Foldwren

Usability testing with five people

A step by step guide to running a simple usability test with five people, writing good tasks, staying quiet and deciding which problems to fix.

A usability test is simple: you ask someone to do a task with your website or app, and you watch what happens. No lab, no one way mirror, no special software. A phone, a quiet table and half an hour will do.

Five people is the number most designers reach for, because by the fourth or fifth session you tend to be watching the same problems repeat. That repetition is the signal you are after.

What a usability test is not

It helps to clear a few things up first.

  • It is not a survey. You are not collecting opinions.
  • It is not a demo. You are not showing people how the thing works.
  • It is not a test of the person. If they get stuck, the design failed, not them. Say so at the start.
  • It is not proof. Five people will not tell you how the whole market behaves. They will show you where the potholes are.

Step one: write tasks, not questions

The quality of a test comes down to the tasks you set. A good task describes a realistic goal and leaves out the instructions.

Weak: "Click on Services and find the pricing page."

Better: "Your hot water system has stopped working. Find out roughly what a replacement would cost and how soon someone could come."

The first tells the person where to tap. The second is what a real customer of a plumbing business would actually be thinking. Notice it avoids the words used in the menu, so you are not handing over the answer.

Write three to five tasks covering the things that matter most to your business. For a physio clinic these might be:

  1. Book a first appointment for a sore shoulder next week.
  2. Find out whether you can claim through your health fund.
  3. Change an existing booking to a different day.
  4. Work out where to park when you arrive.

Step two: find five people

Choose people who resemble your customers. They do not need to be existing ones. A neighbour who has never seen your site is often more useful than a regular who already knows where everything is.

Avoid people who built the thing, people who work for you, and your partner. They know too much and they are too kind.

Step three: run the session

Sit beside them, or join a video call where they share their screen. Use their own phone if you can, because that is how real customers will arrive.

A script that works:

  1. Set them at ease. "We are testing the website, not you. You cannot do anything wrong. If something is confusing, that is exactly what we want to find."
  2. Ask them to think aloud. "Say whatever is going through your head as you go, even if it seems obvious."
  3. Read the first task and hand over.
  4. Stay quiet. This is the hard part. When they hesitate, wait. When they ask "should I tap this?", reply "what would you do if I were not here?"
  5. Take notes on what they do, not on what you think it means. "Scrolled past the booking button twice" is a note. "Button needs to be bigger" is a guess.
  6. Ask a couple of open questions at the end. "What was the hardest part?" "Was anything not what you expected?"

Do not explain, defend or apologise for the design. The moment you help, you stop learning.

Step four: sort what you saw and fix the worst

Straight after each session, write down the top three problems while they are fresh. Once all five are done, put the problems in a list and mark how many people hit each one.

Then give each a rough severity:

  • Blocker: the person could not complete the task.
  • Struggle: they got there, but slowly or with obvious frustration.
  • Niggle: a brief hesitation or a comment, with no real harm done.

A blocker that stopped three of five people from booking goes to the top. A niggle mentioned once can wait.

Resist the temptation to act on suggestions. Testers will happily redesign your site for you: "you should put a big video here". Thank them and note the underlying problem instead. People are reliable at revealing where they got stuck and unreliable at prescribing the fix.

Fix, then test again

Pick the top three to five problems and fix those. Not everything. Small, targeted changes are easier to check.

Many fixes turn out to be modest:

  • Renaming a menu item to the word customers actually use.
  • Moving the main button higher, or pinning it to the bottom of the screen.
  • Removing a form field nobody understands.

Then run another round with five new people. New, because the first group has now learned the layout. If the same tasks go smoothly this time, you have real evidence the changes worked.

Start smaller than you think

If five sessions sounds like a lot, start with one. Tomorrow, hand your phone to someone who has never seen your site, give them a single realistic task, and watch in silence for five minutes. It is a slightly uncomfortable experience and almost always a useful one. Most owners come away with a short list of fixes they could not have found any other way.

Your order