A UX audit is worth running only if it ends in a backlog: every finding tied to a flow, a severity rating and a named owner, the way a building inspection report works.
An inspector who says the place feels dated changes nothing; one who logs a cracked beam, its location, its risk and who repairs it gets work done. The right inspection also depends on the trigger, whether a metric drop, a redesign, accessibility pressure or one leaking flow.
What is a UX audit and what does it end in?
A UX audit is an expert review of usability, accessibility and conversion paths, done without live participants, that ends in severity-rated findings and a prioritized backlog. It combines heuristic evaluation, cognitive walkthrough, analytics funnel analysis and a WCAG 2.2 AA check, applied to the key conversion flows.
A finding is only schedulable when it names a flow, a severity rating and an owner. We recommend choosing methods by trigger, because that is what decides which ones are worth the reviewers' hours. If the open question is whether users understand a new concept, a test with participants answers it better than an audit.
The trigger sets the method and the output:
Which audit type fits which trigger?
Audit type follows the trigger: funnel analysis for a metric drop, heuristic evaluation before a redesign, a WCAG 2.2 AA check for accessibility pressure, a cognitive walkthrough for one leaking flow. Each trigger also needs a stop condition, so the audit ends when its question is answered.
When triggers overlap, run the funnel review first. The funnel review is the cheapest method, and it shows which flows deserve the slower manual ones. Finish the audit before the redesign starts, so the new design does not repeat the old problems.
Funnel analysis assumes event tracking exists on the flow. If it does not, fall back to heuristic evaluation and add the missing events to the backlog.
Teams that want an outside reviewer can see how this is scoped in Netguru's UX review service.
How is a UX audit different from usability testing?
A UX audit is an expert review of a live product, while usability testing watches real participants attempt tasks.
A heuristic evaluation checks key conversion flows against Nielsen's heuristics, and it needs no recruiting or scheduling, so it's the cheaper first pass. Testing is the better tool for questions a reviewer cannot answer from the screen, such as whether buyers understand a label or why they pick one path over another. Our usability testing guide covers that method.
We recommend running the audit first, then testing only the questions it leaves open. Reviewers can settle contrast, focus order, and inconsistent labels. Participants are needed for comprehension and motivation.
Skip the audit in two cases:
- A funded rebuild that replaces every flow. If nothing carries over, there is little to review. If even one flow carries over, audit that flow first so the new design does not repeat its problems.
- No users yet. With no live product and no analytics, there is nothing to evaluate. Test a prototype with participants instead.
What are the six steps of a UX audit?
A UX audit runs in six steps (kickoff, scoping, access, analysis, report, and readout) and ends in a prioritized backlog where every finding has a flow, a severity, and an owner.
- Kickoff: Name the trigger (metric drop, redesign, accessibility pressure, or one leaking flow) and the person who will accept the findings.
- Scoping: Fix the key conversion flows, devices, and WCAG 2.2 level in writing. Ask which flows carry revenue, which breakpoints matter, and who owns each flow's backlog.
- Access: Collect the items in the table below before analysis starts.
- Analysis: Run a heuristic evaluation with two or more reviewers, a cognitive walkthrough per flow, analytics funnel analysis, and a manual WCAG 2.2 AA check. Capture a screenshot or recording timestamp for each finding.
- Report: Rate each finding 0-4 and attach a screen, heuristic, effort estimate, and owner.
- Readout: Walk product and engineering through the findings, then leave with tickets in their tracker.
Access requests cause most stalls, because funnel review cannot start without event data. The checklist:
If any row is missing at the end of scoping, we recommend recording the gap as a limitation in the report.
How do you run a heuristic evaluation with two reviewers?
A heuristic evaluation with two reviewers works when each person scores the same flows against Nielsen's 10 usability heuristics alone, then the two lists are merged under fixed rules.
A single reviewer misses issues because attention follows background. A designer tends to log layout and labeling problems, while a product manager tends to log flow and logic problems. Nielsen Norman Group recommends several evaluators, because their findings overlap only in part.
- Align the setup. Give both reviewers the same flows, devices, and log template: finding, screen, heuristic, severity, and evidence (screenshot or recording timestamp).
- Review alone. No comparing notes until both logs are closed. Early discussion makes the second reviewer anchor on the first one's findings.
- Merge and dedupe. Pool both logs in one sheet and match entries by screen and root cause, not by wording. Two entries on one cause become a single finding that lists both heuristics. Keep findings only one reviewer logged, and ask the other to confirm them on the screen.
- Resolve rating conflicts. Each reviewer rates on the 0-4 scale before merging. A one-point gap takes the higher rating. A gap of two or more points gets reviewed with the screen open, and the rating that stands is the one with the clearer stated user impact. The flow owner breaks any remaining tie.
The NN/g severity guidance is a useful reference for the rating discussion.
Two reviewers from the same discipline share the same blind spots, so we recommend pairing design with product or engineering. The output is still expert judgment, so findings about user behavior need confirmation from the analytics funnel.
How does a cognitive walkthrough find first-time-user problems?
A cognitive walkthrough finds first-time-user problems by stepping through one of the key conversion flows as a newcomer and asking four questions at every step. Pick the flow where the funnel leaks, usually signup, onboarding or checkout, and write down the user's goal before opening the first screen.
The product team makes poor first-time reviewers because of familiarity bias. People who built the flow know which icon is tappable and where "Continue" leads, so they pass steps a newcomer would stall on. Give the walkthrough to a reviewer who has not worked on the flow. If that is impossible, have the team answer only from the persona's goal and prior knowledge.
At each step, the reviewer answers four questions yes or no:
- Will the user try to achieve the effect this step produces?
- Will the user notice that the correct action is available?
- Will the user connect that action with the effect they want?
- After acting, will the user see progress toward the goal?
Every "no" becomes a finding. Record the flow name, step number, the question that failed and a screenshot, and add the recording timestamp if session replay shows hesitation at that step. In our judgement, a "no" on question 2 usually points to a hidden or low-contrast control, and a "no" on question 3 to a vague label.
Rate each finding on the 0-4 severity scale, then assign it an owner in the backlog. Walkthroughs also suit products that are already live.
Before rewriting whym, a real-time interpreting app live on the App Store and Google Play, Netguru reviewed it across five test scenarios, using cognitive walkthroughs to find where the existing product lost people. The new version shipped three months after the project started (whym case study).
How do you match analytics funnels to screens?
Match an analytics funnel to screens by listing every funnel step as a named screen or state, then confirming that each one fires its own event. Do this for each of the key conversion flows before you read any drop-off number.
Unmapped steps are the most common failure. If the only events are "checkout started" and "payment submitted", a large drop between them could sit on the address form, the shipping options, a coupon field or a validation error. Analytics funnel analysis cannot separate these. Add one event per screen, including error states, so each drop-off points to a single screen.
When tracking is thin, work down this list and record which source each finding came from:
- Session recordings and heatmaps on the flow's screens, filtered to sessions that exit there, with the recording timestamp captured.
- Backend logs: failed form submissions, declined payments, API errors.
- Support tickets and chat transcripts, tagged by screen.
- A manual walk-through of the flow on real devices, with a screenshot at every step.
Log the missing events as a finding in their own right, with an owner and a severity rating.
Funnel data shows where people leave, not why. Treat each drop-off as a hypothesis and test it against what the heuristic evaluation or cognitive walkthrough found on that screen, for example a WCAG 3.3.1 error identification failure. If neither review explains the drop, that screen needs testing with participants.
UX audit checklist: Navigation, IA, forms and input
The navigation, information architecture and forms checklist below has nine rows, each answerable yes or no, and each names the method and the evidence a reviewer must capture. Run the heuristic evaluation rows with two or more reviewers and merge their findings afterward.
The form rows map to two of Nielsen's 10 usability heuristics: error prevention, and helping users recognize, diagnose and recover from errors. A missing 3.3.3 suggestion can turn a typo into an abandoned form.
Information architecture failures can look like content problems until a walkthrough shows where visitors get lost. On the Polpharma API website, visitors struggled to reach product data and the API catalog. A usability audit led to a new information architecture built around a product finder with search and filters, and after the redesign and a Webflow rebuild, form completions nearly doubled (Polpharma API case study).
UX audit checklist: Content, labels, microcopy, feedback and errors
A content, microcopy, feedback and error checklist maps each yes/no question to one of Nielsen's 10 usability heuristics, a method and the evidence to capture. Form error identification and recovery get a row each.
Rate every "no" on the 0-4 scale before merging reviewers' notes. Where event tracking logs field errors, check the two error rows against it to see which messages users hit most.
WCAG 2.2 AA checklist a reviewer can check by hand
A reviewer can check seven WCAG 2.2 AA success criteria by hand with a keyboard, a browser and a contrast checker, and each check ends in a yes/no answer plus a piece of evidence.
The hand check is fast and catches the failures users hit first, but it is not a conformance statement: WCAG 2.2 AA conformance covers every criterion on complete pages and processes, and some criteria need screen reader testing.
Timing matters as much as the checklist. For the zgłaszam.to Notify whistleblowing platform of the law firm Domański Zakrzewski Palinka, Netguru audited both the product and its website against WCAG while they were still under construction.
Each violation was described in an Excel checklist with screenshots and prioritized by its impact on users and the seriousness of non-compliance, which gave the firm a route to AA-level compliance (accessibility audit case study).
There is no certification for UX audits themselves. IAAP certifies individuals, and we recommend bringing an IAAP-certified accessibility specialist (the Web Accessibility Specialist credential covers technical evaluation) when you need a conformance claim or face Directive (EU) 2019/882 obligations. Deeper methods sit in our accessibility testing guide.
Performance as users perceive it, and key conversion flows checklist
Check perceived performance per key conversion flow, pairing each yes/no question with Core Web Vitals field data and a session recording timestamp.
Field data comes from the Core Web Vitals reports for the live URLs. Flows with too little traffic for field data fall back to throttled lab runs, and the report should say so.
Lab metrics show how fast a page loads. Recordings and heatmaps show whether users noticed the delay. Log every "no" against its flow name so it can be rated on the severity scale and assigned an owner. For mobile commerce apps, see how a UX audit finds conversion barriers in mobile apps.
How do you rate severity and prioritize findings?
Rate every finding on a 0-4 severity rating scale, then sort the 3s and 4s by engineering effort to produce the prioritized backlog. The scale follows Nielsen Norman Group's severity guidance, which weighs frequency, impact and persistence.
Report the ratings as a list per flow. A single UX score hides the one blocker behind ten cosmetic fixes.
Severity inflates when nobody owns the scale. Name one scale owner, usually the audit lead, who applies the definitions above. A 4 needs evidence of a blocked user in a key conversion flow, such as a recording timestamp or a funnel step. Without that, the finding is a 3 at most.
Effort comes second, from a different person. Hold a short feasibility review with an engineer and tag each 3 and 4 as small, medium or large. Order the backlog like this:
- Every 4, whatever the effort.
- Small-effort 3s.
- Large-effort 3s, each with a named owner and a release target.
- 2s and 1s, batched into a polish ticket.
What is included in a UX audit report?
A UX audit report holds one row per finding, and those rows become the prioritized backlog. Each row is filled in from the heuristic evaluation, cognitive walkthrough, funnel review or WCAG 2.2 check that surfaced it, and rated on the 0-4 severity scale.
The finding schema, and where each field lands in a ticket:
Write every row so it can be pasted into the tracker unedited and still make sense to someone who never read the report. If a developer needs the PDF to understand the ticket, the row is incomplete.
A report built this way can drive the next design phase directly. At Powermeals, a Swiss meal-prep and delivery app, the UX audit report shaped new designs across several parts of the app, including meal plan subscriptions, a refer-a-friend feature and meal filtering by allergies and dietary needs (Powermeals case study).
Component libraries need their own method, covered in our design system audit guide.
FAQ
How many reviewers does a UX audit need?
At least two for the heuristic evaluation. Each reviews the same flows alone, then the findings are merged and rated on one severity scale, which removes much of the personal bias. Nielsen Norman Group recommends several evaluators because each one finds a different share of the problems.
Can you run a UX audit on your own product?
Yes, with the checklists above, but familiarity is the main risk: people who built a flow pass steps a newcomer would stall on. Give the cognitive walkthrough to someone who has not worked on the flow. Bring in an outside reviewer when the team needs an independent view, for example before a funded redesign.
Turn the audit into work, then plan yours
An audit turns into work when every finding sits in a prioritized backlog with a flow, a severity rating, and an owner.
To plan yours, name the trigger first, pick the methods that fit it, and start with the key conversion flows: search, product page, checkout or signup, and onboarding. Run the audit before any redesign so the new screens do not inherit the old defects. If you want a second reviewer on those flows, you can improve your product's UX with Netguru.












