First-Click Testing: A Worked Study, Results Table, and Test Script

Decide what the first click needs to tell you

A library has added online membership renewal, but the team disagrees about where the entry point belongs. One person expects readers to use their account; another prefers a prominent card on the home page. A first-click test can show where people would begin a renewal task on a particular design. It cannot show that they can complete the renewal, understand the eligibility rules, or receive a valid membership confirmation.

This worked example follows a public-library home page from test planning to a written decision. Its illustrative responses and counts produce a research plan, a transparent results table, and a revision decision whose strength matches the evidence.

The useful question is bounded: 'On the mobile home page, where would an existing member start if their membership were about to expire?' Avoid starting with 'Does the page work?' That question bundles together navigation, comprehension, account access, and the renewal transaction. If the actual concern is whether people can finish, plan a task-based usability study that continues beyond the first action.

Specify the page, routes, and scoring before collecting clicks

The fictional home page contains five relevant areas: Books and events, Library card, My account, Help, and a search control. In this design, Library card leads to membership information with a renewal link; My account offers a renewal action after sign-in; search can find the renewal page. Help contains opening hours and contact details but no renewal route. These are explicit properties of this example, not assumptions about how libraries usually organize their sites.

All three viable starting points belong in the accepted set. Calling only the designer's preferred card 'correct' would penalize working alternatives and overstate the navigation problem. Conversely, a link does not become a viable start merely because a participant selected it. Check where it actually leads before coding the response. If the design is still a static image, document the intended destinations and test those routes later in an interactive version.

Define what counts as the first action. Does opening a menu count? Is scrolling possible? Is the image a full page scaled to fit, or a realistic viewport? A screenshot that exposes a footer link without scrolling can produce a different result from the actual mobile page. Record the representation and its limits so the team does not compare unlike tests.

For this exercise, the test presents the initial 390-pixel mobile viewport with the navigation visible. It records the first selected area and allows an explicit 'I would not know where to start' response. That option avoids forcing an arbitrary click, although it remains a response to the test rather than a complete account of real-world abandonment.

Write a situation without naming the answer

The task prompt is part of the measurement. 'Click Library card to renew your membership' tests whether someone can follow an instruction, not whether the page communicates a starting point. 'Renew your card' is better, but can still repeat the exact label you are trying to evaluate. Describe the situation in the language a member might use and check that the intended goal remains unambiguous.

Use this prompt for the teaching study: 'You already borrow books from this library. Your membership will expire next month, and you want to keep borrowing after that date. Looking at this page, select where you would begin.' It gives a reason and a goal without identifying a navigation label. It also avoids assuming the person wants to replace a lost physical card or obtain a first membership.

After the selection, ask 'What would you expect to find after choosing that?' An optional second question is 'Was there another place you considered?' Ask these after the first choice if an unprompted initial action is the main outcome. Do not explain the correct route and then collect an explanation as though it had been spontaneous.

Pilot the instructions and test mechanism before recruiting the full sample. Check that participants understand they are choosing a starting point, that image loading does not consume the response timer, and that a tap is recorded where it occurred. If the pilot reveals an ambiguous task, revise it and keep pilot responses separate from results collected with the final wording.

Recruit for the question and keep comparisons interpretable

For a question about current members, recruit people who fit that use case. Familiarity with the existing library site may affect expectations, so record it and decide whether you need separate observations from frequent and infrequent users. A group of colleagues who designed the navigation cannot stand in for members discovering it. The recruitment criteria should explain whose decision the study represents.

The invented analysis below uses 24 responses to make the arithmetic easy to inspect. Twenty-four is not a recommended minimum or a promise of precision. Choose a sample based on the decision, the audience differences that matter, and the uncertainty you can tolerate. A small exploratory study can reveal a plausible problem without estimating its prevalence accurately. A comparison intended to support a numerical claim needs a sample and analysis planned for that purpose.

If comparing two layouts, keep the task, device conditions, recruitment criteria, and scoring rules aligned. Showing both versions to the same person can introduce learning: once someone discovers renewal under Library card, their next choice is no longer a fresh first encounter. Separate randomly assigned groups can avoid that particular carry-over, though they introduce sampling variation and still require an appropriate analysis.

Do not mix a static first-click test with a fully interactive task and call the difference a design effect. The materials and stopping conditions have changed. Also consider who the image-based format excludes. A flat screenshot cannot establish that navigation works with a screen reader or keyboard; test those interactions in an accessible, functioning interface with relevant participants.

Show the counts before explaining the pattern

Suppose the fictional study produces the distribution below. The denominator is all 24 included responses, including the explicit no-start response. There are 16 accepted starts: eight on Library card, five on My account, and three on Search. That is 16/24, or 66.7% after rounding. It is an accepted-first-choice rate under these rules, not a renewal completion rate.

Excluding the no-start response would change the result to 16/23, or 69.6%. Neither figure should appear without its denominator and rule. Here the no-start response is meaningful data, so the primary report retains all 24. A response from a broken image load would be a different issue: document the technical problem, apply a predetermined exclusion rule, and report the exclusion instead of hiding it inside the percentage.

A click map helps locate clusters, but the table connects those locations with their destinations and counts. Seven Help selections deserve investigation because Help offers no renewal route in this version. The three Search selections are not automatically failures: search was an accepted starting point, and its eventual success remains untested. A visually scattered map can therefore contain several reasonable choices.

If recording time, state when timing starts and stops and inspect technical delays. A long pause may reflect reading, distraction, uncertainty, or a loading problem. Time alone cannot identify which one occurred, and there is no universal number of seconds that certifies a navigation defect. Keep timing separate from the participant's explanation and the observed destination.

Use explanations to form a repair hypothesis

An illustrative response after a Help selection is: 'I expected Help to explain what happens when membership runs out.' It suggests a mismatch between the section's promise and its contents. It does not prove that changing the label alone will solve the problem; the placement, visual emphasis, and surrounding text may also contribute.

There are at least two reasonable design directions. The team could make the renewal route clearer in membership navigation, or it could add a useful route from Help because members reasonably expect assistance there. The first-click result identifies a starting-point mismatch. Choosing between those repairs also requires looking at the actual information architecture and the member's subsequent task.

Write the proposed change and its rationale separately. For example: 'Add a membership renewal link to Help and clarify the Library card summary. Seven of 24 teaching responses selected Help, which currently offers no renewal route.' Then state the uncertainty: 'The study did not isolate why Help was chosen or observe completion.' This leaves room for evidence to change the proposed solution.

The 29.2% describes the share selecting Help. It cannot predict a 29% increase in renewals from changing the navigation: the exercise observes starting choices, not completed renewals or the effect of a redesign. A full task study or a suitably designed production experiment is needed to assess those outcomes.

Finish with a decision and a complete-task follow-up

A useful report ends with a concrete action, not just a heatmap. For the teaching example, the decision is to revise the Help and membership entry points, then run an appropriate follow-up. Repeating a comparable first-click test can examine the entry-point question. Observing the full interactive journey can examine whether the route actually supports renewal, including sign-in, eligibility, and confirmation.

The follow-up task can retain the same situation while allowing people to continue. Define success before observing it: a member reaches the correct renewal outcome for the test account and understands the resulting membership status. A click on Renew is not enough if a later error prevents completion. Record assistance and alternate paths rather than counting every person who eventually reaches a page as unassisted success.

Keep the original test materials with the report: the exact page version, prompt, accepted areas, participant criteria, included and excluded responses, and analysis. When a teammate asks whether a redesign helped, these details make a fair comparison possible. They also prevent a future reader from turning an exploratory navigation result into a claim about all users or all tasks.

Use the memo below for your study. Replace the example counts and route definitions with your own evidence; keep the results fields blank until you have collected it.

References

• Lyssna: First click testing guide
• User Interviews: First-click testing field guide

Continue reading

• Card Sorting vs Tree Testing: A Worked Navigation Example