User Testing Guide: How to Get Real Feedback That Improves Your Site

Why User Testing Produces Insights Nothing Else Can

⠀

User testing is the practice of observing real people attempt specific tasks on your website while verbalizing their thought process. It is the only research method that directly reveals what users think as they interact with your design — not what they remember afterward, not what they say they would do, but what they actually think in real time.

The insight quality from even a small number of user tests is consistently surprising to teams that have only used quantitative methods. Analytics tells you that visitors are dropping off at your pricing page — user testing shows you that they're confused by your pricing structure because it uses industry terminology they don't recognize. Analytics tells you that form completions are low — user testing shows you that a specific field is generating anxiety because visitors aren't sure if their answer is correct.

These qualitative insights drive design improvements that quantitative data alone can never produce. They reveal the "why" behind the "what" — and the "why" is where the improvement opportunity lives.

⠀

When to Use User Testing

⠀

User testing is most valuable at specific points in the design and optimization process.

Before launch: Testing wireframes or prototypes with target users before development begins catches fundamental usability problems while they're cheapest to fix. A thirty-minute test that reveals a critical navigation problem saves weeks of post-launch remediation.

After a redesign: Testing the new design against specific tasks from the old design reveals whether the redesign improved, maintained, or degraded usability. Comparative user testing produces data that informs whether to launch or iterate.

When analytics shows a problem without explaining it: A high drop-off rate on a specific page is a quantitative signal that something is wrong. User testing diagnoses what.

For ongoing optimization: Regular monthly user testing with one to two participants builds a continuous understanding of how users' needs and behaviors evolve.

⠀

Recruiting the Right Participants

⠀

User testing produces useful insights only when the participants represent your actual users. Testing with the wrong people produces misleading data that can drive counterproductive design decisions.

The primary recruiting criterion: participants should closely match the demographic, technical sophistication, and use-case characteristics of your target audience. A SaaS product's target users may be marketing professionals with intermediate technical skills. An e-commerce site's target users may be general consumers with broad age and technical ranges.

How to find participants:

• Existing customers (most relevant, often most willing to help if asked well)

• Customer panels or advisory groups

• Professional recruiting services (UserZoom, UserTesting, PlaybookUX) for quick recruitment against specific criteria

• Social media posts targeting your specific audience segment

• Usability testing platforms that provide panel access

⠀

Five participants is the widely cited minimum for discovering the majority of usability issues on a design. Research by Jakob Nielsen shows that five testers reveal approximately 80% of usability problems. This makes user testing accessible — you don't need 50 participants to get actionable insights. For more complex sites or specific segment comparisons, 8-12 participants per segment is more appropriate.

⠀

Designing Test Tasks

⠀

Test tasks are the core of a user testing session. They should be specific, realistic, and free of leading language that suggests how the task should be completed.

Scenario-based tasks work better than direct instructions. Instead of "Find the pricing page," use "You're interested in starting a subscription — find out how much it costs." The scenario provides realistic motivation context without revealing where the answer is.

Good task design principles:

• State the user's goal, not the navigation path

• Use language from the user's world, not internal terminology

• Test tasks that represent real user needs your site serves

• Include at least one task for each major conversion goal (purchase, sign up, contact, download)

⠀

Design four to six tasks for a 45-60 minute testing session. One task typically takes 5-10 minutes including the think-aloud narration and follow-up questions.

Pilot test your tasks. Before running your full testing sessions, run one pilot test with a colleague. Pilot testing reveals ambiguous task wording, unrealistic task scenarios, and session pacing problems.

⠀

⠀

⠀

The Think-Aloud Method

⠀

The think-aloud method — asking participants to verbalize their thoughts as they work through tasks — is the most reliable way to collect qualitative behavioral data in user testing.

Brief participants before the session begins: "As you work through these tasks, please say everything you're thinking out loud — what you're looking for, what you expect to happen when you click something, whether something confuses you. There are no right or wrong answers; we're testing the website, not you."

This framing removes performance anxiety and encourages genuine verbalization. The "testing the website, not you" statement is particularly important — it clarifies that confusion or difficulty reflects on the design, not on the participant.

Active listening during the session: Your role during the session is to listen and observe, not to guide or assist. When a participant struggles, resist the urge to help. Their struggle is your data. If they ask "Is this right?" respond with "What do you think?" to keep them in self-directed mode.

Ask follow-up questions after tasks complete: "Can you tell me more about why you went there first?" "What were you expecting to find on that page?" "What would you do next if you were doing this for real?"

⠀

Moderated vs. Unmoderated Testing

⠀

User testing can be conducted in two modes: moderated (a facilitator is present and interacts with the participant) and unmoderated (participants complete tasks independently, recorded by software).

Moderated testing allows follow-up questions, task redirection when participants are completely lost, and the kind of nuanced probing that produces deeper insights. It requires scheduling coordination and is more time-intensive. It produces richer qualitative data per session.

Unmoderated testing is conducted through platforms like UserTesting, Lookback, or Maze. Participants self-schedule, receive recorded task instructions, complete tasks, and verbalize their experience without a live facilitator. Results arrive faster and at lower cost per participant. The tradeoff is the loss of real-time follow-up questioning.

For most teams, a hybrid approach works well: moderated testing for deep investigation of specific problems, unmoderated testing for broader validation and faster turnaround.

⠀

Analyzing and Presenting User Testing Findings

⠀

Raw user testing data — video recordings of sessions — requires structured analysis to produce actionable findings.

For each task: Note whether each participant completed the task successfully (success/partial success/failure), the time taken, and the key behavioral observations and direct quotes.

Pattern identification: After reviewing all sessions, look for patterns across participants. If three out of five participants struggled with the same form field, that's a pattern. If two participants mentioned the same confusing terminology, that's a pattern. Individual observations are anecdotes; repeated patterns are findings.

Prioritize by severity: Rate each finding on two dimensions — frequency (how many participants experienced it) and severity (how much it impacted task completion and user experience). High-frequency, high-severity findings are your priorities.

⠀

⠀

Present findings with evidence. Each finding should be supported by specific observations and, ideally, direct participant quotes. "Participants were confused by the pricing structure" is an assertion. "Four of five participants couldn't identify which plan included the feature they needed, with two saying 'I can't tell what's different about the plans'" is a finding with evidence.

⠀

Blakfy's User Testing Process

⠀

At Blakfy, user testing is integrated into every major design project. Before any significant design decision is implemented at scale, we validate it with real users. This approach prevents expensive post-launch fixes and consistently produces designs that perform better in A/B testing than designs produced without user validation.

The feedback loop from user testing to design iteration to retesting is where the most significant improvements emerge. Each round of testing surfaces refinements; each round of redesign incorporates them. Three cycles of user testing on a key page consistently produce a substantially better design than the first iteration — regardless of the initial design quality.

⠀

Frequently Asked Questions

⠀

How is user testing different from surveys?

Surveys ask users what they think or would do; user testing observes what they actually do. The gap between self-reported behavior (surveys) and actual behavior (user testing) is large and consistent. Users say they would search the navigation; they actually look at the hero image first. User testing is behavioral observation; surveys are self-reported opinion.

Can you do user testing on a website that isn't live yet?

Yes. User testing can be conducted on clickable prototypes (built in Figma or similar tools), on staging environments, or even on paper wireframes (for very early-stage testing). Testing early — before significant development investment — is preferable because the cost of acting on findings is lower.

How much does user testing cost?

DIY moderated testing with recruited participants from your existing customer base can cost almost nothing beyond time. Remote unmoderated testing through platforms like UserTesting costs approximately $50-100 per participant. Professional moderated user testing through a research agency costs $300-500 per participant. Given the cost of post-launch redesign work that testing can prevent, the ROI is consistently strong.