What a usability test looks like, with real examples
Gideon Awolesi ·
A usability test is simpler than it sounds. You give people from your user group a realistic task, watch them try to do it with your product or prototype, and write down where they get stuck. That's it. What makes it useful is how you set it up and what you do with what you see. Here's what it looks like in practice, with examples from our own projects.
01What happens in a session
A moderator gives the participant a goal, like "borrow against your tokens", and asks them to think out loud while they work through it. The moderator doesn't help. When someone gets stuck, they ask what the participant expected to happen. A session usually covers three to five tasks, and the screen and audio are recorded with the participant's consent.
02Example: two rounds on a DeFi lending app
For Euler, a DeFi lending protocol, eight participants went through moderated remote sessions with the first design. We tested navigation, clarity, and trust signals. The first round showed that the mobile inputs were cramped and the top navigation was overloaded, so we made the tap areas bigger and simplified the navigation.
In the second round, we revised the tooltips and data displays, and compared a detailed layout with a simpler version. One participant said: "The new interface is so much easier to navigate—it finally feels like I'm in control of my funds, rather than the interface controlling me."
03Example: testing an ICU workflow before anything was built
For Bayer, we designed views that let clinicians follow a patient's progression and medication response without digging through screens. We ran moderated prototype testing across the whole workflow before anything was built, so the problems we found cost a prototype change instead of a code change.
04Example: weekly sessions and a hospital pilot
Some products need testing where they're used. For a surgical tool scanner app, we held weekly reviews with a small group of operating-room nurses and revised the interface from their feedback. The beta then ran in two pilot hospitals, with focused sessions with surgical teams to find problems and decide what to change next.
05Example: several small rounds on an AI product
For AiListz, an AI content creation platform, small groups tested the product over several rounds. Their feedback led to clearer menu labels, shorter prompts, and help with choosing a template or persona. Testing showed that people could find the main tools with less explanation.
06What you get at the end
A good test report lists every problem with how many participants ran into it and how badly it stopped them, ranked so your team knows what to fix first, with clips of the moments that matter. It should end with a decision: build it, change it and test again, or stop.
Our Evidence Sprint builds the flow you're least sure about as a working app, tests it with 5 to 8 people from your user group, and ends with that decision, in three to four weeks from €9,500.