Learning objective
By the end of this lesson, you can turn raw research observations into a focused design question without embedding a preferred solution.
Why it matters
Teams often treat the first description of a problem as objective:
Customers need a chatbot because support is slow.
This sentence already contains a solution and a causal claim. If the real difficulty is an unclear hand-off, a chatbot may add another channel without reducing effort.
A design frame should guide ideas while remaining open to different mechanisms.
Core concept: evidence, need, insight, question
Build the frame through four distinct moves:
- Evidence: what a person did, said or encountered.
- Need: the progress or capability that matters in that situation.
- Insight: the non-obvious explanation that makes the need important.
- Question: an open invitation to explore solutions.
Use:
How might we help [person] achieve [important progress] when [meaningful circumstance]?
Avoid product names, channels and features in the question.
Visual explanation
Text alternative: observations are interpreted into a need and contextual insight. Together they form an open design question that can produce several solution mechanisms.
Worked example
Research at a clinic finds:
- coordinators reschedule patients while on the phone;
- the interface confirms the new time but hides reminder status;
- coordinators write the change on paper until the patient replies;
- duplicate reminders are a frequent complaint.
Weak frame: How might we add a smarter rescheduling assistant?
Evidence-based frame:
- Need: know that every dependent update is complete before ending the call;
- Insight: the confirmation message verifies one record, but the coordinator is responsible for the whole chain;
- Question: How might we help coordinators see and recover the complete state of an urgent reschedule while the patient is still available?
This frame can produce interface, workflow, policy and communication options.
Common mistake
Mistake: making the question so broad that it cannot guide action.
Weak: “How might we improve healthcare?”
Better: “How might we help clinic coordinators confirm every consequence of an urgent reschedule before ending the patient call?”
Quick check
Which question is best framed?
A. How might we build an AI assistant for support?
B. How might we delight all customers?
C. How might we help account administrators understand who must act next when a support case moves between teams?
D. How might we redesign the ticket screen?
Answer: C. It identifies a person, progress and situation without fixing a solution.
Practical prompt
Take one proposed feature from your current work. Write:
- the observed behaviour or event behind it;
- the need revealed by that evidence;
- one contextual insight;
- a “How might we” question that contains no feature.
Summary
- A requested feature is not automatically the problem.
- Keep evidence, need and insight separate.
- A strong design question is specific about the situation and open about the solution.
Next lesson
Continue to Jobs to Be Done: Reconstruct the Decision to investigate why a person moves from a current solution to a new one.