How to run a 5-person usability test on your app for about $30

You built the onboarding, so you can't really see it anymore. You know which button matters, what the icon means, and why step three asks for a phone number. Someone opening the app for the first time knows none of that. Most of them won't tell you where they got lost, either. They just close it.

A small usability test gets that view back. It doesn't take a research team or a sales call. It takes five strangers and an afternoon of watching. This is how I'd run one. I run a testing marketplace, so there's a disclosure near the end, but every step before it works the same however you find your five people.

Why five people

Jakob Nielsen's well-known argument is that five users find about 85% of the usability problems in a design, and each extra user mostly repeats problems you've already seen. His advice is to split the budget into several small rounds: test with five, fix, test with five more.

The same article has two caveats. The math assumes your users are roughly one group. If you have two very different audiences, say teachers and students, test three or four from each. And five people find problems; they don't measure anything. If you want a conversion rate, that's analytics.

Step 1: pick one flow

Not "the app". Pick one path with a clear end that every new user has to get through:

• Sign-up to the first useful screen
• Creating the first thing your app makes (a habit, a budget, a playlist)
• The paywall, from the moment it appears to a decision
• Checkout

If you can't choose, take the flow right after install. Every user sees it, and you've looked at it so often you no longer notice it.

Step 2: write tasks as goals, not instructions

Most first tests go wrong here. If the task says "Tap the + button and add a habit", you've told the tester where the button is, which is the exact thing you wanted to find out.

Write what the person wants to get done, in their words, and leave the how to them:

Three to five tasks fill 10 to 15 minutes. Put them in the order a real user would meet them.

Step 3: the script

This is the whole thing I'd send. Copy it and swap in your tasks.

Thanks for doing this. Please record your screen and talk out loud the whole time: what you're looking at, what you expect to happen, anything that confuses you. There are no wrong answers. We're testing the app, not you.

1. Install the app from [link] and open it. Before you tap anything, tell us what you think it does.
2. [Goal task 1]
3. [Goal task 2]
4. [Goal task 3]
5. If you could change one thing about what you just used, what would it be?
6. Would you keep this app after today? Why or why not?

The first task does more work than it looks like. "What do you think this does?" tells you whether your first screen explains the app. If three of five people guess wrong, that's your first finding, and nobody has reached the flow yet.

Don't tell them how long anything should take, and don't explain the app beforehand. Your real users won't get a briefing.

Step 4: find five people who have never seen it

They need to be new to the app and reasonably close to your audience. Roughly from cheapest to most expensive:

• Friends of friends, not friends. Friends are polite, and they already know what you were going for. Someone one step removed is a better stand-in for a stranger.
• Communities your users already spend time in: a subreddit, a Discord, a Facebook group. Read the rules first. Many don't allow test requests, and the ones that do usually expect something back.
• Your waitlist. People who signed up but haven't opened the app yet are close to your audience by definition. Offer a small thank-you.
• A paid testing service. You pay per person and skip the scheduling. Prices vary a lot: some of the big platforms sell by quote or annual contract, and smaller marketplaces charge per tester.

Whichever route you take, ask for a screen recording with their voice, not a written summary. People describe what they think they did, and it often isn't what they did. A recording shows the four seconds of hesitation before they tap the wrong tab.

Step 5: watch with a sheet open

Watch each recording once all the way through, then again with a sheet:

What I write down:

• Hesitation. A pause of several seconds before a tap usually means the screen didn't make the next step obvious.
• Backtracking: into a screen, back out, back in again.
• Saying one thing and doing another, like "OK, that's easy" followed by the wrong tap.
• Workarounds: finishing the task by a route you didn't design, like closing the app to get "back".
• Their words. If everyone calls it "my list" and your tab says "Collections", the tab is wrong.

Severity doesn't need to be scientific. High means the task failed or nearly did. Medium means they got there but struggled. Low is a moment of friction.

Step 6: fix three things, then test again

After five recordings you'll usually see the same two or three problems again and again. Fix those first. It's tempting to redesign everything after one round, and I'd resist it. The point of small tests is that the next round tells you whether the fix worked.

Then run five new people through the same tasks. They have to be new, because the first five have learned your app by now.

What it costs and how long it takes

Recruiting yourself is free, and you pay in time. Expect a day or two of messages before five people actually send a recording.

On the marketplace I run, a screen recording costs $5.99 per tester, so five people come to $29.95. Some numbers from our side: 54 recorded campaigns went live after 1 June and got at least five recordings. For half of them, the fifth recording arrived within 15.4 hours, and 44 of the 54 had it within three days. Across all 825 recordings since March, the median session lasted just under 14 minutes.

Disclosure

I run TestFi. Testers there record their screen and talk through your tasks on their own phones, and each recording comes with a transcript and timestamps for the moments they got stuck. If you'd rather not recruit, it's here. Everything above works the same with five people you found yourself.

If you try the first-screen question ("what do you think this does?"), I'd like to hear in the comments what people guessed.