weight and source every finding. Whatever doesn’t hold up is dropped.
- 04Results and handoverA prioritised report, the findings as a backlog for your team and a meeting to agree the next steps.Prioritised report, backlog and a meeting on next steps.
A review is only as good as the questions it starts with
In a conversation of about 60 minutes, we lay the groundwork. Your knowledge feeds directly into the assessment. So I don’t review by the book, but by what really matters to your users and your business.In about 60 minutes we lay the groundwork. I review what matters to your users and your business.
At the end of the kickoff, we have agreed: the jobs to be done and user groups, the process and channels in focus, known problems, the competitors to compare against and the concrete test tasks, for example: “Family holiday in South Tyrol, Whitsun holidays, 7 nights, max. €2,000”.
- Prioritised list of findings, by severity in three levels: minor, moderate, severe.
- Every finding with a screenshot, mapped to page, state and subtask.
- Backed by sources: heuristics, standards and studies, so internal debates aren’t about taste.
- What works well, too, so it isn’t accidentally lost in the redesign.
- Sorted by impact and effort, so you know what to tackle first. Quick wins are flagged.
- As a backlog to work with: all findings also as an Excel list for your team.
- Recommendations for the user test: which questions remain open and how to test them.
“Family holiday in South Tyrol” case, two travel portals compared · schematic
| Subtask | Portal A | Portal B |
|---|---|---|
| Enter destination and family occupancy | good | minor |
| Price and amenity filters | moderate | good |
| Filters persist when dates change | severe | good |
| Compare hotels | minor | moderate |
| Total price transparent | good | severe |
| Booking up to checkout | moderate | minor |
Examples from the “Family holiday in South Tyrol” case · schematic
Even experienced experts have blind spots. That’s why I have two AI models assess independently of me and of each other: one without industry knowledge, as unbiased as a first-time user, and one that works with expertise from your domain.Even experts have blind spots. So two AI models assess independently: one without, one with industry knowledge.
In benchmarking, each AI rates each competitor without memory, so no portal colours the next.
In the end, I decide as a human. Every finding in the report is checked, weighted and sourced by me. From more than 20 years of UX research, I know which findings really cost revenue in practice.In the end, I decide: every finding is checked, weighted and sourced by me.
AI models are fast, but not infallible. They sometimes invent problems or sources. Three fixed rules ensure that only robust findings reach you.AI sometimes invents problems or sources. Three fixed rules ensure only robust findings reach you.
One scale for all
Expert and AI rate against the same predefined criteria, so findings can be compared directly.
A fixed finding format
Every finding states page, subtask, problem, heuristic, severity and recommendation. Nothing stays vague.
Every source checked by me
No evidence without verification. I check every heuristic, standard and study before it goes into the report.
How severity is rated
Irritates or slows down, but users get on without help.
Leads to detours or errors. Some users give up here.
Prevents the task or leads to costly wrong decisions, for example a misunderstood total price.Prevents the task or leads to costly wrong decisions.
Frequency, impact and whether users can work around the problem are assessed, based on Nielsen’s severity rating.
Mini-Check
One subtask, to get to know each other
- Short kickoff (30 minutes)
- One subtask, e.g. filters or checkout
- Expert walkthrough and 2 AI reviews
- Top findings with severity and quick wins
- Fully credited if you book a Check
Check
One process, one channel
- Kickoff call
- Video documentation of all pages and states
- Expert walkthrough and 2 AI reviews
- Prioritised report with sources
- Impact/effort matrix
- Backlog as an Excel list
- Results presentation (online)
Benchmark
Your process compared with 2–3 competitors
- Everything in “Check”
- The same tasks on all portals
- Benchmark matrix by subtask
- Strengths and weaknesses side by side
- What you can adopt from competitors
- Where you’re ahead and how to protect that lead
Add-on modules
Optimised wireframes and a clickable prototype of the critical process steps. Ready to use as the test object for the user test.Wireframes and a clickable prototype of the critical steps, ready to test.
After your redesign, I check whether the problems are fixed, with a before/after comparison you can show internally.After the redesign: are the problems fixed? With a before/after comparison.
Projects for Amazon, Audi, BMW, Volkswagen, Commerzbank, ING-DiBa, comdirect, ZDF, Sky and ProSiebenSat.1, among others. Plus interim roles in the automotive and travel industries.Projects for Amazon, Audi, BMW, Volkswagen, Commerzbank and ZDF, among others.
No, and that’s not the aim. The check uncovers what can be identified with expert knowledge and heuristics. How real users think, feel and decide is only shown by a user test. The check makes sure that test is worth its budget.No. The check finds what experts can identify. How users think and decide is only shown by a user test.
The AI delivers suggestions, not results. I check every finding and verify every source. Anything that can’t be traced is dropped.
Public competitor websites are uncritical. For internal systems, we agree the approach in the kickoff.
About 60 minutes for the kickoff, access to the system (for non-public areas) and everything you already know about your users. I’ll take care of the rest.
Websites, shops, booking and application flows, configurators, apps and customer portals, whether desktop or mobile, live or as a prototype.
Individual reviewers typically miss some of the problems. Several independent assessments increase coverage. The AI reviews make this fast and affordable, and the weighting stays with the expert.Individual reviewers miss problems. Several independent assessments find more; the weighting stays with the expert.
Let’s start with the kickoff.
In a free initial 20-minute call, we clarify whether the UX-Check is the right starting point for your project. If it isn’t, I’ll tell you.Free 20-minute call: is the UX-Check the right start? If not, I’ll tell you.