Every team says they'd test more if they had the budget. In my experience, budget is rarely the real blocker. The blocker is that usability testing sounds like a big project: a lab, a recruiting agency, a research team and a 40-page report.
You don't need any of that to learn something useful. You need five people, a prototype, a few clear tasks and one afternoon. That's enough to find the problems that are hurting your product the most right now.
This is the exact way I run small tests, the same approach I first practised during my UI/UX diploma at AND Academy and now use on client work at Ksolves.
Why five users is enough (most of the time)
The idea comes from Jakob Nielsen at Nielsen Norman Group, who has argued for years that small, repeated tests beat one big study. You can read his reasoning in Why You Only Need to Test with 5 Users.
The short version: the first few people will trip over the biggest problems. By the fifth, you're mostly seeing the same issues again. So rather than spending everything on one large round, test with five, fix what you found, then test again.
Two honest caveats:
- Five works when your users are similar. If you have two very different groups (say, shop owners and their customers), test a few of each.
- Five is for finding problems, not for measuring things. If you need numbers you can defend, like a success rate, you need more people.
Step 1: Decide what you want to learn (30 minutes)
Write down one question the test should answer. Not "is the app good?" but something specific:
- "Can new users add money to their wallet without help?"
- "Do people understand what the Cry Report is telling them?"
- "Can a shop owner find last month's invoices?"
Then turn it into 3 to 5 tasks. A good task is a realistic goal, not an instruction.
- Weak: "Tap the Transfer tab and send money."
- Better: "Your friend paid for dinner. Send her ₹600 for your share."
The weak version tells people where to click. The better one tests whether they can find it.
Tip
Read each task aloud to a teammate before the sessions. If they ask "what do you mean?", rewrite it now, not in front of a participant.
Step 2: Recruit five people for free or close to it (1 to 2 days, in the background)
You don't need a recruiting agency. You need people who roughly match your users and haven't seen the design.
Places I'd look:
- Your existing users: a message in your app, newsletter or support inbox.
- WhatsApp communities related to your product. On Pukaar.ai, the team had a WhatsApp community of 400+ parents, which is a goldmine for anyone designing for that audience.
- Friends of friends who fit the profile (not your design team, not your co-founder).
- Customers in a real location, like a café, a clinic waiting area or a shop, if your product fits that context.
Offer something small as thanks: a voucher, a coffee or early access. Keep each session to 20 to 30 minutes so it's easy to say yes.
Avoid testing only with colleagues. They know too much and they're too polite.
Watch out
Ask for consent before you record a screen or a face, and say where the recording will be kept. Store notes with participant numbers (P1, P2), not names.
Step 3: Prepare the prototype and script (1 hour)
The prototype only needs to support your tasks. A clickable Figma prototype is usually enough. I sometimes deploy a quick live prototype on Vercel so people can try it on their own phone, which feels far more real than a laptop screen.
Test on the devices your users actually own. If your audience is on budget Android phones, testing on your new iPhone will hide problems.
A simple script:
- Intro (2 min): "We're testing the design, not you. There are no wrong answers. Please think out loud."
- Warm-up (2 min): One or two questions about how they do this task today.
- Tasks (15 min): Read each task, then stay quiet.
- Wrap-up (5 min): "What was the most confusing part? What would you change?"
The hardest part is staying quiet. When someone gets stuck, count to ten in your head before saying anything. And never ask "Did you find that easy?" It invites a polite yes. For more on asking good questions, read user interviews that work.
Step 4: Run the sessions in one afternoon (3 hours)
Book five sessions back to back with short gaps. Remote over Google Meet or Zoom works fine; ask people to share their screen. In person is better for mobile, because you can see their thumbs.
Bring one teammate to take notes if you can, ideally a developer or product manager. Watching a real person struggle with a feature does more to change a team's mind than any report I could write.
Note-taking template for each task:
| Participant | Completed? | Where they hesitated | What they said | Severity |
|---|---|---|---|---|
| P1 | Yes, slowly | Looked for "Send" under Profile | "I thought it'd be here" | Medium |
Write down what people did, not what you think they meant. Quotes are gold.
Step 5: Find the patterns the same day (1 hour)
Don't wait a week. Do the analysis while the sessions are fresh.
- Put every observation on a sticky note (FigJam or Miro is fine).
- Group similar notes together. This is basically affinity mapping.
- Count how many of the five hit each problem.
- Rate each problem: high (blocked the task), medium (slowed them down), low (minor annoyance).
Note
Severity is about the task, not the volume of the complaint. A participant who grumbles loudly about a colour is a low; a quiet one who gives up on payment is a high.
A problem that blocked three out of five people is your top priority, full stop. A problem only one person mentioned might still matter, but it can wait for the next round.
Step 6: Share findings in one page, then fix and retest
Nobody reads a long research report. I share one page:
- What we tested and with whom.
- The top 3 to 5 problems, each with a screenshot, a quote and how many people hit it.
- A suggested fix for each.
- What we'll test next round.
Then fix the high-severity issues and run another small round. Three quick rounds will teach you more than one big expensive study, and they fit naturally into sprints. If you want to show the impact afterwards, read measuring UX impact.
Common mistakes I'd avoid
- Testing too late. Test rough wireframes. Changes are cheap then.
- Leading tasks. If the task uses the button label, you've given away the answer.
- Helping too soon. Silence is uncomfortable, but that's where you learn.
- Treating opinions as findings. "I'd prefer blue" isn't a usability problem. "I couldn't find the button" is.
- Testing once and stopping. The value is in the loop.
Work with us
Need research done for you? Our UX research and usability testing services cover interviews, surveys and usability tests, with clear recommendations.
Key takeaways
- Five similar users are enough to find your biggest usability problems in one round.
- Write realistic tasks that state a goal, not a set of click instructions.
- Recruit from real users and communities, not colleagues, and test on their kind of phone.
- Analyse the same day, rank problems by severity and share one page of findings.
- Small, repeated rounds beat one large, expensive study.
Test this week, not next quarter
The best time to test is before you think you're ready. Pick one flow, find five people and block one afternoon. You'll come out with a short list of fixes that matter, and a team that has seen real users with its own eyes.
If you'd like to see how research shaped a real product, read the Pukaar.ai case study. Want help running a quick test on your app? Tell me about your project.