Choosing the right method

There are dozens of research methods, but most product questions are answered by a small set. Choosing well depends on two things: whether you want to learn what people do or what they say, and whether you are exploring a problem or evaluating a solution.

The core methods

• Interviews: one to one conversations about past behaviour, goals and frustrations. Best for exploring a problem space and understanding why something happens.
• Moderated usability tests: you watch a participant try realistic tasks with a prototype or live product while thinking aloud. Best for finding where a design fails.
• Unmoderated usability tests: participants complete tasks on their own with a tool recording their screen. Faster and cheaper, but you cannot ask follow up questions.
• Surveys: many people answer the same questions. Useful for measuring how common something is, weak for discovering why.
• Analytics review: product data shows what people do at scale but never explains the reason.

Matching questions to methods

Look back at our three studio questions. "Can people complete the booking flow?" is an evaluation question about behaviour, so a moderated usability test with five people is a strong choice. "Why do trial members not return?" is exploratory and about motivation, so interviews come first. Once you have a hypothesis, such as "people forget because nothing reminds them", a survey of all trial members can tell you how widespread that is.

A useful habit is to pair methods. Analytics show a drop at the payment step, a usability test shows why, and a follow up check in the data confirms the fix worked.

How many participants

For qualitative work, small numbers go a long way. Five usability test participants from the same audience usually reveal the most serious problems in a flow. For interviews, start with five to eight and stop when new sessions stop surprising you. Surveys need far more responses before percentages mean anything, often a hundred or more.

Pick the cheapest method that can change the decision. If a quick hallway test answers the question, you do not need a two week study.