Skip to main content
Free plan, no cardStart free

UXUX testing · updated 2026-10-25

UX testing tool: usability and UI checks on the journeys your testers record

Compare two recordings of a website and get every control measured against WCAG 2.2 and usability rules, friction signals, lab vitals, a journey outcome, a gate verdict and a defect ledger.

For: Product designers, front-end teams and QA leads who want usability defects found before a release with a cited standard on every finding, and teams with no budget for a recruited-user study on every change.

acme-shop / checkout-flow.recIllustrative
1 selector healed#pay-now-btn[data-test="pay"]

Capabilities

Every control measured against a cited standard

Target size (WCAG 2.2 SC 2.5.8, 24 by 24 CSS pixels with the inline-link exception), accessible name (4.1.2), generic link text (2.4.4), placeholder-only labels (3.3.2), redundant entry (3.3.7), contrast on the control’s own background (1.4.3), a missing image alt attribute and the document language (3.1.1). Each mark carries its severity, P0 to P3, and whether it is deterministic or heuristic.

Friction read from the session, not asked in a survey

Rage clicks (three or more in two seconds), dead clicks, error clicks (a click followed within a second by a JavaScript error), hesitations of ten seconds or more, corrections, Back navigations, validation messages, first-party API failures and quick returns to the previous page, counted per page with the definitions published.

Usability rules from the geometry

Label-to-control proximity, left-edge alignment, choice count (Hick: at most seven primary actions), fields before a step (Miller), one dominant call to action (Von Restorff), feedback within 400 ms (Doherty), consistent help across pages (3.2.6), a derived palette and type scale with their drift, and a streamlined cognitive walkthrough of each step: label clarity, discoverability, feedback.

Lab vitals and a journey outcome

LCP, INP, CLS, FCP and TTFB rated good, needs improvement or poor at the web.dev thresholds, stated as one lab sample rather than a field 75th percentile; the journey is judged likely success, likely failure or unknown from a published decision table, with its confidence.

Regression, defect or fixed, with a reviewer in the loop

A mark that passed in the base and fails in the target is a regression; a failing mark is a defect at P0 critical, P1 warning, P2 and P3 info; a mark that cleared is reported fixed. Accept files a bug in Project Management with the evidence attached; Reject records the change as intended, and the decision is remembered on later runs for as long as the evidence stays the same.

Read more: Regression, defect or fixed, with a reviewer in the loop

A gate verdict, a Definition of Done and website-level evidence

Each run carries a verdict (pass, tracked or fail) under the policy you choose, hard gate, baseline and track, or track only, counting new violations apart from known ones. Nine Definition-of-Done items are read from evidence (met, not met, needs a live audit, needs a design file, unknown), and the website block shows device and engine coverage, task success with a 95 percent interval, time on task, edge states exercised, a debt ledger with closure within 90 days, and the reviewer-approved baseline.

How it works

  1. Record the journey twiceRecord the flow before and after a change with the test recorder, on any plan; the validation needs two recordings of the same website.
  2. Pick the base and the targetOn Visual Testing, open UI/UX Validation, choose the website and the two recordings. The automatic run compares a new recording against the reviewer-approved baseline, or against the previous recording until one is approved.
  3. Run and read the marksEvery page lists its controls with the checks that failed, the friction counts, the vitals rating and the journey outcome; the run shows what regressed, what is a defect and what was fixed.
  4. Decide each findingAccept to file a bug in Project Management with the evidence; Reject to record that the change was intended. The decision carries to later runs while the evidence is unchanged.
  5. Read the verdict and the ledgerThe gate verdict, the change class with the frameworks it calls for, the bounded live audit of up to two pages, the Definition of Done and the website evidence sit under the summary. Your CI job polls the run and decides what blocks.

What you receive, and what we do not do

Evidence ledger
Per control
The check that failed, its cited standard and oracle, severity P0 to P3, a deterministic or heuristic basis, the before and after values, and a screenshot box where one exists.
Per page
Steps, time on page, the friction counts with their definitions, the five vitals with their rating and the lab caveat, the usability rows from the geometry, the walkthrough answers and the journey outcome with its confidence.
Per run
Regressions, defects and fixes with deterministic ids; each finding’s defect category, cited standard, users affected and priority; the gate verdict with new and known counts; the change class and its reason; the Definition of Done item by item with the evidence beside it.
Per website
Viewport and engine coverage with the gaps named, task success with an adjusted-Wald 95 percent interval and its n, time on task p50 and p75, edge states exercised, the debt ledger (first seen, last seen, resolved, open age, closed within 90 days) and the approved baseline.

Limits we state

  • The people in the recordings are your testers, not recruited users: task success and time on task are a lab proxy, said so on every run, and no SUS or CES score is computed, because those need people answering questions.
  • Two recordings of the same website are needed; a site with one recording shows nothing to compare.
  • Vitals are one lab sample from the recording, not a field 75th percentile; Google grades on real visitors.
  • Hover, disabled and loading states cannot be judged from a recording and are marked as needing the live audit, which runs on at most two pages per run; design-file fidelity is marked as needing a design file, because none is connected.
  • The gate verdict is data on the run, not a mechanical block: a CI job that polls the run decides what fails.
  • Runs made before 2026-10-24 carry no behaviour, gate, plan or website evidence, and the tab says so.

Plans: The UI/UX Validation tab is on the Visual Testing page; visual regression is included from Solo Pro and sold alone as a single tool on Free. The recordings it compares come from the test recorder, which is on every plan including Free. Filing a bug uses Project Management, included from Solo Pro and a single tool on Free. See the metering table.

Questions buyers ask

Is this a usability testing tool?

Partly. A usability testing tool in the research sense recruits people, sets them tasks and measures how they fare; Dokimore measures the journeys your own testers recorded, so task success, time on task and the friction counts are a lab proxy, labelled as such on every run, and no SUS or CES score is produced. What it adds is a cited standard on every finding, from WCAG 2.2 target size to Nielsen’s severity scale, so a usability defect is a reviewable fact rather than an opinion.

What is the difference between UI testing and UX testing here?

UI testing checks that the interface works: a recorded flow replays as a regression test and each step passes or fails. UX testing checks whether it can be used well: the same recording is measured control by control against usability and accessibility rules, the session’s friction signals are counted, and the journey’s outcome is judged. One recording feeds both; the regression testing page covers the first.

Does it replace an accessibility audit?

No. It checks the WCAG 2.2 criteria that apply to the controls a tester used, such as target size, accessible names, link text, contrast and image alt attributes. The accessibility module audits whole pages with axe-core against WCAG 2.2 A and AA and produces a VPAT draft from the scan; the two read the same recordings.

Can it block a release?

It can tell your pipeline to. Every run carries a verdict of pass, tracked or fail under the policy you choose (hard gate, baseline and track, or track only), counting new violations apart from known ones. The verdict is written on the run and into the notification; a CI job that polls the run decides whether to fail the build.

Which standards and sources does it cite?

WCAG 2.2 success criteria by number, the web.dev Core Web Vitals thresholds, Nielsen’s severity ratings, the Gestalt and UX laws it measures (Hick, Miller, Von Restorff, Doherty), Spencer’s streamlined cognitive walkthrough, the adjusted-Wald interval of Sauro and Lewis for task success, and the WebAIM Million 2026 shares for the six most common accessibility errors. Each row names its oracle, so a reviewer can check the rule and not only the verdict.

What happens when I reject a finding?

It is recorded as intentional and does not come back on later runs of that website while the evidence behind it is unchanged. If the finding changes, for example the same control fails a different check or the after value moves, it is proposed again, names the earlier decision, and counts as new for the gate.

Try it on your own application

The Free plan asks for no card and runs for 14 days after the trial. Or bring one flow from your product and we record it live.

  • Free plan, no card
  • 14-day Solo Pro trial
  • Demo replies within one business day
  • Data hosted in India