Pick from the menu of UX research methods first and you will end up forcing your question to fit it. The researchers who get listened to do the opposite: they start with the question and the decision it informs, then choose the method that answers it honestly. This guide walks through how to make that choice, so you spend your effort on the study that actually moves the work forward.
Start with the decision, not the method
Before you think about interviews or surveys, write down two things:
- The question you are trying to answer. Not "let us do some research on onboarding," but "why do new users drop off before they finish setup?"
- The decision it will inform. Who will do what differently once you have the answer?
If you cannot name the decision, you are not ready to pick a method. You are ready to talk to whoever owns that decision and sharpen the question. Good method selection is mostly good question selection wearing a lab coat.
The two axes that decide most of it
Almost every UX research method sits somewhere on two axes. Knowing where your question falls narrows the field fast.
Qualitative or quantitative
Qualitative methods (interviews, usability tests, field studies) tell you why something happens and how people experience it. You get depth, context, and the language people actually use. You do not get statistical confidence.
Quantitative methods (surveys, analytics, A/B tests) tell you how many and how much. You get scale and confidence, but little explanation. A funnel can tell you that 40 percent drop off at step three. It cannot tell you why.
Most real questions need both at different moments, which is why the qualitative versus quantitative choice is rarely an either or.
Attitudinal or behavioural
Attitudinal methods capture what people say: their stated preferences, opinions, and recalled experiences. Behavioural methods capture what people do: where they click, where they hesitate, what they abandon.
The gap between the two is where a lot of bad product decisions live. People are sincere and still wrong about their own behaviour. When the stakes are high, lean toward methods that observe behaviour rather than collect opinions.
Match the method to the stage of work
The same question feels different depending on where you are in the product cycle.
- Early and exploratory (we are not sure what to build): generative methods. User interviews, field studies, diary studies. You are looking for unmet needs and the shape of the problem.
- Mid and formative (we have a direction, is it right): evaluative methods. Usability testing, concept tests, card sorts. You are pressure testing a design before it ships.
- Late and summative (it is live, how is it doing): measurement methods. Surveys, analytics, benchmarking. You are tracking whether the thing works at scale.
A common mistake is reaching for a usability test when the real question is generative, or running yet another round of interviews when you actually need to measure. Name the stage and the right family of methods usually presents itself.
A short field guide to the common methods
- User interviews · best for understanding motivations, context, and mental models. Cheap to run, easy to do badly. Strong when you want the "why" behind a behaviour.
- Usability testing · best for finding where a design confuses people. You watch real tasks and learn where they stumble. Five to eight participants surface most major issues.
- Surveys · best when you need numbers at scale or a read across a population. Powerful for measurement, treacherous for discovery, because you only learn what you thought to ask.
- Field studies and contextual inquiry · best when context shapes behaviour. You go to where people actually work or live. Expensive, and sometimes the only honest way to see the truth.
- Diary studies · best for behaviour that unfolds over time, like habit formation or a multi day task.
- Card sorting and tree testing · best for information architecture and navigation questions.
- Analytics and A/B tests · best for measuring real behaviour at scale, once you already know which behaviour matters.
How much rigour does this decision deserve
Not every question needs a flagship study. A useful filter: size the research to the cost of being wrong.
If the decision is cheap to reverse, a quick unmoderated test or five interviews is plenty. If you are about to commit a quarter of engineering time or change something that touches every user, invest in a study you can defend, with cleaner recruitment and a larger or more carefully chosen sample.
Rigour is not about always doing the biggest study. It is about doing the right sized study for the risk in front of you.
A simple way to decide
When you are stuck, run the question through these four prompts:
- What decision will this inform, and who owns it?
- Do I need to know why, or how many? (qualitative or quantitative)
- Do I need what people say, or what they do? (attitudinal or behavioural)
- What stage am I in? (generative, evaluative, or measurement)
Four answers, and the field of sensible methods is usually down to one or two. Pick the lighter one if the cost of being wrong is low, the heavier one if it is high.
The method is the easy part
Choosing a method is not really the hard skill. The hard skill is asking a sharp question and staying close to the decision it serves. Get those right and almost any competent method will give you something useful. Get them wrong and the most rigorous study in the world will answer a question nobody was asking.
People of Research is a community and newsletter for UX researchers, behavioural scientists, market researchers, and anyone who makes decisions with evidence. We share the methods, the messy bits, and the people behind the work. Join us and get the good stuff in your inbox.