Skip to main content

UX and research · Beginner · 12 min

Turn a problem statement into how-might-we questions

By Padding · Updated 18 Nov 2026
Any notes app or document, plus sticky notes or a spreadsheet · November 2026

You will have a sorted set of how-might-we questions drawn from one problem statement, with three workable questions chosen to explore and a note on why the rest were set aside.

On this page

A problem statement tells the team what is wrong. It does not tell anyone where to start looking for answers. A how-might-we question is the bridge: a short question, starting "How might we...", that turns one part of the problem into an open invitation for ideas.

The hard part is not the phrasing. Anyone can put "How might we" in front of a sentence. The hard part is scope. A question that is too broad ("How might we improve healthcare?") gives the team nothing to push against. A question that is too narrow ("How might we add a reminder text?") has already picked the answer. This tutorial spends most of its time on judging scope, because that is where these questions usually go wrong.

The running example is Larkfield Clinic, a fictional community clinic building an appointment booking app. If you worked through Write a problem statement before you design, this is the statement the team agreed there (Draft 5). If you did not, use it as given:

Patients who book weekday morning appointments around work or childcare find out on the day they cannot attend, but the phone line is too busy to cancel, so they are marked as no-shows and wait two to three weeks to rebook, while 1 in 7 booked appointments go to waste.

You need somewhere to write: a document, a spreadsheet with one question per row, or sticky notes on a wall. Paper works too.

Step 1 · Break the statement into parts you can question

Each part of a problem statement is a different place to start asking questions. Split yours into four labelled lines before you write a single question:

  • Person: patients who book weekday morning appointments around work or childcare
  • Situation: they find out on the day they cannot attend
  • Obstacle: the phone line is too busy to cancel
  • Consequence: they are marked as no-shows and wait two to three weeks to rebook, while 1 in 7 booked appointments go to waste

Add one more line, Assumptions, and list anything the statement takes for granted. For Larkfield: that the phone is the only way to cancel, that a slot given up on the morning cannot be refilled, and that anyone who does not attend has to be marked as a no-show. These are useful later, because questioning an assumption often produces the most interesting questions.

Some statements do not name the obstacle separately; it sits inside the situation ("find out on the day, too late to cancel"). In that case, write the obstacle line as what stops the person dealing with the situation, and check it with whoever owns the statement. If the person, situation or consequence is missing, the problem statement needs more work before this exercise will help. Go back and fix it first.

Check it worked

Your notes show five labelled lines: person, situation, obstacle, consequence and assumptions. Each line is a phrase taken from, or directly implied by, the problem statement, and none of them names a feature, screen or technology.

Step 2 · Write questions quickly, one part at a time

Set a timer for 10 minutes. Go through the five lines in order (the four parts plus Assumptions, which counts as a line of its own here) and write at least two questions for each, so at least 10 in total. Do not judge them yet; bad questions are useful raw material for the next steps.

Some angles that help you get past the obvious first question:

  • Aim at the moment it goes wrong. "How might we let patients who cannot come tell the clinic before their appointment starts?"
  • Shrink or remove the obstacle. "How might we make cancelling take less effort than not turning up?"
  • Turn the consequence around. "How might we get a slot freed on the morning to someone who can use it?"
  • Challenge an assumption. "How might we make the no-show mark fair to people who tried to cancel?"
  • Look at someone else in the picture. "How might we let reception know early that a morning slot is about to go unused?"

Write every question in full, starting with "How might we". Keep each one to a single sentence. Number them as you go so you can refer to them later.

A Larkfield session might produce 15 to 20 questions, including some like these:

  1. How might we improve healthcare for working people?
  2. How might we let patients who cannot come tell the clinic before their appointment starts?
  3. How might we add more phone lines for the morning rush?
  4. How might we help patients find out sooner that their morning will not work?
  5. How might we get a slot freed on the morning to someone who can use it?
  6. How might we make the no-show mark fair to people who tried to cancel?
  7. How might we make the clinic less stressful?
  8. How might we add a cancel button to the Booked screen?

Check it worked

You have at least 10 numbered questions, every one starting "How might we", with at least two questions for each of the five lines: person, situation, obstacle, consequence and assumptions. If a line has fewer than two, spend two more minutes on it.

Step 3 · Pull hidden solutions out of the questions

Read each question and ask: does this name the answer? Words such as button, text, app, screen, notification, portal or a named feature are the usual signs.

When you find one, ask "What would that achieve?" and rewrite the question around the answer to that. The solution becomes one possible idea for the new question, not the question itself.

From the Larkfield list:

  • Question 3, "add more phone lines for the morning rush", achieves the clinic hears about a cancellation in time to do something with the slot. Rewrite: "How might we make sure the clinic hears about a cancellation while there is still time to use the slot?"
  • Question 8, "add a cancel button to the Booked screen", achieves a patient can tell the clinic quickly that they are not coming. That is already question 2, so mark 8 as a duplicate.

Keep the original solution ideas in a separate "Ideas parked" list. They are not wasted; they become candidates when you start generating ideas.

Check it worked

No question in your list names a feature, channel, screen or technology. Your "Ideas parked" list holds every solution you removed, and each rewritten question has its old number kept beside it so you can trace where it came from.

Step 4 · Test each question's scope with two prompts

Now judge scope. For each question, write down quick answers to two prompts:

  • "Name three ideas that would answer this, as different from each other as you can." If you can only think of one idea, or all your ideas are versions of the same thing, the question is too narrow.
  • "Would every one of those ideas help the person in the problem statement?" If the question also invites ideas that have nothing to do with the problem (redecorating the waiting room, a new website), it is too broad.

A workable question passes both: it invites at least three genuinely different ideas, and all of them would reduce the problem you wrote down.

Trying this on Larkfield:

  • Question 1, "improve healthcare for working people": ideas include longer opening hours, a crèche and cheaper prescriptions. Three different ideas, but most have nothing to do with morning cancellations. Too broad.
  • Question 2, "tell the clinic before their appointment starts": a cancel link in the reminder message, a reply-to-cancel text, a recorded line that only takes cancellations and never queues. All different, all aimed at the problem. Workable.
  • Question 3 rewritten, "the clinic hears about a cancellation while there is still time to use the slot": close to question 2. Workable, but mark it as a near-duplicate of 2.
  • Question 7, "make the clinic less stressful": ideas spread everywhere and few touch missed morning slots. Too broad.

Be honest about the three ideas. Write them down in a word or two beside the question, rather than deciding in your head that you could think of some.

Check it worked

Every question has a label (too broad, too narrow or workable) and two or three idea words beside it. You can point to the idea words that justify each label.

Step 5 · Move the broad and narrow questions to a workable level

Do not throw away questions that failed the scope test. Most of them are one move away from workable.

  • To widen a narrow question, ask "Why would we want that?" and write the answer as the new question. "How might we let patients cancel by replying to a message?" becomes, after asking why, "How might we let patients cancel without having to get through to reception?"
  • To narrow a broad question, ask "What part of this matters for our problem?" and anchor it to the person, situation or consequence. "How might we make the clinic less stressful?" becomes "How might we take the worry out of telling the clinic on the day that you cannot come?"

After each rewrite, run the two prompts from the previous step again. Some questions will need two moves. If a question still fails after two attempts, set it aside with a one-line reason, such as "too broad: about clinic experience in general, not morning cancellations".

Check it worked

Your list now has three groups: workable questions (original and rewritten), set-aside questions each with a one-line reason, and duplicates. Every rewritten question has passed the two prompts on its new wording.

Step 6 · Choose three questions to explore

You rarely have the time to explore every workable question. Pick three, using these tests in order:

  1. Evidence. Does the question point at something you have seen in research or support data, not something you suspect? For Larkfield, the statement already carries one figure, from the clinic's booking records: 1 in 7 booked appointments missed without notice. That gives questions about getting a freed slot used evidence behind them. A question about making the no-show mark fairer has none yet, until someone checks how many no-shows had tried to phone first.
  2. Spread. Do your three questions point in different directions? Three versions of "let them cancel more easily" leave the other parts of the problem unexplored. Aim for questions drawn from different parts of the statement.
  3. Reach. Could the team act on ideas from this question with the people and time you have? A question about changing how the whole clinic schedules may be right but out of reach for this project; note it for later.

For Larkfield, three questions that pass all three tests:

  • How might we help patients find out sooner that their morning will not work? (situation)
  • How might we let patients who cannot come tell the clinic before their appointment starts? (obstacle)
  • How might we get a slot freed on the morning to someone who can use it? (consequence; evidence: 1 in 7 booked appointments missed without notice)

Together they cover learning earlier, telling the clinic in time, and limiting the damage when a slot is still given up late. None of them assumes the answer is in the app, which keeps options such as a cancellation-only phone line or a reception process open.

Check it worked

You have exactly three chosen questions, each tagged with the part of the problem statement it came from, and no two tagged with the same part unless you can say in one sentence why that part matters most.

Step 7 · Write up the short list

Put the result on one page so the team can review it and refer back to it during idea sessions. Use this structure:

Text
Problem statement
<the sentence, unchanged>

Questions to explore
1. <question>  (part: <person/situation/obstacle/consequence>; evidence: <one line>)
2. ...
3. ...

Workable, not chosen this round
- <question>: <one-line reason, e.g. "out of reach this project">

Set aside
- <question>: too broad / too narrow, <one-line reason>

Ideas parked
- <solution ideas pulled out of questions>

Share it with whoever owns the problem statement and ask one thing: "Do these three questions cover the part of the problem you care most about?" If the answer is no, swap in a question from the "not chosen" list rather than writing a new one from scratch.

Check it worked

Your page has all five sections filled in. Someone who was not in the session can read it and tell you, without asking, why each chosen question was picked and why at least one other question was not.

Next steps

Your three questions are the starting points for idea generation. Before you start sketching answers, list what each question quietly assumes and decide which assumption to check first in Map your assumptions and choose which to test first.

If your problem statement felt shaky while you split it into parts, tighten it with Write a problem statement before you design. To see where this exercise sits in a whole project, from research to delivery, read Plan a project with the double diamond. If you have a journey map, Find opportunities in a journey map gives you another source of evidence for choosing between questions.

Need this done for you?

Padding does this work for growing companies, from the first diagnosis to a system that keeps working.

Digital Transformation

Get new tutorials

New tutorials, about twice a month.