UX audit checklist: what to review, how to rank findings and what an audit costs

The short answer

A UX audit checks whether people can complete the tasks that matter on your site or product, and explains why they fail. A complete one combines seven checks: a heuristic review against Nielsen Norman Group's 10 heuristics by three to five people working separately, funnel analysis in your analytics, session recordings and heatmaps, usability tests with about five people per round, and checks of accessibility, content and speed. Each finding gets a severity, a reach and a business value, which sorts the list into what to fix now.

Published prices we read on 17 October 2026 run from a free automated report to USD 500–1,500 for a basic audit on Upwork and from USD 60,000 for a Nielsen Norman Group design review. A small team can run the do-it-yourself version below for about USD 500 in tools and recruiting, plus participant incentives and its own time.

Scope it before you check anything

An audit that reviews "the website" finds a long list of small problems and misses the one that costs money. Scope it around tasks instead.

1. Name the user groups: for example, first-time visitors, trial users and admins. Groups that behave differently are tested separately.
2. List three to six tasks per group in the user's words: "find out what it costs", "book a demo", "invite a teammate", "export last month's report".
3. Gather what you already know: analytics, support tickets, the questions sales hears, survey answers and past research.
4. Fix devices and browsers: if most traffic is on phones, the review and the tests happen on phones.
5. Pick the evaluators: three to five people, ideally including someone who didn't design the screens.

The checklist below follows that scope. Each table can be copied into a spreadsheet, one row per check per task.

The UX audit checklist, area by area

1. Heuristic review

Nielsen Norman Group's 10 usability heuristics, first published in 1994 and last reviewed in January 2024, are "broad rules of thumb", not a checklist of features. Walk through each task and ask the questions in the middle column.

Run it the way NN/g describes in its guide to conducting a heuristic evaluation: ideally three to five evaluators, each reviewing independently before anyone compares notes, because each person misses problems the others catch. NN/g is also clear that a heuristic evaluation can't replace user research, so treat findings that rest on an assumption as questions for the tests.

2. Task and funnel analysis

Analytics tells you where people drop out and how many; it doesn't tell you why. Build one funnel per task, for example with GA4's funnel exploration, which shows how many users complete or abandon each step.

If tracking is missing or wrong, fix it before you draw conclusions; our comparison of GA4 and PostHog covers which tool records what.

3. Session recordings and heatmaps

Recordings show behavior at the steps the funnel flagged. Watch sessions that reached the worst step, not a random sample, and note what people did just before they left.

Check privacy settings before you record. PostHog and Microsoft Clarity both mask content you mark as sensitive, and a recording tool may need the visitor's consent where your privacy law requires it.

4. Usability testing with five users

Nothing else in the audit shows you why people fail. Nielsen's model estimates that a test with five users finds about 85% of a design's usability problems, and he argues that three rounds of five, fixing between rounds, improve a design more than one round of 15 (NN/g). With two distinct user groups, NN/g suggests three or four people from each; with three or more groups, three from each.

Five-user tests find problems; they don't measure how often they happen. For a success rate or a time-on-task you can track, NN/g's summary of quantitative sample sizes puts the appropriate number for most studies at 40 participants.

5. Accessibility, where it overlaps with UX

A UX audit isn't an accessibility audit, but the audited tasks should pass the WCAG checks that affect everyone. WCAG 2.2, a W3C Recommendation since 5 October 2023, added nine success criteria, several of them on this list.

Automated scanners help, but the W3C says evaluation tools can't determine accessibility, only assist, so test the tasks with a keyboard and a screen reader. Our free WCAG 2.2 checklist lists every Level A and AA criterion; for full conformance, see what an accessibility audit costs.

6. Content and copy

Most failed tasks on marketing sites are content failures: the visitor can't tell what you offer, what it costs or what happens next.

7. Performance

Slow pages fail tasks before anyone reads them. Check field data for the templates the tasks use against Google's Core Web Vitals thresholds, at the 75th percentile of visits:

• Largest Contentful Paint within 2.5 seconds
• Interaction to Next Paint of 200 milliseconds or less
• Cumulative Layout Shift of 0.1 or less

If a template fails, our playbook on improving Core Web Vitals covers the fixes, and our platform speed report shows how sites on each platform score with real users.

How to prioritize the findings

A raw findings list is a to-do list nobody finishes. Score each finding on three things, and keep them separate so you don't count the same thing twice.

Severity uses NN/g's 0 to 4 scale from severity ratings for usability problems:

• 0: not a usability problem
• 1: cosmetic, fix only if time allows
• 2: minor, low priority
• 3: major, high priority
• 4: catastrophe, fix before release

NN/g builds severity from a problem's frequency, its impact and its persistence, plus its market impact. Because analytics measures frequency directly, rate severity on impact and persistence, and take frequency from the data.

Reach is the share of sessions or accounts that get to the step, from your funnel. Value is what the task is worth to the business: an illustrative scale is 3 for a revenue task (checkout, demo request), 2 for activation or retention, 1 for everything else.

Priority = severity × reach × value. Here's how four hypothetical findings sort:

Then add an effort estimate and sort high-score, low-effort items to the top. A severity-4 problem in a release-critical flow gets fixed regardless of its score.

What a UX audit should deliver

• The scope: user groups, tasks, devices and methods, agreed before the work started
• A findings log: one row per finding, with where it happens, the evidence (clip timestamp, funnel numbers or the heuristic it breaks), severity, reach, value and a recommended change
• Annotated screens showing each problem and the proposed fix
• Highlight clips from the usability tests, a few minutes each, for the people who weren't there
• A prioritized plan: fix now, next and later, with effort estimates
• A readout with the team that will make the changes
• The raw material: recordings, notes and analytics exports, so the next audit can compare

Push back on a report that lists generic best practices without evidence from your own users, or scores screens without saying which tasks were tested.

What UX audits cost: published prices

Few providers publish a price for a UX audit; most quote after a call. These are the published prices we could read on 17 October 2026.

The spread reflects scope more than skill. An automated report reads one page; a marketplace audit is usually one person's heuristic review; NN/g's price buys senior researchers, several products or competitors, and a full report. None of these prices includes usability testing with recruited participants unless the provider says so.

What moves a quote: the number of tasks and user groups, whether there are test rounds and how many, recruiting and incentives, whether the product sits behind a login with roles and test data, and the languages. Dardo's UX audit is quoted per scope, without a published starting price.

A do-it-yourself UX audit for a small team

The cash cost is small; most of the effort goes into recruiting, sessions and analysis. These are the tools and recruiting services, at prices published on 17 October 2026.

A worked budget for one round of five moderated consumer sessions: Clarity at no cost, five sessions recruited through Respondent at 5 × USD 40 = USD 200, and Lookback's USD 299 plan, which also covers a second round. That's USD 499 before incentives, which you set yourself. A second round of five adds USD 200 in recruiting, for USD 699 in total. If your own customers will join a call, recruiting costs nothing but your time.

The steps:

1. Agree the user groups and three to six tasks, using the scoping list above.
2. Build one funnel per task and note the worst step for each.
3. Three people run the heuristic review separately, then merge their findings; NN/g's article on heuristic evaluation links a fillable workbook.
4. Watch recordings of sessions that reached each worst step.
5. Run five usability sessions, fix what's clearly broken, then run five more.
6. Check the tasks against the accessibility, content and performance tables.
7. Log every finding with its evidence, score it and agree the top five fixes with whoever will build them.
8. Re-test the fixed tasks, and compare the funnel a month later.

The weak point of a DIY audit is distance: the people who designed the screens are the least able to see their problems. Bring in at least one evaluator from outside the team, even a colleague from sales or support.

When to bring someone in

Hire help when nobody on the team is independent of the design, when the product sits behind a login with several roles, when the audience is hard to recruit, such as clinicians or procurement managers, or when you need the findings to settle an internal argument. If you already know what's broken and have the traffic to test changes, conversion rate optimization is the better spend; if the findings point to a missing flow rather than a fix, that's product design work.

To run it yourself, start with the scoping list and the seven tables above. If you'd rather have Dardo run it, our UX audit page explains how we scope and test, and a website redesign checklist helps when the audit points to a bigger change.