CESAWS journal

Usability Testing Checklist for Websites and Apps

WhatsApp
Usability Testing Checklist for Websites and Apps - CESAWS

Test five practical tasks before launch, including navigation, form recovery, accessibility and staff handoff.

Consider a team reviewing a new booking workflow before launch. The immediate goal is evidence that representative users can complete the task. Use the worked scenario with a researcher, a designer and a support coordinator. The examples describe how to plan the work, assign decisions and check the result before committing to a wider rollout.

Define the test outcome

Begin with the participant task prompts. The research owner can bring it to the review and identify what the team already knows. In this example, instructions name the exact button to click. A workable first response is to describe the goal without telling the participant how to get there. Confirm that the proposed response fits the operating conditions before budgeting the remaining work.

Collect evidence from an actual transaction before holding the scoping meeting. Use this method when the team needs to define the test outcome. A blank form makes it easy for everyone to agree that the fields look reasonable. A completed record reveals who supplies each value, how reviewers recognise a problem and what people currently do when information is missing. Bring a record that represents everyday work and one that was uncomfortable to handle. Remove any personal information unnecessary for the discussion. Record what makes the two cases different before changing the workflow.

A useful inventory distinguishes the document the customer sees, the working record staff edit and the evidence managers use after the event. The participant task prompts should show how this decision contributes to evidence that representative users can complete the task. A label can be identical across all three and still represent different things. Ask the person who performs the task to explain the last decision they made and where they looked for reassurance. If they had to open another tool or consult a colleague, capture that dependency. That is a genuine part of the work and can affect the estimate even if the screen looks simple.

Time-box the investigation. Confirm the outcome with the research owner before signing off on this part of the work. Start with the outcomes the project must support, then add only details that change a decision, permission or next action. Write down one uncertainty and name the person who can resolve it. Leave unverified claims visibly open. When the team later changes the design, keep the original evidence next to the proposed change so reviewers can tell whether the problem under discussion has actually been addressed. That comparison becomes particularly valuable when several people remember the old process differently.

Before continuing, ask the research owner to identify a representative item in the participant task prompts. Use it to show how the response would work and which inputs remain uncertain.

Recruit relevant users

Make the participant coverage sheet specific enough to support a decision. Ask the project coordinator to describe this borderline case: staff who know the application complete it faster than customers. The working decision is to include people who match real roles, devices and experience. Write down who needs to act and which evidence must be available if another person takes over.

Turn the agreement into a choice that someone can check. Use this method when the team needs to recruit relevant users. Start by describing the first option and the situation in which it is appropriate. Then write down what changes in a second situation. Use the terms staff and customers already know. Where a term has a different meaning on each team, keep the internal codes in the workflow documentation and define the plain-language label people need to see. If staff consistently interpret a decision differently, bring the people who own each interpretation together before estimating implementation.

Test cases are easier to approve when they include a boundary. The participant coverage sheet should show how this decision contributes to evidence that representative users can complete the task. Pick a value close to a threshold, a request close to the end of a service window or a record missing one required detail. Have two reviewers describe what should happen, independently. Compare the result with the written rule and check the timing as well as the visible response. If the rule needs a judgement call, name the role that can make it and say where the next person can find the outcome.

Capture the alternatives that the project explicitly considered. Confirm the outcome with the project coordinator before signing off on this part of the work. A quote may look cheaper because it assumes an existing function will stay in place, or a workflow may appear shorter because a decision has been moved to a different team. Write a short reason for the selected approach next to the estimate. When a dependency later changes, that record helps people determine which part should be re-estimated. The decision is complete when a new reviewer can apply it to the example without needing the original conversation.

Hand the participant coverage sheet to a second reviewer and ask them to apply the decision independently. If a different result emerges, the project coordinator should clarify the underlying rule before work is approved.

Test discovery

Put the designer in the review with the people who will receive the output. Use the navigation observation notes to describe how authority moves through the case. Suppose participants search for a service under a different term. The team should find where that wording belongs and retest the route. Check who is allowed to do this, when they can do it and how they hand the outcome back.

Check who can do the work, who can approve it and who needs to see the result. Use this method when the team needs to test discovery. These are separate questions. An employee may see a status without being allowed to edit the fields that determine it. A reviewer may have authority for one department while a central team needs the aggregate outcome. Confirm the exact action and which record it affects. Test a user who belongs to more than one group, because overlapping access often behaves differently from the simplest demonstrations.

Ask which changes need an independent check. The navigation observation notes should show how this decision contributes to evidence that representative users can complete the task. Some edits only affect the current draft. Others can change what a customer is told, make an earlier report incorrect or move ownership to another staff member. Confirm who needs to know, and which previous values should remain available for review. One practical check is to replay a normal case using each relevant role. Compare what each person can see before and after the decision, then verify the outcome from the perspective of someone who received the handoff.

Write down what happens if the usual person cannot act. Confirm the outcome with the designer before signing off on this part of the work. Include someone who no longer works on the account, a team member covering a short absence and a customer who needs access restored. The handover should have an intentional owner and a clear point at which a temporary arrangement ends. Avoid designing around unrestricted support access as a default. Give staff the narrow action they actually need and leave a trail that an authorised reviewer can follow. Approve the roles before creating large sets of production accounts or moving historical records.

For acceptance, inspect the result as the designer and as the next person who needs it. Write the difference between permitted actions and useful visibility into the navigation observation notes.

Test form entry

Use the representative booking example with a screen or a written walkthrough. Have the researcher present the scenario without revealing the expected answer: users cannot tell why a phone field failed. The intended response is to show useful errors and preserve correctly entered values. Ask participants what they saw, what they would do and how they would know they had succeeded.

Prototype the moment when a person decides whether to continue. Use this method when the team needs to test form entry. The interface needs to answer a practical question at that point. Put the context, available action and consequence where it is visible on the device people actually use. Check the copy with realistic words, dates, technical names and numbers. A concise draft title can fit perfectly even when the real title occupies several lines and pushes the expected action below the fold.

Run a walkthrough with someone who does not already know the expected path. The representative booking example should show how this decision contributes to evidence that representative users can complete the task. Provide a goal and a sample record, then see which detail the person looks for first. Ask what they expect to happen before they take the action, and again after the response appears. Note where the interface and the process owner are telling different stories. A confirmation that sounds final can confuse people if a human must approve it later. A status that sounds temporary can create unnecessary follow-up when the work is complete.

Write confirmations and errors with the same level of care as the first screen. Confirm the outcome with the researcher before signing off on this part of the work. The confirmation should identify what happened, whether anything remains pending and who should act. An error should identify the affected step and preserve useful work wherever appropriate. Use a phone, a keyboard and a slower connection during testing. Include a path where users return later, after they have forgotten the details of the previous session. This checks whether the interface still communicates enough for the person to recover and finish without calling your team for routine guidance.

Ask a participant to finish the task without explanation from the researcher. Add the observed reading, decision and result to the representative booking example, then retest confusing steps.

Check interrupted journeys

Run the scenario where a customer closes the browser halfway through an appointment request. Identify the point at which the normal path stops working. Ask the support lead to mark the expected recovery in the return path scenario. The proposed handling is to decide what can be recovered and what must be confirmed again. If that requires an extra decision, record who is responsible and what the customer or staff user should see.

Review what happens when the expected information is incomplete, late or contradicted by a newer record. Use this method when the team needs to check interrupted journeys. A checklist that only follows a happy path will not show which team has to repair a failure. Start with the visible effect. Determine whether the action was accepted, rejected or is still pending. Keep the underlying reasons available for the appropriate operator while using language the customer or staff member can act on.

Talk through what a safe retry means for your workflow. The return path scenario should show how this decision contributes to evidence that representative users can complete the task. Repeating an action may be appropriate when it was never accepted. Repeating a different action can create duplicate requests, alter an accepted order or send a misleading notification. Decide where the source of truth sits and how the reviewer finds it. Try the same interruption at two points in the workflow. One should happen before a record is accepted and the other immediately afterwards. Verify the resulting states, including any customer confirmation and follow-up task.

When an exception needs a person, identify the role and the latest useful time for response. Confirm the outcome with the support lead before signing off on this part of the work. Show the information the owner needs alongside the action, instead of assuming staff will search for related records across several systems. Choose whether the item stays in a queue, returns to its originator or moves to a separate review step. Write down how you will identify similar cases after release. A few real examples are often enough to distinguish a broken promise from a legitimate exception, and they make follow-up decisions easier to explain to the people doing the work.

Use the return path scenario to record the final state after a repeat attempt. Have the support lead confirm whether further action would be safe and describe how the team would notice a duplicate or unresolved item.

Review device and access support

Give the accessibility reviewer the keyboard and text-size evidence and an outcome to verify. Assume a mobile keyboard covers the only way to continue. The proposed check is to test reachability and a way to recover without losing work. Compare the initial information, the action actually taken and the final record. Save the result alongside the rule so a second reviewer can see exactly what was tested.

Decide how results will be checked while you still have time to change the workflow. Use this method when the team needs to review device and access support. An attractive dashboard is of little help if its numbers cannot be tied back to a source record. Start with one question someone in the business needs answered and select a small set of evidence they can review. For a count, define included states and date boundaries. For a handoff, identify the source and destination and verify the record that connects the two. For an edited item, keep its previous value where the reviewer needs it.

Choose a sample that exercises variation. The keyboard and text-size evidence should show how this decision contributes to evidence that representative users can complete the task. Include something accepted on the first attempt, something corrected and something the team declined. Reconcile on both sides of any handoff. A row count can match even if the wrong people own the records or if categories moved unexpectedly. Check key field values and agree how staff can trace a mismatch. If the sample cannot support a conclusion, use it to name the next test rather than reporting a confirmed result.

Agree who approves the outcome and how another person can repeat the check. Confirm the outcome with the accessibility reviewer before signing off on this part of the work. Save representative inputs, the observed result and any useful code, account or transaction reference in a place the authorised team already uses. Put questions affecting approval next to the evidence rather than scattering them across messages. An open question can stay in the record with a named owner and deadline. A decision-maker should be able to distinguish verified behaviour from a provisional assumption when planning the release or estimating follow-up work.

The accessibility reviewer should sign off against records that can be traced back to the source. Include the expected and observed outcomes in the keyboard and text-size evidence, along with any change still waiting for approval.

Check staff counterparts

Ask the operations lead to use the accepted request handoff to rehearse the routine task. Include an exception where the buyer sees success but staff cannot find the request. The planned action is to trace the accepted record and verify owner responsibility. Capture who can make the change, what notice the rest of the team needs, and the point where the issue should move to someone with authority.

Before handover, map recurring work to the people who will actually perform it. Use this method when the team needs to check staff counterparts. Many projects describe the initial setup in detail while leaving routine ownership uncertain. Decide who can maintain information, who handles a service exception and who must be consulted if the process changes. Match the responsibility to the tools that role can access. Check that a person who owns the work is able to find and update the relevant record, with an approval step when one is genuinely needed.

Ask the operating team to demonstrate one normal task and one recovery action. The accepted request handoff should show how this decision contributes to evidence that representative users can complete the task. Give them the written material that will be available after the project team leaves. Observe which decisions they can make independently and which details still require someone to interpret the system. Update terminology, instructions or permissions when the evidence shows a gap. Record the point at which they should stop and escalate, because “contact support if necessary” leaves an important operational decision undefined.

Keep the material sized for the job. Confirm the outcome with the operations lead before signing off on this part of the work. A short runbook that identifies the right account, field and contact can be more maintainable than a long manual. Use a complete example where possible and show who approves later changes to the instructions. Recheck it after a staff change, an integration update or a release that alters the original action. If external support is involved, agree how the internal owner submits a useful report and what output they should receive. Staff should know which conclusions they can draw themselves and which require review.

Give an authorised colleague the accepted request handoff and ask them to do the routine job. The operations lead can confirm whether the handover is usable when that person can complete the work and identify when it needs escalation.

Prioritise observations

Add the severity and action log to the launch record, with the product owner as its reviewer. Test this case before completing the release: a visual issue is memorable but rarely affects completion. The release plan should rank evidence by impact on target tasks and who is affected. Ask the reviewer to record the evidence and any follow-up rather than approving from a screen capture alone.

Put the release sequence in the order someone will carry it out. Use this method when the team needs to prioritise observations. List the prerequisites, which checks can be made in advance and which require the new environment or real handoff. Include the people who can decide whether a problem is severe enough to delay further changes. Decide what gets checked first when several teams are waiting on the same release. Use a short sample with known expected results so a new operator can tell whether the deployment is working.

Plan what recovery means before switching over. The severity and action log should show how this decision contributes to evidence that representative users can complete the task. In some projects the team can pause publication while fixing a problem. In others, a data change or new process makes an older version unreliable without reconciliation. Write down which changes can be reversed, what evidence must be retained and who owns the decision. Rehearse the procedure in a test environment where practical. Record any exception that has no realistic recovery path and agree how to avoid making that change before the team is ready.

After the switch, check one end-to-end action from the perspective of a real user and the person processing the result. Confirm the outcome with the product owner before signing off on this part of the work. Inspect confirmed records and any notices staff must act on. Keep operating notes for delayed feed updates, approval decisions or unusual traffic rather than accepting a single successful screen as proof. Agree when monitoring can return to its usual level. Use a shared launch record to make any postponement, partial rollout or follow-up visible to the people who need to act on it.

Keep the severity and action log in the release record. Ask the product owner which result should pause further changes and what minimal proof is required before approving a restart.

Retest the significant changes

Review the before-and-after observations with the QA reviewer once there is enough evidence to see the result. Suppose a navigation adjustment makes a second route harder. The useful follow-up is to repeat both tasks on representative devices. Contrast the first observation with a second representative case and note which part of the work could explain the outcome.

Choose measures that tell you whether people got the result they were trying to achieve. Use this method when the team needs to retest the significant changes. A new count is useful when the business can identify the action it represents and compare it consistently across periods. Record the time window, the people or records included and when staff consider an outcome final. Where the event is recorded by one system but its value is decided by another team, compare a small sample to confirm that both sides describe the same transaction.

Use changes to ask better questions before assigning causes. The before-and-after observations should show how this decision contributes to evidence that representative users can complete the task. A rise or fall can reflect differences in demand, data collection, staffing or the actual experience. Split the cases by a dimension you can act on, such as role, territory, entry path or device. Examine representative records from the groups that behave differently. Document when other changes happened and whether an unusual batch or marketing event could explain the result. Review meaningful cohorts after staff have had time to finish the work rather than comparing a partially reviewed period with a completed one.

Decide who will apply the result. Confirm the outcome with the QA reviewer before signing off on this part of the work. A metric with no decision attached is easy to collect and hard to use. For each figure, identify how it affects the next content edit, operating practice or project choice. If the team disagrees about the measure, write down the result each person expected and work backwards to define it. An evidence-based priority can then go into a backlog with an owner, a proposed change and a way to verify that the change helped the people who use the process.

Agree the next business decision with the QA reviewer, then check whether the before-and-after observations contains enough evidence to support it. Where evidence is missing, write the next verification task.

Build a release decision

Assign the accepted task and issue register to the business lead as an ongoing responsibility. Agree what should trigger review. One such case occurs when one supported user group still cannot finish a required step. Ask the owner to assign the blocker or agree a workable support route before launch. Capture the decision and the evidence that would merit revisiting it again.

Plan how the decision will be maintained after the original release team moves on. Use this method when the team needs to build a release decision. An operating rule can become inaccurate when people change roles, products are reorganised or a connected service alters its behaviour. Give an owner a reason to return to the decision, such as repeated exceptions, an upcoming renewal or an approved process change. Keep enough of the original context to distinguish what is necessary from what was simply convenient during the first rollout.

Review the current state with somebody who deals with exceptions. The accepted task and issue register should show how this decision contributes to evidence that representative users can complete the task. A monthly summary might hide a recurring problem if people are fixing it informally. Ask where they work around the process and how they would recognise a stale record. Compare their examples with the written expectation. Consider whether new information belongs in the interface, in a training document or in a role that handles the case. Choose a narrow change whose effect can be checked against normal work before expanding it to more teams or audiences.

Record a future owner, the evidence they will review and the smallest change that would warrant testing again. Confirm the outcome with the business lead before signing off on this part of the work. Make this part of the same place the team already uses to review production issues, approvals or content changes. Before closing the project, confirm where draft decisions, historic records, support messages and release notes are kept. Review access as well as accuracy: a useful record is one a new authorised team member can actually find. Take a fresh sample after the next substantial change and compare it to the agreed case, keeping the record current as the business evolves.

Put the next review into the regular planning cycle. The business lead should note which changes to conditions will make the accepted task and issue register outdated and who can approve its replacement.

From the brief to the first review

Start with realistic instructions

Give users a goal and expected outcome without naming buttons. Ask them to find a service, compare an option and complete a representative form. Watch where they hesitate and note what information they expected earlier.

Test failure and return paths

Include an invalid entry, a back navigation and a draft opened on another visit. Check keyboard access, focus and phone layouts. Ask what the user believes happened after each system response.

Turn observations into release decisions

Separate harmless hesitation from a failure to complete a core task. Prioritise repeatable blockers, name an owner and test the same task after correction. Record issues postponed to a later release.

Common question

How many people should test each journey?

Choose participants who represent the intended roles and devices. Continue until the team understands what prevents completion and has retested the most consequential fixes.

Related reading and services

Discuss your project with CESAWS

Planning a digital project?

Tell us what you are building and what you need help with.

Discuss your project with CESAWS