What a UX audit should leave you with
A useful audit does not end with a scorecard and a feeling. It ends with a ranked list of evidenced problems that your team can start fixing on Monday.
Most UX audits are bought once, skimmed once, and filed. That is rarely because the site had no problems. It is because the audit was written to demonstrate expertise rather than to transfer a work list. Knowing the difference before you commission one saves the fee and, more importantly, the quarter.
The test is simple: could your team open the report on Monday and start fixing things in a sensible order, without the author in the room? Everything a good audit contains exists to make the answer yes.
Evidence, not taste
A finding is a claim about your site, and claims need evidence. “The hierarchy feels cluttered” is taste; you can neither verify it nor act on it. “The homepage's primary button links to the login screen, so a first-time visitor hits a dead end” is a fact anyone can check in ten seconds, and once checked, nobody argues about whether to fix it.
Insist on findings tied to specific pages and elements, reproduced from the live site rather than screenshots of an old deploy. An honest audit also separates what was confirmed from what is inferred. Colour-contrast ratios, keyboard behaviour and screen-reader output need checking in a real browser; a report that asserts them from a glance is padding its severity counts.
A framework you can interrogate
Heuristic review sounds academic, but the point of scoring against a published framework (Nielsen's ten heuristics, WCAG 2.1 AA) is accountability. A framework forces the auditor to look at the unglamorous categories: error recovery, system status, consistency. It also means two people can disagree about a finding by pointing at a shared definition instead of trading adjectives.
The framework is the checklist, not the deliverable. If the report spends more pages explaining the methodology than listing what is wrong with your site, the ratio is backwards.
The conversion path, mapped
The highest-value section of any audit is usually a table nobody thinks to ask for: every call-to-action on the site and where it actually goes. Sites accumulate wiring errors the way houses accumulate extension cords. Two buttons with the same label pointing at different pages. A hero that sends new visitors to a login form. A “contact us” anchor that scrolls nowhere because an ID changed case.
These are not design problems and no redesign will find them. They are QA failures, they are invisible in a design file, and they sit directly on the revenue path. In one recent audit, the single highest-impact finding was that the most prominent button on the most visited page pointed at the wrong destination: a few hours to fix, and quietly costing conversions since launch.
Ranked by impact times effort
Twenty findings of equal visual weight is a wall, and walls get filed. The author's judgement about which problems are costing you most, and what each costs to fix, is half of what you are paying for. Expect every finding to carry:
- A severity: what it costs in conversions, trust, or excluded users.
- An effort estimate a developer would recognise: hours, days, week-plus.
- A recommended fix, specific enough to become a ticket without translation.
- A stated confidence: confirmed on the live site, or needs verification.
The ordering matters more than the count. The best audits usually conclude that three or four fixes recover most of the damage, and say so plainly instead of presenting all twenty as equally urgent to justify the fee.
A summary someone else can forward
The full report is for the team doing the work. The people approving the work need one page: what is wrong, what it costs, what to fix first. If the audit cannot be summarised on a page, it has not finished its own prioritisation. The forwardable summary is also the practical test of independence: it should stand on its own whether or not the auditor is hired to fix anything.
That last point is worth making explicit before you commission. An audit written as a pitch will inflate problems the author wants to be paid to fix. Ask directly: is the report actionable by our own team, or by any team we hire? If the answer is hedged, keep looking.
The Monday test
Evidence per finding, a shared framework, the conversion path mapped, severity times effort, a forwardable summary. Together they produce the only outcome that matters: on Monday, your team knows what to fix first and why.
This is the shape of our own fixed-fee UX audit: one week, every finding evidenced from your live site, ranked so the order of work is obvious. If you suspect your site is leaking somewhere but can't see where, send us the URL.