Guide
How to write good "How might we" questions
A good "How might we" question turns one observed problem into an invitation for ideas. Start from a specific problem or insight, write it as "How might we..." on its own note, and check that it allows several solutions without naming one. Then group the notes into themes and vote on which to pursue. In AppSprint's Full Product Sprint, the Opportunities step collects anonymous notes, groups them into themes with AI and gives everyone private votes.
What a "How might we" question is
"How might we" (HMW) rewrites a problem as an opportunity. "Might" makes it safe to suggest something that may not work, and "we" makes it a shared job. The phrasing is older than the design sprint: it is generally traced to Procter & Gamble in the 1970s and was spread widely by the design firm IDEO.
In the book Sprint by Jake Knapp, John Zeratsky and Braden Kowitz, the team writes HMW notes while listening to expert interviews. The notes are grouped into themes and dot voted, and the winners help the Decider choose a target.
Getting the scope right
Weak HMW questions are too broad to act on, or so narrow that the answer is already inside them. Take one problem: new customers leave during setup, before they invite a teammate.
| Version | Question | Verdict |
|---|---|---|
| Too broad | How might we improve onboarding? | Any idea fits, so it guides nobody |
| Too narrow | How might we add an invite button to step two? | A solution with a question mark |
| About right | How might we make inviting a teammate feel worth doing in the first five minutes? | One clear problem, many possible answers |
Write and sort them step by step
Gather raw material (before the session)
Bring the problem statement, customer complaints, interview notes or an expert to question.
Write alone, one question per note (10 to 15 minutes)
Whenever you hear a problem or an interesting fact, turn it into a question beginning "How might we". Aim for quantity.
Check each note (5 minutes)
Does it come from a real problem? Could you answer it in at least three ways? Does it avoid naming a feature? Rewrite the ones that fail.
Group into themes (10 minutes)
Post all notes without names, put similar ones together and label each group. A miscellaneous group is fine.
Vote (5 minutes)
Give each person a few votes to place privately, then reveal the counts together. See anonymous dot voting.
Choose the focus (5 minutes)
The Decider picks the questions to work on, using the vote as advice.
Common mistakes
- Hiding a solution in the question: "How might we build a chatbot for support?"
- Writing them out loud as a group, which produces one person's questions.
- Polishing the wording. A rough question on the right problem beats a polished one on the wrong problem.
Running it remotely
Online, the risks are visible authorship and slow sorting. Collect notes in a way that hides who wrote what until all are posted. Let one person do a rough first grouping, then correct it together.
Doing this in AppSprint
The step is called Opportunities, and it exists only in the Full Product Sprint. Each person writes up to ten "How might we" notes. Notes stay private until the host shows them, then appear shuffled with no author shown to anyone. See anonymous ideas and voting.
AI groups the notes into up to six themes, which the team can edit together. Each person then casts one private vote for a theme and eight for individual notes. Counts appear only after the reveal, and the vote is advisory: the decision-maker, AppSprint's term for the Decider, confirms the choice. On this step the host can also invite an expert guest by link. The expert joins the video call without an account and sees none of the room's content.
The Focused Product Sprint and One-Hour Sprint have no HMW step. For HMW notes on a free-form canvas, use a whiteboard.