How to Run a Heuristic Evaluation of Your Website

Where the ten heuristics came from

The heuristics most people mean when they say "heuristic evaluation" were developed by Jakob Nielsen. In his own account of their history, Nielsen describes creating a first set in 1989 together with Rolf Molich, while working at the Technical University of Denmark, and refining the list in 1994 ¹. The 1994 revision was based on a factor analysis of 101 usability principles against 249 usability problems, an attempt to find the smallest set of principles that explained the largest share of real problems.

The result was a list of ten broad rules of thumb. They are called heuristics because they are not precise guidelines; they describe qualities of good interfaces that a reviewer can recognize and apply to almost any design, from a phone app to a checkout page.

The ten heuristics, applied to websites

The names below follow Nielsen's list as published by the Nielsen Norman Group ². The website examples are ours.

1. Visibility of system status

The site should always tell people what is happening. On a website, that means showing that a form is submitting, that an item was added to the cart, which step of checkout someone is on and which page of the menu they are viewing.

2. Match between system and the real world

Use words, concepts and order that visitors already understand. When labels are disputed, card sorting and tree testing show which words your audience expects. A menu item called "Solutions" may mean something to your team and nothing to a visitor looking for prices. Dates, currencies and units should follow US conventions for a US audience.

3. User control and freedom

People make mistakes and change their minds. Give them clear ways out: a visible close control on pop-ups, the ability to edit a cart, a back button that does not wipe a form and an easy way to undo an action.

4. Consistency and standards

Follow the conventions visitors know from other sites. Links should look like links, the logo should lead to the home page and the same action should have the same label everywhere. Internal consistency matters too: if one form marks required fields, all forms should.

5. Error prevention

Better than a good error message is a design that prevents the error. Use input types that bring up the right keyboard, constrain dates to valid ranges, confirm before destructive actions and show costs before the final step.

6. Recognition rather than recall

Do not make people remember information from one page to use on another. Keep filter choices visible, show the plan someone selected throughout checkout and label icons rather than relying on memory of what they mean.

7. Flexibility and efficiency of use

Let experienced visitors move faster without confusing new ones. Examples include search with useful suggestions, saved addresses for returning customers and a quick reorder option.

8. Aesthetic and minimalist design

Every extra element competes with the ones that matter. Remove content that does not help the visitor's current task, and keep promotions away from forms and checkout steps.

9. Help users recognize, diagnose, and recover from errors

Error messages should say what went wrong in plain language, point to the field involved and explain how to fix it. "Invalid input" fails this test; "Enter a five-digit ZIP code" passes.

10. Help and documentation

The best site needs no instructions, but some tasks are complex. Help should be easy to find, focused on the task at hand and written as concrete steps.

How to run the review

Define the scope

Choose the pages and tasks to review, the audience you are reviewing for and the devices to use. A review of everything produces a long, shallow list. Two or three core tasks reviewed carefully are more useful.

Use more than one reviewer

Different reviewers notice different problems. If you can, ask two or three people to review the same scope independently, without discussing their findings until each has finished. Then merge the lists and remove duplicates.

Walk the tasks, then sweep the pages

Start by completing each task as a new visitor would, noting every point of confusion, effort or doubt. Then go back through each page and check it against the ten heuristics one by one. The first pass finds problems in context; the second catches what the first missed.

Record each problem the same way

For every problem, record the page and step, a screenshot, a description of what happens, the heuristic it breaks, a severity rating and a suggested fix. A consistent format makes the list easy to merge, sort and import into a backlog. Our free audit checklist can help structure a first pass.

Rating severity

A list of problems without priorities is hard to act on. Rate each problem on a short scale that combines three factors: how many visitors it is likely to affect, how badly it disrupts the task and whether people can work around it. Our own audits use four levels, described on the usability audits page: critical, serious, moderate and minor.

When reviewers disagree on a rating, discuss the reasoning rather than averaging the numbers. The conversation may reveal that one reviewer imagined a different kind of visitor, which is useful information in its own right.

Common mistakes

• Citing a heuristic without a problem. "Violates consistency" is not a finding. Describe what the visitor experiences.
• Reviewing as yourself. You know the site. Review as a specific new visitor with a specific goal.
• Proposing redesigns. Suggest the smallest change that solves the problem. Large redesign ideas belong in a separate discussion.
• Treating findings as facts. Every finding is a prediction. Label them that way in the report.
• Ignoring accessibility. The ten heuristics do not cover every accessibility barrier. Add basic WCAG checks, as described in our accessibility testing guide.

What heuristic evaluation cannot do

A heuristic review is quick and inexpensive, which is why it is so widely used. Its weakness is that reviewers are not your visitors. They may flag problems that real visitors handle easily, and miss problems that come from your audience's particular knowledge, vocabulary or circumstances.

The practical answer is to combine methods. Use a heuristic review to clear away obvious problems, then test the most important findings with a small group of real visitors. Our guide to usability test sample sizes explains how small that group can be, and our guide to writing test tasks explains how to give participants tasks without leading them.