What is usability testing?

How many participants do you need?

Fewer than most people expect for finding problems, and more than most people expect for measuring anything.

For finding problems: five per user group. A widely used planning heuristic rather than a threshold. The reasoning behind it is that the same problems recur across participants quickly, so each additional session returns less than the one before. Running three rounds of five, fixing between rounds, beats one round of fifteen.

For measuring anything: far more. Comparing task completion rates between two designs is a statistical comparison and needs a sample sized accordingly. Usability testing is a qualitative instrument, and treating five participants' completion rates as a measurement is a common error.

Distinct user groups need their own five. If technical evaluators and procurement buyers use the site differently, they are two groups.

Key takeaway: five participants to find problems, a sample sized for the comparison to measure whether one design beats another.

Moderated or unmoderated?

Moderated allows follow-up. When a participant hesitates, you can ask what they expected. That single capability is what makes it the better instrument for complex B2B flows.

Unmoderated is faster and cheaper and scales, at the cost of losing the ability to ask why. It suits comparing two navigation structures or checking whether a specific task is completable.

A practical combination is unmoderated first to find where people struggle, then moderated sessions focused on those points.

What does usability testing not tell you?

Whether people want it. Usability testing measures whether an interface can be used, not whether the offer is compelling. A perfectly usable page for a product nobody wants converts at zero.

How common a problem is. Five participants hitting an issue tells you it exists. Quantifying prevalence requires analytics or heatmaps.

What people would actually do. Participants know they are observed and are attempting an assigned task rather than pursuing their own goal. They persist longer than real visitors, so tasks that get completed in testing can be abandoned in reality.

Whether your fix works. That is an A/B test.

Anything reliable from the wrong participants. Testing a B2B SaaS buying flow with people who would never buy the product produces confident findings about the wrong audience. Recruitment quality decides whether the study was worth running at all.

Related terms

Five second test · Tree testing · Voice of customer · Session recording · Cognitive load

Service: Conversion Rate Optimization.

FAQ

Is five participants really enough?

For surfacing the problems that recur across users in one group, yes, and returns diminish quickly after that. It is a planning heuristic rather than a guarantee: a problem that one visitor in fifty hits is unlikely to appear in five sessions. For measuring differences between designs, five is far too few, since that is a statistical comparison.

How is usability testing different from session recordings?

Session recordings capture real visitors pursuing their own goals, with no ability to ask questions. Usability testing uses recruited participants on assigned tasks, with the ability to probe reasoning. Recordings show what happens at scale, usability testing explains why.

Can you run usability testing on a live site?

Yes, and it is the usual approach. Testing a prototype before building is cheaper, since problems surface before engineering time is spent. On Webflow there is a third option: publish the redesign to the staging subdomain, which is served separately from the live domain and kept out of search results, and run sessions against that.