research methods
Expert review
What is an expert review, and how does it differ from user testing?
An expert review is a UX specialist evaluating a design against their knowledge and established principles to predict where users will struggle - no participants required. It's fast and cheap and catches a lot of obvious problems, but it's a prediction of usability, not a measurement of it: an expert flags what's likely to go wrong, where user testing shows what actually does.
Also known as: expert review, design review, expert evaluation, UX audit
The demo
A UX expert reviewed this sign-up screen and predicted where people would struggle. Flip to what real users actually did - and see how close the expert's guess was, and where it wasn't.
What this demo shows (text version)
A sign-up screen is evaluated two ways. The "expert's predictions" view lists problems a UX specialist flagged from principles alone - unclear password rules, a low-contrast button, an ambiguous label, a missing "why we need this". The "what users hit" view shows what real testing found. Most of the expert's calls were right, but a couple of their flagged issues never actually troubled anyone (false alarms), and users got stuck on something the expert sailed past - a confusing error message that only appears mid-flow.
That comparison is the whole point: an expert review predicts usability quickly and cheaply and catches the obvious problems, but it's an educated guess shaped by the expert's own blind spots, not a measurement. Run it first to clean up the glaring issues, then test with real users to catch the surprising things only they reveal. Predict, then measure.
An expert review trades real users for an expert's trained eye: quick, cheap, and good at catching obvious, principle-violating problems early, before you spend money testing. But it's an educated guess, prone to the expert's own blind spots and false alarms, and it can't surface the genuinely surprising things only real users do. Use it to clean up the glaring issues first; it complements user testing, never replaces it.
The expert flagged a stack of likely problems in an afternoon, no recruiting needed - and most were real. But put it beside what users actually hit and two things show up: a couple of the expert's "issues" never bothered anyone, and users tripped over something the expert walked straight past. That's the line between predicting usability and measuring it.
The appeal is speed and cost: an experienced reviewer can audit a design against heuristics and patterns in hours, with no recruiting, scheduling or incentives, and surface a large share of the obvious problems - the missing labels, the inconsistent buttons, the unclear flow. It's ideal as an early cleanup pass, so you don't burn precious user-testing sessions on issues an expert could have caught for free.
But it's a prediction, not evidence, and it has characteristic failure modes. Experts carry their own assumptions and expertise (they're nothing like a first-time user), so they raise false alarms on things that won't actually bother anyone, and - precisely because they "get" the design - miss the confusions only a real novice reveals. Two reviewers will also disagree.
So sequence them: expert review first to fix the glaring, cheap-to-find problems, then user testing to catch what prediction can't - the surprising, the context-specific, the things people do that no one anticipated. Structured variants (heuristic evaluation, cognitive walkthroughs) sharpen the method, but none of them substitute for watching real users. Predict, then measure.