How to run a heuristic evaluation
A heuristic evaluation is a fast, low-cost way to catch usability problems before you spend a session on real users. Here is the full process, from picking evaluators to scoring severity.
A heuristic evaluation is a usability inspection method where three to five trained evaluators review an interface against a set of established usability principles, flag every place the design breaks one, and rate each issue by severity. It is one of the fastest and cheapest ways to catch usability problems, and you can complete one in a day or two without recruiting a single participant. The tradeoff is that it relies on expert judgment rather than real behavior, so it should clear out the obvious problems, not stand in for testing on real users.
This guide walks through what a heuristic evaluation is, the 10 heuristics most teams use, how many evaluators you actually need, how to score severity, and the step-by-step process. It also covers where heuristic evaluation stops being useful and real user testing has to take over.
What a heuristic evaluation is (and is not)
The method was introduced by Jakob Nielsen and Rolf Molich in the early 1990s. The idea is simple: instead of watching users struggle, you have people who understand usability principles inspect the interface and predict where users will struggle. Each evaluator works through the product against a checklist of heuristics, which are broad rules of thumb rather than rigid specifications.
Because it does not involve recruiting, scheduling, or moderating sessions, a heuristic evaluation is cheap and quick. A small team can turn one around in a couple of days. That speed is exactly why it is popular as an early cleanup pass.
What it is not is a substitute for observing real people. Evaluators predict problems; they do not measure them. They can tell you a form violates the error-prevention heuristic, but they cannot tell you that 40 percent of your actual buyers abandon that form because the error only appears after they hit submit. Heuristic evaluation finds issues fast. Testing on real, verified users tells you which issues actually cost you conversions, task completion, or trust. If you want the fuller landscape of methods and when each one fits, our complete guide to user research lays them out side by side.
Nielsen’s 10 usability heuristics
Nearly every heuristic evaluation uses the same core framework: the Nielsen Norman Group’s 10 usability heuristics. These have held up for three decades because they describe how humans interact with systems, not any specific technology.
| # | Heuristic | What it means |
|---|---|---|
| 1 | Visibility of system status | The system keeps users informed about what is happening through timely feedback. |
| 2 | Match between system and the real world | Language, concepts, and conventions match the user’s world, not internal jargon. |
| 3 | User control and freedom | Users can undo, redo, and exit unwanted states easily, without being trapped. |
| 4 | Consistency and standards | Words, actions, and layout follow platform conventions and stay consistent internally. |
| 5 | Error prevention | The design prevents problems before they happen rather than just showing error messages. |
| 6 | Recognition rather than recall | Options and information are visible so users do not have to remember things. |
| 7 | Flexibility and efficiency of use | Shortcuts and accelerators let experienced users work faster without hurting novices. |
| 8 | Aesthetic and minimalist design | Interfaces avoid irrelevant or rarely needed information that competes for attention. |
| 9 | Help users recognize and recover from errors | Error messages are in plain language, state the problem, and suggest a solution. |
| 10 | Help and documentation | When help is needed, it is easy to search, focused on the user’s task, and concrete. |
Some teams supplement these with domain-specific principles. Accessibility guidelines, mobile-specific rules, or a company’s own design system standards can be added as extra heuristics. The 10 above are the reliable baseline.
How many evaluators you need
This is the question that trips up most first-timers. A single evaluator is not enough. Research from Nielsen found that one person catches only about 35 percent of the usability problems in an interface. Different evaluators notice different things, so a lone reviewer leaves most issues on the table.
The evidence points to three to five evaluators as the sweet spot. At that range, the combined findings uncover roughly 75 percent of usability problems. Beyond five people, each additional evaluator finds fewer new issues while adding cost, so returns diminish quickly.
A few practical notes:
- Evaluators should work independently first. If they compare notes too early, they anchor on each other and miss issues. Combine findings only after everyone has done an individual pass.
- Double experts find more. Evaluators who understand both usability principles and the specific domain (say, enterprise security software or a clinical tool) catch more relevant problems than generalists. This is where recruiting the right domain expertise matters even for an internal method.
- You can use non-specialists carefully. If you lack trained evaluators, well-briefed team members using a clear heuristic checklist still add value, though expect lower coverage.
Severity ratings: prioritizing what you find
Finding 60 problems is useless if you cannot tell the catastrophic ones from the cosmetic ones. That is what severity ratings are for. The standard scale runs from 0 to 4.
| Rating | Label | Meaning |
|---|---|---|
| 0 | Not a problem | I do not agree this is a usability problem at all. |
| 1 | Cosmetic | Need not be fixed unless extra time is available. |
| 2 | Minor | Low priority, fix if there is time. |
| 3 | Major | Important to fix, high priority. |
| 4 | Catastrophe | Imperative to fix before the product can be released. |
Severity is a blend of three factors: frequency (how often the problem occurs), impact (how hard it is for users to overcome), and persistence (whether users hit it once or repeatedly). A rare, easily-overcome glitch scores low. A frequent, blocking, repeated problem scores high.
Have each evaluator rate severity independently, ideally after all problems have been collected, then average the scores. Averaging smooths out individual bias and gives you a defensible priority order to hand to design and engineering.
The step-by-step process
Here is the full workflow from setup to handoff.
1. Define scope and prepare materials
Decide what you are evaluating. A checkout flow, an onboarding sequence, a settings dashboard. Trying to evaluate an entire product at once produces shallow findings. Pick the flows that matter. Prepare a short brief describing the product, the target users, and the key tasks, plus the list of heuristics and the severity scale everyone will use.
2. Brief your evaluators
Give each evaluator the same context: who the users are, what they are trying to accomplish, and any domain rules. Grounding evaluators in the real user helps them evaluate against actual needs rather than their own preferences. If you have persona notes or prior research, share them. Our walkthrough on how to turn product research into better product decisions is a useful primer on keeping the whole team anchored to the user.
3. Run independent evaluation passes
Each evaluator goes through the interface at least twice. The first pass builds a feel for the flow and overall structure. The second pass focuses on specific elements against each heuristic. For every issue they find, they record the location, which heuristic it violates, and a short description. Two passes matter because you catch flow-level problems on the first and detail-level problems on the second.
4. Aggregate the findings
Collect every evaluator’s list into one master set. Merge duplicates where multiple people flagged the same issue, but keep a count of how many evaluators caught each one. Problems that many evaluators independently spotted tend to be the most severe and the most real.
5. Rate severity and prioritize
Send the combined list back to evaluators so each person can score severity on the 0 to 4 scale. Average the ratings. Sort the list so major and catastrophic issues rise to the top. This ranked list is your primary deliverable.
6. Report and recommend
A good heuristic evaluation report does not just list problems. For each significant issue it names the heuristic violated, describes the problem, states the severity, and suggests a concrete fix. Tie recommendations back to the underlying principle so the team understands why the change matters, not just what to change.
Where heuristic evaluation stops and testing begins
Heuristic evaluation has a well-known blind spot: false positives and false negatives. Evaluators sometimes flag things that never trouble real users, and they miss problems that only appear when a real person with real goals sits down with the product. Experts also tend to over-focus on interface details and under-weight whether the product solves the actual job the user came to do.
That is why the strongest teams treat heuristic evaluation as step one, not the whole staircase. Use it to strip out the obvious violations cheaply, then validate the cleaned-up design with real users doing real tasks. Watching a live participant hesitate, backtrack, or misread a label surfaces the behavioral truth no inspection can. Techniques like analyzing user interview data from raw conversations to actionable insights show how much richer the signal gets once real people are involved, and avoiding the 5 common user interview mistakes that ruin your research keeps that signal clean.
For a broader view of how inspection methods sit alongside moderated and unmoderated testing, our complete walkthrough to product research methods maps the full toolkit.
Pairing heuristic evaluation with real users on CleverX
The catch with the validation step is recruiting. If your product serves developers, compliance officers, procurement leads, or any hard-to-reach B2B audience, finding qualified people to test with is often slower than the evaluation itself. This is where CleverX fits.
CleverX is a B2B research platform with more than 8 million verified professionals, plus B2C reach, across 150+ countries. Every participant’s identity and professional background is verified, so when your heuristic evaluation flags a suspected problem in an enterprise workflow, you can put the fix in front of the exact role that lives in that workflow, usually within about two to five days. AI Interview Agents let you run moderated-style sessions at scale, and pricing is pay-as-you-go, so you only pay for the studies you run.
The pattern that works: run your heuristic evaluation to clear the obvious issues fast, then confirm the redesign with verified users who match your real audience. If you recruit B2B participants regularly, our guide to recruiting B2B research participants and the broader B2B user research playbook go deep on getting the right people fast. For consumer or mixed studies, the 8 methods that actually work for recruiting user research participants covers the alternatives.
Heuristic evaluation is a sharp, cheap first cut. The real users are where the answers live. When you are ready to validate what your evaluation surfaced, recruit verified participants on CleverX and close the loop.
Frequently asked questions
What is a heuristic evaluation?
A heuristic evaluation is a usability inspection method where a small group of evaluators reviews an interface against a set of recognized usability principles, most commonly Nielsen’s 10 usability heuristics. Each evaluator independently notes where the design violates a principle, then the team combines findings and rates each issue by severity. It is a fast, expert-driven way to surface obvious usability problems before you invest in testing with real users.
How many evaluators do you need for a heuristic evaluation?
Research from the Nielsen Norman Group suggests three to five evaluators. A single evaluator finds only about a third of usability problems, while three to five people working independently uncover roughly 75 percent of them. Adding evaluators beyond five gives diminishing returns for the added cost, so three to five is the practical sweet spot for most teams.
What is the difference between a heuristic evaluation and usability testing?
A heuristic evaluation uses trained evaluators to inspect a design against usability principles, with no real users involved. Usability testing observes actual users trying to complete real tasks. Heuristic evaluation is faster and cheaper and catches many surface-level issues, but it cannot tell you how real people actually behave, where they get confused, or which problems matter most in practice. The two methods complement each other.
What are Nielsen’s 10 usability heuristics?
They are visibility of system status, match between the system and the real world, user control and freedom, consistency and standards, error prevention, recognition rather than recall, flexibility and efficiency of use, aesthetic and minimalist design, help users recognize and recover from errors, and help and documentation. They were first published by Jakob Nielsen in 1994 and remain the most widely used framework for usability inspection.
How do you rate severity in a heuristic evaluation?
Severity combines how frequently a problem occurs, how much it impacts users, and how persistent it is. A common scale runs from 0 to 4: 0 means it is not a usability problem, 1 is cosmetic, 2 is minor, 3 is major, and 4 is a usability catastrophe that must be fixed before release. Evaluators score independently, then the team averages the ratings to prioritize fixes.
When should you run a heuristic evaluation?
Run one early, when you have a working prototype or a live product but before you commit to expensive usability testing. It works well as a cleanup pass to remove obvious problems so that later sessions with real users focus on deeper behavioral questions. Many teams run a heuristic evaluation, fix the clear issues, then validate the redesign with verified participants.