How might we · Ideation · Design thinking
How to write how might we statements, with examples
By T. Krause
A how might we question that already contains the answer won't get you many ideas. Here is how to write one that works, with three made-up clinic problems rewritten, checked and tried out on a brainstorm.
To write a how might we statement, take one problem you found in research and rewrite it as a question that begins "How might we". Name the outcome you want and leave the solution out. Then check it's wide enough to allow many different answers and narrow enough that it still names that one problem.
"Statement" and "question" mean the same thing here, and it's written as a question. Nielsen Norman Group, or NN/g, says the format keeps a team on the right problem and stops people pushing "their pet solutions". IDEO.org's Design Kit adds a rule that's easy to break: "A properly framed How Might We doesn't suggest a particular solution".
We make Prompt Board, and our how might we template opens on the three problems below. A wall and a marker work too.
Three clinic problems, rewritten
Everything in this section is invented: the clinic, its problems, its team and the ideas. There's no such clinic, and nobody did this research.
The staff of a small medical practice have listened to patients and read their complaints. They wrote down three problems, and a first question for each. These six notes are on our sample board.
| Problem | First question |
|---|---|
| Patients forget appointments | …remind patients the way they prefer? |
| Phone line busy at 8am | …let people book without calling? |
| Forms filled twice | …ask for details only once? |
These are good first drafts. Each one comes from a problem on the board, which is NN/g's first tip. Now the team tests them.
"Remind patients the way they prefer"
This question has already chosen reminders. A reminder might be the right idea, but the question rules out a booking closer to the date, a shared calendar entry, or a follow-up call about why someone might not come. The outcome is that patients turn up, so that's what the rewrite names.
How might we help patients keep the appointments they book?
"Let people book without calling"
This is closer, since it leaves room for several ways to book. It's written from the clinic's side, though, and NN/g warns about the same pattern. Its poor example is "How might we stop users from calling us?", because it treats calls as the problem and misses why people call. These patients call at 8am because that's when they can get an appointment.
How might we get people an appointment on their first try?
That could still mean online booking. It could also mean more phone staff at 8am, or holding back slots for later in the day.
"Ask for details only once"
This one needs no change. It names one problem and an outcome, and there are several ways to do it, such as shared records, a pre-filled form or fewer questions. A question can pass without a rewrite, so don't change one just to look thorough.
Five checks on every question
NN/g's article gives a checklist. These are its questions, applied to the clinic.
Is it based on an existing problem or insight? "…keep the appointments they book?" comes from a note on the board. "…improve the patient experience?" could be written without any research.
Does it track a desired outcome? "…an appointment on their first try?" names what the patient wants. "…reduce phone calls?" names what the clinic wants.
Is it written positively? "…keep the appointments?" passes. "…stop patients missing them?" fails.
Is it broad enough to ensure many creative ideas? "…help patients keep appointments?" allows many kinds of answer. "…text patients a day before?" allows one.
Does it suggest a solution? "…ask for details only once?" doesn't say how. "…build a booking app?" does.
NN/g's advice on positive wording prefers verbs like "increase" and "create" over "reduce" and "prevent". We apply this loosely. If the negative form is much clearer, keep it, since a clear question matters more.
How wide is wide enough
Design Kit says to broaden a question if it doesn't allow a variety of solutions, and that it shouldn't be so broad that you can't tell where to begin. The same problem at three widths shows what that means.
| Width | Question |
|---|---|
| Too narrow | How might we text patients a day before their visit? |
| About right | How might we help patients keep the appointments they book? |
| Too wide | How might we improve healthcare in our area? |
Our own test takes two minutes. Write three different answers. If all three are the same kind of idea, the question is too narrow. If you can't tell where to start, it's too wide.
For the question in the middle row, one person writes "a reminder by text", "offer a slot nearer the date" and "ask what would stop them coming". Those are three different kinds of idea, so it passes.
From a list to one brainstorm
NN/g says "there's no limit on how many HMW questions you should produce". If you have too many, combine the overlaps or choose by likely impact on the project.
The clinic team gives each person two dots and puts one question in Top questions. On the sample board, that's the booking question, in its first form. After the rewrite, it reads "How might we get people an appointment on their first try?" They copy it to the top of a brainstorming board, and now everyone's writing ideas about the same thing.
If your questions came out of a research session, an affinity map is a good way to reach them. Each named group can give you one question or more.
Why the three words work
Tim Brown of IDEO explained the three words in Warren Berger's 2012 article for Harvard Business Review. "How" assumes there are solutions, "might" says ideas can work or fail, and "we" says you'll build on each other's ideas. That's his view of how the wording affects a team, and the article doesn't offer a study behind it.
Where the phrase came from
NN/g says the template "was first introduced by Procter & Gamble in the 1970s and adopted by IDEO". Berger's article traces it to Min Basadur "and his early days as a creative manager at Procter & Gamble", and it describes how the phrase later reached IDEO through a designer who had worked with him. We read the article's text in a copy reposted elsewhere, since HBR's own page didn't show us all of it. We haven't seen a document from P&G at the time, so this rests on Berger's account and NN/g's summary.
Frequently asked questions
What is a how might we statement?
It's a problem rewritten as an open question that begins "How might we", used to start ideas. For example, "Patients forget appointments" could become "How might we help patients keep the appointments they book?"
When do you write how might we questions?
After research and before a brainstorm. Design Kit places them after you've defined themes and insights, and NN/g says to write them at the end of discovery.
How many should you write?
As many as your findings suggest. NN/g sets no limit. Then combine the overlaps and pick one or a few to brainstorm on.
What is a bad how might we question?
One that contains its answer, such as "How might we build a booking app?" Another is one with no problem behind it, such as NN/g's example "How might we improve the user experience of the product?"
Can you use how might we outside design?
Yes. Any team with a named problem can use it. The clinic in our example could as easily be a sales team or a school office.
Write your own
Put your findings in the first frame, one per note, and rewrite each as a question. Open the how might we template, replace the clinic notes, run the five checks and vote. Then take the winning question to a brainstorming session.