User Experience April 9, 2025 13 min read

Heuristic Evaluation Step by Step: The 10 Usability Heuristics Applied to One Example

Jakob Nielsen's ten usability heuristics help you spot common weaknesses in an interface without organizing a test first. This article walks you through the method step by step using an online booking flow.

A small bike repair shop has just launched a new website with online booking. The owner is happy, and so is the agency that built it. Yet people keep calling to make appointments, and some show up for appointments they never meant to book. Something about the booking flow isn’t working. But what exactly?

There’s no budget right now for a proper test with real customers. This is exactly the kind of situation heuristic evaluation was made for: you review an interface against proven rules of thumb and catch many typical weaknesses before you invite a single person.

In this article, you’ll learn the method through one running example. The “Sprocket” bike shop is made up, but its problems aren’t: you’ll find many of them on real websites. By the end, you’ll know how to prepare your own evaluation, how to apply each of the ten heuristics, and how to turn your findings into a list your team can actually work with.

What a heuristic evaluation is

A heuristic is a rule of thumb. It isn’t always right, but it usually is. In a heuristic evaluation, experts go through an interface and note every spot where it breaks one of these rules. No real users are involved. That makes the method fast and cheap, but also limited. More on that later.

The best-known rules of thumb come from Jakob Nielsen. He developed the method with Rolf Molich in 1990 and refined the ten heuristics in 1994, based on an analysis of 249 usability problems. In 2020, the Nielsen Norman Group refreshed the wording and added new examples, but the ten heuristics themselves haven’t changed since 1994. You’ll find a compact overview in our resource 10 Usability Heuristics.

If you’ve come across the term “expert evaluation”: heuristic evaluation is one type of it. To see how it differs from the cognitive walkthrough, another type in which you play through tasks from the user’s point of view, read How to Use Expert Evaluation Correctly. This article is all about practice: how do you actually apply the ten heuristics?

Preparation: define the scope and pick a task

The most common beginner mistake is trying to review “the whole website.” Two hours later, you have fifty notes about the homepage, the legal notice and the footer, and not a single insight into the flow that really matters. That’s why the Nielsen Norman Group recommends narrowing the scope: to one task, one section, one user group or one device.

For the Sprocket bike shop, that means reviewing only the path from the homepage to a confirmed repair appointment, on a smartphone, from the perspective of someone booking for the first time. That’s small enough to be thorough and big enough to explain all those phone calls.

Phrase your scope as a task, the way a real person would approach it. For example: “Book a tune-up for your bike next week, on a day when you’re free in the morning.” With a concrete task in mind, you don’t just judge whether something looks nice, but whether it gets you to your goal.

Also decide up front how you’ll document your findings. A simple spreadsheet is enough. There’s an example further down.

Alone or as a team?

Ideally, three to five people review the same interface, each working independently. The reason is simple: everyone spots different problems. Across six projects, Nielsen found that individual evaluators discovered only about 35 percent of the usability problems on average. So when you review alone, you’ll likely miss more problems than you find.

That doesn’t make a solo evaluation pointless. It just makes it incomplete. If you don’t have colleagues to help, ask someone on the team who knows the product well but didn’t help build it. Or review it twice yourself, a few days apart.

Running the evaluation: two passes

Plan for about one to two hours per person. That’s the timebox the Nielsen Norman Group recommends for the independent review. Split it into two passes:

First pass: get to know it. Complete the task once, the normal way. Only note where you hesitate or feel surprised. The goal is to get a feel for the flow.
Second pass: evaluate. Go through every step again, this time with the list of ten heuristics next to you. For each screen, ask yourself which rule it breaks.

One important point: if you’re working as a group, don’t look at each other’s notes until everyone is done. Once you know what others found, you unconsciously look for confirmation and miss things of your own.

The 10 heuristics applied to the booking flow

Now for the hands-on part. For each heuristic, you’ll find a short explanation, the finding from Sprocket’s booking flow and a possible fix.

1. Visibility of system status

The interface should always keep people informed about what’s going on, and do so promptly. When you submit something, you want to know it arrived.

Before: After tapping “Request appointment,” a spinner turns with no text. A few seconds later, you’re back on the homepage. There’s no confirmation, not even an email. Many people pick up the phone just to be sure.

After: A confirmation page shows the date, time and service and says an email is on its way. Above the form, a progress indicator reads “Step 2 of 4,” so people know how far they still have to go.

2. Match between the system and the real world

The interface should speak the language of the people using it. Internal terms, abbreviations and jargon don’t belong there.

Before: The services are called “Service S1,” “Service S2” and “Service S3.” That’s how they’re listed in the shop’s point-of-sale system. First-time customers have no idea what’s behind them.

After: “Basic tune-up: adjust brakes and gears,” “Full service: check everything, replace worn parts,” “Repair: I know what’s broken.” Each one with a price and an approximate duration.

3. User control and freedom

People tap things by accident or change their minds. They need a clearly marked emergency exit to cancel or undo an action.

Before: The booking flow has no back button. Using the browser’s back button wipes out everything you’ve entered. Once a repair is booked, there’s no way to change or cancel it online.

After: Every step has a “Back” link, and entries are kept. The confirmation email includes a link to reschedule or cancel.

4. Consistency and standards

The same things should look the same and have the same name. And a website should follow the conventions people know from other websites.

Before: The main button says “Next” in the first step, “Continue” in the second and “Proceed” in the third. Sometimes it’s a green button, sometimes a gray text link. The date shows up as “11/03” in one place and “2026-11-03” in another.

After: One consistent “Next” button in the same style on every step, and “Confirm booking” at the end. One date format that matches local conventions.

5. Error prevention

Good error messages matter. But a design that keeps errors from happening in the first place is even better.

Before: The calendar lets you pick Sundays and past dates, even though the shop is closed then. The booking goes through anyway and gets canceled by phone later.

After: Closed days and fully booked slots are grayed out and can’t be selected. Before submitting, a summary shows all details so typos stand out.

6. Recognition rather than recall

People shouldn’t have to remember what they saw one step earlier. Everything they need should be visible or easy to retrieve. That takes load off working memory.

Before: In the last step, a free-text field labeled “Your request” asks you to describe once more which service you chose. What you selected in step 1 is nowhere to be seen.

After: A small summary above the form shows the service, date and time throughout the flow. The free-text field now reads “Anything else you’d like us to know? (optional)”.

7. Flexibility and efficiency of use

Experienced users should be able to get things done faster without making it harder for newcomers. Shortcuts help one group and don’t get in the way of the other.

Before: Regulars who bring their bike in for a tune-up every spring have to type in their name, phone number and bike model every single time.

After: The confirmation email includes a link, “Book your next appointment with these details,” that pre-fills the form. No customer account required.

8. Aesthetic and minimalist design

Everything on a screen competes for attention. Content that isn’t needed for the task at hand distracts from what matters. Minimalist doesn’t mean “as empty as possible,” it means “nothing unnecessary.”

Before: Between booking steps, there’s a newsletter banner, an Instagram feed and a promo for gift vouchers. On a smartphone, you have to scroll to find the Next button.

After: During booking, the page only shows what’s needed to book. The newsletter sign-up moves to the confirmation page, where it no longer gets in anyone’s way.

9. Help users recognize, diagnose and recover from errors

If an error does happen, the message should explain in plain language what went wrong and how to fix it. Error codes don’t help anyone.

Before: At the top of the page, red text reads “Error 422: Invalid input.” It doesn’t say which field is the problem.

After: Right below the phone field, it says: “Please enter your phone number including the area code.” All other entries are kept. Short bits of text like this are called microcopy, and they often make a big difference.

10. Help and documentation

Ideally, an interface needs no explanation. When it does, help should be easy to find, focused on the task and brief.

Before: What to bring to the appointment and how long a repair takes is only covered in an FAQ PDF linked at the very bottom of the footer.

After: Two short hints right in the flow: the approximate duration next to each service, and on the confirmation page, “Please drop off your bike by 9 a.m. Don’t forget the key to your lock.”

Not every violation is a problem

Once you’ve just learned the heuristics, you’ll suddenly see violations everywhere. That’s normal. Still, it’s worth pausing on each finding. The Nielsen Norman Group stresses that heuristics are guidelines, not strict rules. Whether a violation is a real problem depends on context.

An example from the bike shop: the booking flow doesn’t offer customer accounts. You could log that under heuristic 7 as a missing shortcut. But the shop chose not to offer accounts on purpose, because most people only book once or twice a year and don’t want to manage yet another password. That’s a trade-off, not an oversight. Write it down anyway, but note that it may be a deliberate decision.

The reverse also applies: if you can’t map a finding to any heuristic, write it down anyway. The ten rules are a tool for looking more closely, not a filter that throws everything else out.

Documenting and rating your findings

A heuristic evaluation is only as good as its list of findings. A note like “form is confusing” doesn’t help anyone. Describe each finding so that someone who wasn’t there can follow it. A spreadsheet with these columns works well:

Location: Where exactly does the problem occur? (step, screen, element)
Heuristic: Which rule does it break?
Description: What happens, and why is that a problem for people?
Severity: How urgent is it? (see below)
Suggestion: A first idea for how to fix it.

An entry for the bike shop could look like this: “Location: step 4, after submitting. Heuristic: 1, visibility of system status. Description: After submitting, no confirmation appears and you land on the homepage. People don’t know whether the booking went through, so they call or book twice. Severity: 4. Suggestion: confirmation page with a summary plus a confirmation email.” Adding a screenshot makes the finding even clearer.

Nielsen’s severity scale

So your team knows what to tackle first, every finding gets a severity rating. Nielsen suggests a scale from 0 to 4:

0: not a usability problem
1: cosmetic problem only, fix it if there’s time left
2: minor problem, low priority
3: major problem, important to fix, high priority
4: usability catastrophe, must be fixed before launch

Three questions help you rate a finding: How often does the problem occur? How much does it get in people’s way, can they still get through? And do they stumble over it every time, or only the first time? The missing confirmation affects everyone who books, there’s no way around it, and it happens with every booking. That’s a 4. The inconsistent button labels are irritating but barely slow anyone down. More like a 1 or 2.

Nielsen recommends assigning severity ratings only after the actual review, because while you’re searching, your focus is on finding problems. If there are several of you, each person rates the combined list independently, and you take the average afterward. He considers a single person’s ratings unreliable, while the mean of about three ratings is good enough for most purposes.

Consolidating and moving forward

Once everyone is done, pool your findings. Merge similar entries into one, for example on a digital whiteboard by grouping sticky notes by topic. Then discuss where you agree, where you don’t, and which three to five problems you’ll tackle first.

For the Sprocket bike shop, a clear order emerges: confirmation after submitting (heuristic 1), no closed days in the calendar (heuristic 5), a clear error message at the field (heuristic 9) and understandable service names (heuristic 2). These four changes probably account for a good share of the phone calls. But that isn’t certain.

What the method can’t do

And this is the most important limitation: a heuristic evaluation shows where an interface breaks proven rules. It doesn’t show how real people deal with it in their everyday lives. You are not the target audience, no matter how well you put yourself in their shoes. That’s why the Nielsen Norman Group explicitly states that the method doesn’t replace user research and works best as a complement to testing.

A sensible next step for the bike shop would be: fix the worst problems from the evaluation, then run a short usability test with five people. That will show whether the changes work and which problems the evaluation missed. Maybe it turns out that many people don’t know whether their bike needs a basic tune-up or a full service. Nobody would have spotted that with heuristics alone.

Conclusion

Heuristic evaluation is a great way to get started with usability work, because all you need is the ten heuristics, a clear task and one to two hours. Narrow the scope, go through the flow twice, describe each finding so others can understand it, and rate how urgent it is.

Keep in mind what the method is: a structured outside view, not a substitute for real people. Used that way, it helps you catch the obvious pitfalls quickly so your tests can focus on the questions that are genuinely open.

Checklist

Scope limited to one flow, one device and one user group
Concrete task written the way a real person would approach it
List of the ten heuristics at hand
Ideally three to five evaluators working independently
Two passes: first get to know it, then evaluate
Each finding noted with location, heuristic, description and suggestion
Severity from 0 to 4 assigned only after the review
Findings consolidated and the most important ones tackled first
Results checked with a usability test

Sources

Back to “All articles”