User testing, or usability testing, asks representative people to complete realistic tasks while an observer watches and, in many formats, the participant thinks aloud. It uncovers problems analytics cannot explain: unclear wording, misleading layouts and missing information. A handful of sessions is often enough to reveal the most serious issues.
Formats
Moderated tests have a facilitator who can ask follow-up questions, either in person or on a video call. Unmoderated tests give participants written tasks and record their screen and voice, which is faster and cheaper to run at scale. Each has a place: moderated sessions for deep understanding, unmoderated for quick checks.
Running a useful session
Recruit people who match your target buyers, not colleagues or friends who already know the product. Write tasks as goals rather than instructions, for example “Find out whether this tool works with your accounting software” rather than “Click Integrations”. Stay quiet while they work and note where they hesitate, what they say and what they expected to see.
- Define two to five realistic tasks tied to business goals.
- Recruit participants who match the target audience.
- Ask them to think aloud and avoid leading prompts.
- Record sessions and note problems with timestamps.
- Group issues by severity and frequency before fixing.
Where it fits in CRO
User testing is a research method, not a validation method. It tells you why visitors struggle and generates strong hypotheses. Whether a fix improves conversions at scale is then a question for an A/B test or careful before-and-after measurement. On low-traffic B2B pages, user testing can be the most practical way to find improvements.
Example
Testing a pricing page with buyers
Hypothetical procurement tool Ashgrove Kite asks five finance managers to work out which plan they would need for a 40-person team. Three of them misread a usage limit as per user rather than per company. The team rewrites the plan table labels and adds a short worked example.
Common mistakes
Testing with insiders
Employees and friends know too much and are too polite. Recruit people who match real visitors.
Asking opinions instead of watching behavior
“Do you like this page?” produces guesses. Watching someone try to complete a task produces evidence.
Questions
For finding usability problems, small rounds of a few participants, repeated after fixes, are a common approach. For measuring task success rates precisely, you need more participants.
