A DIY website usability test: hand over your site and say nothing

You know your website too well to judge it. You know that “Solutions” means the product list, that the enquiry form sits under “Company”, and that delivery times are in a PDF at the bottom of one page. A visitor knows none of this, and nobody sits beside them to explain.

A website usability test takes you out of the picture. One person who does not know your business tries to do three things on the site, and you watch without speaking. Every moment at which you want to say “no, it is up there” marks a place where the website only works when you are in the room.

It takes about an hour per person and needs no software: a table, a device, paper and a pen. The hard part is the silence.

The right tester is someone who could plausibly be a customer but has never dealt with your company. Knowledge of your products, your vocabulary or your menu spoils the result, because the tester then finds things the way you do.

• Good: a friend from another industry, a neighbour, a business contact who buys similar things from other suppliers.
• Good, once: a new employee in their first week, before they learn the product names.
• Weak: a family member who has listened to you talk about the business for years.
• Not suitable: colleagues, and whoever designed or built the website.

Three to five people are enough for a small company website. Test them one at a time, never as a group, because the second person simply follows the first. Ask at least one of them to use a phone.

The tasks come from what the website is for. Pick three things a real visitor arrives wanting to do and that matter to your business: finding the right product, checking whether you can deliver, sending an enquiry, deciding whether you can be trusted.

Describe a situation, not a route. A task that names the menu item is an instruction, and following instructions tells you nothing. Avoid the exact words your navigation uses.

Give each task a clear end, so both of you know when it is finished. Put the easiest task first, so the tester relaxes. Write each task on its own piece of paper.

1. Let the tester use their own device if possible, and start on the home page or on a search engine, as a real visitor would.
2. Say the same opening to everyone: “I am testing the website, not you. You cannot do anything wrong. Please say aloud what you are looking for and what you expect. I will not answer questions about the site until the end.”
3. Hand over the first task and read it aloud once.
4. Watch and stay quiet. Do not nod, sigh or point. If the tester asks “Should I click here?”, answer: “What would you do if I were not here?”
5. Let them get stuck. If nothing moves for about three minutes, say “That is fine, let us go to the next one”. Do not show the answer yet.
6. After the third task, ask three questions: what was hardest, what did you expect to find that was not there, and what does this company do, in your own words.
7. Only now answer their questions, and thank them properly.

Write down what people do, not what they think of the design. “I do not like the green” is an opinion. “Scrolled past the enquiry button twice” is a finding.

After the last session, put all notes side by side and count per task how many testers finished without help. The chart below is an illustrative example, not real data: five testers before the fixes and five new testers afterwards.

In this example the delivery time is the obvious place to start: four of five testers failed there, and it blocks an order. Sort your own findings the same way. Problems that several testers hit come first, and problems that stop a business task come before cosmetic ones.

Most fixes are small and cost little.

If the notes show that the structure itself is the problem, not the wording, the next step is a proper look at navigation and page design, and sometimes a redesign. Visitor statistics complement the test: analytics show where people leave, and the silent test shows why.

Repeat the test after every larger change, and otherwise about twice a year. Always use new people. A tester who has seen the website once is no longer a stranger to it.