How do I run usability tests on a small budget?

Five participants, three or four realistic shopping tasks, and an hour each are enough for a small qualitative usability study, provided you fix what you find and test again. Recruit customers and people who have not bought yet. Explain the study and ask permission before recording.

The benchmark

The five-participant guideline comes from a Jakob Nielsen article published in 2000. Its model assumes that one participant encounters, on average, 31% of an interface's usability problems. Under that assumption, five participants uncover roughly 85%. Rather than run one large study, Nielsen recommends three rounds of five, with fixes between rounds.

These are different kinds of evidence. Raluca Budiu at Nielsen Norman Group explains that a qualitative study produces a list of problems, not a statistically precise estimate of their prevalence. The 31% assumption comes from projects in the early 1990s. Plan for no-shows too: a separate NN/g recruiting survey reported an average of 11% in 2003, roughly one person in nine. Neither figure is a forecast for your store.

The rule

Test a specific journey, rather than “the website.” Start where your analytics show a loss without explaining it: a product page that loses visitors, or a cart that is frequently abandoned. Five sessions should leave you with a short, actionable list of changes.

Recent customers are easy to recruit, but they have already overcome the obstacles you want to observe. One practical starting mix is three customers and two newsletter subscribers who have never ordered. Adjust that mix to the audience you are studying.

Keep research participation separate from marketing permission. Tell people what the study involves, what data you will collect, whether you plan to record, who can access the recordings, and when you will delete them. Ask before recording and offer an unrecorded session where practical.

The applicable privacy rules depend on your business and participants. EU GDPR can cover businesses outside the EU when they target or monitor people there; it is not a French-only rule. Use the European Commission's applicability guidance to identify the relevant scope, then check the requirements for your research setup. A French CNIL example is not a worldwide recording-permission standard.

The steps

1. Write the tasks. Use three or four real shopping situations: “You are looking for a gift for your sister, around the middle of this store's price range.” Avoid instructions that reveal the interface answer, such as “Click New Arrivals.” The wording is a research task, not a sales script.
2. Recruit. Send a short invitation to recent customers and subscribers who have not ordered. Explain the incentive, whether it is store credit or a product. Invite six people if you need five completed sessions.
3. Explain participation and recording. Before the session, provide a plain-language notice covering the purpose, data collected, recording arrangements, retention period, and contact for questions or applicable rights. Request recording permission separately. Confirm the appropriate privacy requirements before your first study.
4. Run the session. Allow five minutes for introductions, time for the tasks, and ten minutes for questions. A video call with screen sharing is sufficient. Ask the participant to think aloud, then let them explore without helping. Three closing questions from a Club brand's research guide are: “At what point did you decide whether to add this product to your cart?”, “What would make you hesitate now?”, and “How would you explain this product to someone else?”
5. Take notes in pairs. One person moderates; the other records the task, outcome, point of difficulty, exact words, and severity. Distinguish success without help, success with help, and failure. Classify issues by whether they prevent a purchase, slow it down, or merely annoy. Compare notes immediately after each session.
6. Fix and retest. Rank problems by frequency and severity, fix the first three, and run another round with five new participants. If your traffic is too low for an A/B test, this cycle can become a principal method of improving the store; see the Playbook on optimizing without an A/B test.

The pitfall

Testing with colleagues or friends is tempting. They know the brand, infer what the interface is meant to do, and may avoid saying it is confusing. The study then confirms what the team already believes.

The other mistake is turning five observations into a population estimate. “Three of five participants could not find the size guide” identifies a serious problem. “60% of our customers cannot find it” claims a precision the study does not provide.

In a mobile product-page audit described by the Club's contributors, a shopper hesitated between two sizes, checked the return policy, came back, and left. Analytics recorded an exit. A session recording showed where uncertainty appeared. A moderated study can help explain what the person was trying to find. Combine that observation with session-recording analysis and an assessment of the privacy rules that apply to your setup.

What to tell your team

Budget for both the first study and the follow-up. This fictional budget uses USD for illustration; choose incentives suitable for your participants and market:

“Five one-hour sessions, a $40 store-credit incentive per participant, $200 in incentives and two days of work. We will fix the three most frequent problems, then run five more sessions next month to check that the fixes worked.”

The figures are fictional. Replace them with your recruiting costs and participants' circumstances.

Sources

• Nielsen Norman Group: Why You Only Need to Test with 5 Users (2000)
• Nielsen Norman Group: Why 5 Participants Are Okay in a Qualitative Study, but Not in a Quantitative One (2021)
• Nielsen Norman Group: Recruiting Test Participants for Usability Studies (2003)
• Nielsen Norman Group: Thinking Aloud: The #1 Usability Tool (2012)
• European Commission: Application of the GDPR