Usability testing on a budget
Five users, one afternoon and a script that finds real problems.

Zander Whitehurst
9 min read
Many teams skip usability testing because they think it needs a lab, a big budget and weeks of planning. It does not. Some of the most useful testing I have ever done involved a laptop, a cup of coffee and a slightly confused friend of a friend.
You can find most of the serious problems in a design with five users and one afternoon. Here is how to do it well without spending much at all.
Why five users is enough
Testing with a small group reveals the majority of the most significant usability problems. After the first few sessions, you start seeing the same issues again and again. Each additional user adds less new information.
That makes small, frequent tests far more valuable than one large study. Test with five, fix what you find, then test again.
Decide what you want to learn
Before recruiting anyone, write down the questions you want answered. If you are testing a prototype, those questions also decide how polished it needs to be. Can people complete sign up? Do they understand what the dashboard is for? Can they find the export option?
Clear questions lead to clear tasks, and clear tasks lead to useful results. Vague goals lead to vague sessions where everyone feels good and nobody learns anything.
Write realistic tasks
Tasks should describe a goal, not a set of instructions. "Click the settings icon and change your password" tells people exactly what to do. "You think someone might know your password. Make your account secure" lets you see whether they can work it out.
Avoid using words that appear in your interface. If the button says "Export", do not ask people to "export their data". Ask them to "get a copy of their data to share with their accountant".
Recruit cheaply
You do not need a recruitment agency. Try:
- Existing customers who have agreed to be contacted
- People in relevant online communities
- Colleagues in other departments who match your audience
- Friends of friends who fit the profile
A small thank you, such as a gift card, goes a long way. Just avoid testing with people who work on the product. They know too much.
Run the session
Keep sessions short, around thirty to forty five minutes. Start by putting people at ease. Tell them you are testing the design, not them, and that nothing they do can be wrong.
Ask them to think aloud as they work. Then, and this is the hard part, stay quiet. Do not help. Do not explain. When someone gets stuck, ask "what are you looking for?" or "what would you expect to happen?" instead of pointing at the answer.
Watching someone struggle with something you designed is genuinely painful. I have physically sat on my hands before. But every struggle is a gift, because it shows you exactly what to fix.
Record and invite the team
With permission, record the screen and audio. Better still, invite engineers, PMs and stakeholders to watch live. Nothing builds shared understanding faster than watching a real person get stuck on something the whole team thought was obvious.
Summarise quickly
After the sessions, list the problems you saw and how many people hit each one. Rank them by severity: does this stop people completing the task, slow them down or just annoy them?
Keep the report short. A page of prioritised problems with a few clips is far more useful than a long document nobody reads.
Fix and repeat
The value of testing comes from acting on it. Fix the most serious problems, then test again. Each round makes the product noticeably better.
The best teams I have worked with test every couple of weeks, almost casually. It stops being a big event and becomes a habit, and that habit is one of the most reliable ways to build products people can actually use. We test real prototypes with real people throughout our AI Product Design Bootcamp.
Memorisely