Usability testing in UX design in five steps

Usability testing is the method of watching real users attempt real tasks with a design, so that teams learn where people struggle before launch. Run in five repeating steps, plan, define tasks, recruit the right participants, observe, and report, it turns opinions about design into evidence that protects revenue and customer loyalty.

Consider the director of digital channels at a bank who approves a redesigned loan application after a convincing presentation. The screens are clean, the team is proud, and the launch goes ahead. Within a month the call center reports a surge of calls from applicants who could not find where to upload their income statement. Nothing was broken in a technical sense. The design simply asked people to think in the way the designers thought. A single afternoon of watching five customers try to complete the application would have shown the problem before it cost money.

That is the argument for usability testing. The brands that today give their users a great experience, as the original version of this article noted, went through a great deal of trial and error until they found solutions that worked, and that is why they test their designs iteratively with real users. This claim is the author's observation and not a measured statistic, but it matches what practitioners report. The companies that look effortless are usually the ones that watched people fail many times and fixed what they saw.

What is usability testing and why does it matter for the business?

Usability tests are a method that allows a team to test the ease of use of a design with real users. The most common form is simple. The team gives the user the design, assigns a series of tasks to complete, and observes while the person tries to perform them. The exercise is repeated with several users and at different stages of the design process, so that the team can understand where people experience difficulties and, with the feedback of the participants, make the pertinent adjustments. The Nielsen Norman Group describes the same essential elements in its introduction to the discipline, namely a facilitator, a participant and a set of realistic tasks, which shows how settled the method is.

The original article names three objectives for a test. The first is to determine whether the test subjects can complete the assigned tasks correctly and without inconvenience. The second is to measure how much the user enjoys the experience. The third is to identify the errors a design may have, so that they can be improved. Together they respond to a final goal, which is to improve the UX design, give the user an excellent experience and, as a consequence, build loyalty. For a business leader the translation is direct. Every obstacle that testing removes before launch is a lost sale, a support call or an abandoned process that never happens.

There is a second, less obvious reason to test. Teams that work on a product for months lose the ability to see it as a newcomer does. Words that seem obvious inside the company, such as the name of a menu or the label of a button, are unfamiliar to customers. Testing restores the outside view. It also changes internal conversations, because a video of a customer who cannot find a button settles an argument faster than any opinion, including that of the most senior person in the room. We discuss why this outside view matters in our article on interface versus experience and other myths, and in our introduction to UX research.

Why must usability tests be repeated throughout a project?

The author insists that usability tests must be iterative throughout the development of the project and understood as a process, with several stages to be followed each time a test is executed. The reasoning is that one large test at the end is the most expensive way to learn something, because by then changes are costly and politically difficult. Small tests repeated along the way catch problems when they are cheap to fix.

A frequently cited argument supports this approach. In a widely read article, the Nielsen Norman Group explains why you only need to test with five users in a single round. Its analysis found that about five participants are enough to uncover most of the usability problems of a design in one test, and that a team gets more value from running several small rounds, fixing the problems between them, than from a single large study. The figure often quoted from that research is that five users reveal roughly 85% of the problems, though the article itself frames it as an average under certain conditions and it should not be treated as a guarantee. The same group points out in a separate piece that a quantitative study, one that aims to measure and compare numbers, needs a much larger sample, so the rule of five applies to finding problems and not to proving statistics.

The stages that follow are the ones listed in the original article. Each is described here with a practical comment for the executive who must sponsor the work.

Step one, how do you plan a test so that nothing is left to improvisation?

Planning is of the utmost importance because, once the participants are in front of you, you cannot improvise. You must be clear about what you are going to test and how. The team therefore defines the specific aspects of the design it wants to examine, which may be the navigation, the names of the sections, or the hierarchy and relationship of the content. It also decides how the test will be run, whether in person or remotely, whether a moderator will guide it, and how much help the participant may receive.

These choices have consequences. A moderated test allows the facilitator to ask why a person hesitated, which produces rich understanding but requires scheduling and skill. An unmoderated remote test scales faster and costs less per participant, but it cannot follow up on an interesting reaction. The team at 18F, the digital service group of the United States government, published a clear introduction to remote moderated usability testing that shows how an organization can run sessions across distances without losing the conversation. For most corporate projects a combination works best, with moderated sessions in the early stages when the team has more questions than answers, and lighter remote checks as the design stabilizes.

A good plan fits on a single page. It states the question the test must answer, the part of the design under examination, the profile of the participants, the format, the date and the person responsible for the report. Writing the question first is the habit that most improves results, because it prevents a test from becoming a general tour of the product.

Step two, how do you define tasks that reveal real problems?

When the team knows what it wants to test, it defines the specific tasks the user must perform. These should be concrete actions that are relevant within the interface, so that the team can clearly measure whether the design allows the user to solve them simply. A weak task says explore the website. A strong task says you want to know the cost of a shipment to another city, find out how.

The team will often have many tasks to try, and it is best to prioritize them by importance. If the list is too long, the exercise becomes tiring, the participant becomes exhausted and stops performing the tasks effectively, and that affects the result. The usual advice is to choose between five and eight tasks that correspond to the most valuable journeys of the business, such as registering, buying, requesting a quote or recovering a password. Tasks should also be written in the language of the user and not in the vocabulary of the interface. If the task says click on the account settings tab, it gives away the answer and measures nothing.

Step three, who should be invited to test?

Before starting, the team must understand who its target audience is and find users who fit that profile, because the aim is to give those people the best experience. This is why the author recommends building your buyer persona first. Our article on how much you know about the people who buy from your business explains how to do that, and our piece on the buyer persona as the key to your business strategy goes deeper.

Recruiting the wrong participants is one of the commonest ways to waste a test. Colleagues who know the product, or friends who want to be kind, will not stumble where a real customer would. If the service is for suppliers, test with suppliers. If it is for older adults who are new to online banking, test with them, and do not assume that the findings from a young group will transfer. A practical approach is to write a short screening question for each criterion that matters, such as whether the person has bought online in the last six months or has used a similar service, and to offer a modest incentive to make participation worthwhile.

Step four, what should you watch for while users interact with the design?

Observe the user carefully while the tasks are performed. The behavior shows when the participant got stuck, how long a task took and whether the person looks comfortable, satisfied, stressed or frustrated. Give the participant the space and confidence to express what they think and feel, because that is where the most valuable observations come from. Silence from the facilitator is a skill, and the temptation to rescue a struggling user is the main enemy of a good test.

After the tasks, the participant can be asked what they felt or why they could not finish a task, in order to understand what may be failing. The author also recommends quantifying that information and taking notes during the process. Useful measures include the percentage of participants who completed each task, the time taken, the number of errors and a simple satisfaction rating. These numbers will not be statistically robust with five people, but they allow the team to compare one round with the next and to show leadership whether the design is improving. Observers from other areas, such as development, marketing and customer service, should attend when possible, since seeing a user struggle changes the mind of a skeptical engineer more than any report.

Step five, how do you turn observations into a report that leads to action?

With all the information collected, the team analyzes and synthesizes it to identify the design problems and decide what can be done to improve them. A good report is short and ordered by severity. It states each problem, the evidence for it, how many participants suffered it and a recommended change. Including a clip or a quotation from a session makes the finding memorable. The report should also record what worked, so that the team does not undo successful elements in the next round.

The author closes with a point that deserves emphasis. Tests are not carried out to know whether something works or not, but to understand whether the team is giving the user a pleasant experience, one that generates satisfaction and not frustration. The user's gestures and actions often tell more than words, and observing them carefully is how a team learns whether it offers an experience or only functionality. A person may say the page is fine and still sigh, squint or hover over the wrong element for ten seconds. Those signals are evidence, and they belong in the report.

What does this mean for the C-level?

For executives, usability testing is a risk-control tool that costs little compared with the investments it protects. A round of five sessions, prepared and analyzed, can be completed in days, and it can prevent a redesign that confuses customers or a launch that drives them to a competitor. The practice also builds a culture in which decisions about the customer experience are made with evidence. The questions a leader should ask are simple. When was our main journey last tested with real customers? What did we change as a result? Who owns the findings? If the answers are vague, the organization is relying on opinion.

It is also worth connecting testing to the commercial agenda. Journeys that are easy to complete convert better, cost less to support and encourage return visits, which is why our articles on the role of UX in improving customer loyalty and UX as a lever that drives the entire sales cycle treat usability as a business lever and not as a design preference.

How can a director start next Monday?

Pick the single journey that matters most to revenue, for example the purchase, the application or the registration, and write the question you want answered about it. Define five tasks in the customer's language, recruit five people who resemble your real customers, and schedule one-hour sessions within the next two weeks. Invite colleagues from product, development and service to observe. At the end of each session, ask the observers to write down the three things that surprised them. Within three days, produce a one-page report ordered by severity, assign an owner to each fix and schedule a second round to verify the changes. Repeat the cycle at each major stage of the project, and the practice will pay for itself.

How does ICX help?

ICX Consulting works with companies to plan, run and interpret usability studies as part of their UX and customer experience strategy, connecting findings to the business metrics that leaders care about. To start with a structured method, we recommend our Heuristic Tests Toolkit. We chose it for this topic because an expert review against established usability principles is the natural companion of user testing. It helps a team find the obvious problems cheaply before it spends the time of real participants on the subtler ones. You can also learn about our approach to user experience and explore how usability applies to web forms.

Frequently asked questions

How many users do I need for a usability test?

For finding problems, about five participants per round is a widely used guideline, repeated across several rounds as the design changes. If you need statistically reliable measurements, such as comparing completion rates between two designs, you will need a considerably larger sample.

Is usability testing the same as asking users whether they like a design?

No. Opinions are useful, but the core of a usability test is observing what people do when they attempt a task. What users say and what they do often differ, and behavior is the more reliable evidence.

Can usability tests be done remotely?

Yes. Remote sessions, moderated or unmoderated, are common and allow access to participants in different cities. Moderated remote sessions preserve the ability to ask follow-up questions, while unmoderated tests are faster and cheaper for simpler checks.

When in the project should we test?

As early and as often as possible. Sketches and prototypes can be tested before any code is written, which is the cheapest moment to change direction. Testing again after major changes confirms that the fixes worked.

Usability testing gives a team the humility to see its work through the eyes of the people it is meant to serve, and the discipline to improve it step by step. If you would like to know how your key journeys perform with real users, request a diagnostic with an ICX consultant and we will review it with you.