Writing unbiased tasks

Tasks are the heart of a usability test. A good task is realistic, specific enough to have an end point, and does not tell people how to do it.

Scenario-based tasks

Give context and motivation: "Your daughter's birthday party is on Saturday. Order a cake that feeds 15 people and can be collected on Friday afternoon." This feels real and gives people what they need to make decisions.

Avoid giving clues

If the button says "Manage subscription", do not write "Manage your subscription". Instead: "You want to pay for your plan yearly rather than monthly. Make that change." Using interface labels turns the test into a word-matching game.

Other tips

• Avoid leading words like "easily" or "quickly".
• Include both exploratory tasks ("Find a holiday that interests you") and specific tasks with a clear answer.
• Let participants use their own details where appropriate, which increases realism.
• Order tasks naturally, but vary order between participants where learning effects matter.

Biased tasks produce misleading results: people succeed because the task told them where to click, and real problems stay hidden. Well-written tasks are the difference between a test that reveals truth and one that confirms what the team wants to hear.

A bank's first test task reads: "Click on 'Payees' and add a new payee." Everyone succeeds, but in real life, call centre logs show people cannot find how to pay someone new.

The rewritten task reads: "You owe your friend Sam £40 for concert tickets. You have their sort code and account number. Pay them back." Three of five participants look under "Transfers" first, revealing the real navigation problem.