UX Research · Step 1 · 6 min read
Research questions that lead to decisions
A study is only as good as the question it answers. How to climb from a vague business worry to a research question you can actually study, why it's different from the questions you'll ask participants, and how to write down, before you start, what result would change the plan.
Most studies that fail, fail before the first participant arrives: they answered a question nobody needed answered. "Do users like it?" can't be studied well and can't change anything. "Where do first-time buyers hesitate in checkout?" can be studied in a week and tells the team what to fix. The difference is a few minutes of thinking, and it's the most valuable part of the whole study.
What you'll be able to do after reading:
- Tell a business question, a research question and an interview question apart
- Turn a vague worry into research questions that can be studied
- Write the decision and the deciding evidence before you start
- Run a short session with your team to agree on what to learn
1. Three kinds of question
Teams mix up three different things and call them all "the question":
- The business question is the worry: "Why do so many accounts stop after the first invoice?" It's real and important, and too broad to study as it is.
- The research question is what the team needs to learn to act on that worry: "How do freelancers bill a returning client today?" It guides the whole study.
- The interview or task question is what you'll say to a participant: "Tell me about the last time you billed a client for the second time." It's written later, in The interview guide.
The GOV.UK Service Manual makes the same distinction when teams write research questions together: "remind them that these are things you need to learn, not questions you will ask users" (GOV.UK, Capturing research questions). If you ask participants your research question directly ("How do you bill returning clients?"), you'll get a tidy, generalised answer. Real behaviour comes out of specific stories.
2. Climb down the ladder
- 1 · Business question"Why do so many new accounts stop after the first invoice?"
What the team is worried about. Too broad to study as it is.
- 2 · DecisionFix onboarding, or build recurring invoices first, this quarter?
What will change depending on the answer. If nothing would change, stop here.
- 3 · Research questionsHow do freelancers bill a client the second time? What do they use instead of us, and when?
What the team needs to learn. Not the questions you'll ask participants.
- 4 · Method and peopleSix interviews with freelancers who stopped, plus the drop-off funnel in analytics
How you'll learn it, and from whom.
- 5 · Evidence that would decideMost bill monthly and copy the last invoice by hand: recurring invoices first
Written before the study, so the result can't be bent to fit.
Each step down is narrower and more concrete:
- Business question. Write the worry as the team says it.
- Decision. What will the team do differently depending on the answer, and by when? Name who decides. If you can't name a decision, read When not to do research before going on.
- Research questions. Two to five things you need to learn. More than five usually means two studies.
- Method and people. Only now choose how and with whom (see Choosing a research method).
- Evidence that would decide. What result would push the decision one way, and what would push it the other?
Step 5 is the one teams skip, and the one that protects you most. Written before the study, it stops everyone, you included, from reading the results to fit what they already wanted. It also makes the readout short: "We said X would mean recurring invoices first. We saw X."
The card Research questions to decisions turns the ladder into a template you can fill in with the decision owner.
3. What a good research question looks like
GOV.UK's advice is that broad, open questions starting with "who", "what", "when", "where" or "why" are more helpful than narrow ones starting with "do", "have", "will" or "can" (GOV.UK, Capturing research questions). From that and from practice, a good research question is:
- Open. It can't be answered with yes or no.
- About behaviour and context, not taste or predictions. People are poor at predicting what they'll do (NN/g, Screening Participants, 2024).
- Narrow enough to finish in one round.
- Tied to the decision. If the answer can't move the decision, cut the question.
- Not already answered. Check analytics, support tickets and past studies first.
| Not ready yet | What's wrong | A research question |
|---|---|---|
| Do users like the new dashboard? | Yes or no, about taste. No decision hangs on "like". | What do people look for first on the dashboard on a Monday, and can they find it? |
| Would people pay for a premium plan? | A prediction. People are poor at predicting what they'll do. | What do freelancers already pay for to get paid faster, and what made them start? |
| Validate the new checkout. | The answer is decided before the study. "Validate" means "confirm". | Where do first-time buyers hesitate or fail in the new checkout? |
| Understand our users. | No edge. The study never ends and never decides anything. | How do small shops decide which orders to ship first on a busy day? |
| How many people use the export button? | A good question, but you don't need a study: analytics already knows. | Look it up first. Then: what do the people who export do with the file? |
4. Start from your assumptions
A quick way to find the right questions is to write down what the team already believes. GOV.UK's planning guide asks teams to agree "which assumptions or beliefs about your users or service you want to test" and "what you need to know now to be able to make an informed decision about what to do next" (GOV.UK, Plan a round of user research).
- List the beliefs the plan depends on: "People stop because setup is confusing."
- For each, ask: how sure are we, and how bad would it be if we were wrong?
- Turn the riskiest ones into open questions: "What happens in a freelancer's first week with the app?"
Beliefs you're very sure of and that are cheap to be wrong about don't need a study. That idea becomes the whole next lesson.
5. Agree on the questions together
Research questions work best when the team writes them together, because then the team wants the answers. GOV.UK describes a session of about an hour (GOV.UK, Capturing research questions):
- The product owner reminds everyone of the outcomes the team is after.
- Everyone writes questions on sticky notes for 5 to 10 minutes.
- Group similar notes, write one summary question per group.
- Each person puts a dot on the 5 questions they think need answering first.
- Keep the top few for the next round; park the rest in a shared list.
A remote board works as well as a wall. What matters is that the product manager and a developer are in the room, so the questions belong to the team and not only to the designer.
6. Common mistakes
- Using the business question as the research question, then drowning in data.
- Asking participants the research question word for word.
- Questions that ask people to predict ("Would you use…?") or judge ("Do you like…?").
- No decision attached, so the findings have nowhere to go.
- Writing the deciding evidence after the study, when it's too easy to bend.
7. Vocabulary
| Term | Meaning |
|---|---|
| Business question | The team's worry or goal, in its own words |
| Research question | What the team needs to learn to act; guides the study |
| Interview question | What you actually say to a participant |
| Assumption | A belief the plan depends on that nobody has checked |
| Deciding evidence | The result, written in advance, that would move the decision |
8. Try it (no tools needed)
- Take a question your team asked recently, for example "Do people like the new home page?"
- Climb down the ladder on paper: what decision is behind it? Who decides, and when?
- Write two research questions, and one sentence that starts "If we see…, we'll…".
Reflection questions: Did the decision turn out to be different from the question you started with? Would anyone on your team disagree with your "If we see…" sentence, and is it better to find that out now?
9. If you already practise
- Put the decision and the deciding evidence at the top of every plan and every readout.
- Keep a living list of the team's open research questions. New studies start from the list, and the list shows stakeholders what research is for.
- When a stakeholder brings a method ("we need a survey"), ask which decision it's for before you talk about methods.
Checklist
Learn more
- GOV.UK Service Manual: Capturing research questions · Plan a round of user research
- Maria Rosala, Discovery: Definition (NN/g, 2020)
Books, courses and videos to go deeper
- Just Enough ResearchErika HallBook · 2nd edition (A Book Apart, 2019) (Amazon)
- It's Our ResearchTomer SharonBook · Morgan Kaufmann, 2012 (Amazon)
Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.
Related
What research is for · When not to do research · Writing a research plan · Problem statements and "How might we" · Defining success · Writing insights that change decisions