# How Might We questions: framing the problem before ideating

> How Might We (HMW) questions turn a research finding into an open problem to ideate on. Learn how to write them, set their scope, and run the session.

*Tags: ux, technique, ideation, workshop, problems*

---


> [!info] Quick Definition
> **How Might We** (HMW) questions turn a research finding into an open problem that invites ideas. They bridge understanding the problem and generating solutions, and their format helps the team avoid jumping straight to the first idea.


## What are HMW questions?

They are questions that start with "How might we…?" and frame a problem found during discovery. Nielsen Norman Group describes them as a structure that frames a problem, not a solution. The question does not say what to build; it says what needs solving.

According to Nielsen Norman Group, the template was introduced by Procter & Gamble in the 1970s and later adopted by IDEO, and it spread from there with design thinking.

## Why does it matter to a designer?

Because the most common temptation in a design team is to jump to a solution: "let's build a new menu". An HMW question forces a pause to ask what you are trying to achieve. That reduces the risk of designing beautiful screens for the wrong problem.

It also organizes conversations. When every finding is turned into a question in the same format, it is easier to compare, prioritize, and decide what to explore first.

## How do you write a good question?

Koji explains the form word by word: "How" assumes a solution exists, "Might" leaves open the judgment on whether it will work, and "We" sets a collaborative tone. In practice, a simple form is "How might we" + an action + the desired outcome for a specific person.

Some practical rules:

- **Start from a concrete finding**, not an opinion. A question comes from a problem you observed in research.
- **Write it positively** and without building in the solution. "How might we reduce waiting time at payment?" works; "How might we add a one-click payment button?" has already decided the answer.
- **Aim at a desired outcome**, something the user or the business wants to achieve.

## How do you set the scope?

Scope is the hardest part. A question that is too broad gives infinite directions; one that is too narrow already contains the solution. Koji gives an example of each extreme:

| Too narrow | Too broad |
|---|---|
| "How might we raise trial conversion by 5%?" | "How might we delight customers?" |

Nielsen Norman Group offers a similar criterion: the scope should be broad, but not so broad that the problem gets lost. Its example of a question that is too broad is "How might we redesign the submission-drafting process?". And "How might we stop users from calling us?" describes a symptom: the calls are a consequence of people being unsure how to complete the process, and that uncertainty is the cause worth addressing.

A test that Koji attributes to Tim Brown is useful: a good question should not be answerable in five minutes, nor take five years.

## How do you facilitate the session?

Nielsen Norman Group describes a process you can apply as is:

1. **At the end of discovery**, the team agrees on the main findings.
2. **Each person proposes HMW questions** based on those findings.
3. **The questions are evaluated** with a checklist: does it start from a problem or insight, does it point to a desired outcome, is it written positively, is it broad enough, and does it suggest a solution?
4. **Overlapping questions are merged** and prioritized, for example by impact.

Koji suggests writing three to five questions per point of view, choosing one, and spending 30 to 60 minutes ideating on it, aiming for more than a hundred ideas before filtering.

### How do you generate variations of the same question?

If the first version comes out too narrow or too broad, rewrite it instead of accepting it. Stanford d.school suggests three moves that start from what you already know about the challenge: amp up the good, explore the opposite, and take it to an extreme. Koji extends the list to seven patterns documented by IDEO and the d.school, including removing the bad, questioning an assumption, identifying unexpected resources, creating an analogy, and playing with scale.

Illustrative example, using the taxes case below: "explore the opposite" leads to asking how the final price could be the first thing people see; "play with scale" leads to asking what the ideal price display would be for someone buying a single flight versus someone planning a five-city trip. Neither dictates the solution, and each opens a different family of ideas.

## When should you NOT use them?

If you have already decided on a solution and only need to execute it, HMW questions add a step with no value. They also do not work without research: without findings, the questions become opinions with question marks.

## Example

Illustrative example: in interviews for a travel app, several users say they stop searching for flights when they see the final price, because taxes appear at the last step. A first attempt at a question would be "How might we show taxes at the start?", which already contains the solution. The team reframes it as "How might we help people compare real prices from the first search, so they trust the result?". The new question opens options: showing the total price, explaining charges, filtering by final price, or comparing other dates.

## Common mistakes

- **Writing the solution inside the question**, which closes the conversation.
- **Asking about symptoms** instead of causes.
- **Filtering by feasibility while writing**, which reduces the ideas.
- **Writing questions without findings** to support them.
- **Keeping the first question** without trying variations in scope and angle.
- **Turning it into an exercise with no consequences** that nobody uses afterward.

To reach the findings that feed these questions, [desk research](/en/wiki/research/desk-research/) and [user interviews](/en/wiki/research/user-interviews/) are the starting points. Afterward, [[scamper|SCAMPER]] helps generate ideas from each question, and [[crazy-8s|Crazy 8s]] speeds up sketching. And to validate the most promising idea before building it, [[concept-testing|concept testing]] is the next step.

## Sources

- [Using "How Might We" Questions to Ideate on the Right Problems, Nielsen Norman Group](https://www.nngroup.com/articles/how-might-we-questions/)
- [How Might We Questions, Stanford d.school](https://dschool.stanford.edu/tools/how-might-we-questions)
- [How Might We Questions: The Complete Framework for Turning Insights Into Innovation Opportunities, Koji](https://www.koji.so/docs/how-might-we-questions)


---

Source: https://www.fernandoux.com/en/wiki/techniques/how-might-we/
