problem-definition（问题定义）

Problem Definition

Scope

Covers

• Turning a vague idea into a crisp, testable problem definition
• Writing a shareable problem statement (1-liner + expanded)
• Capturing Jobs To Be Done (JTBD) and target segments
• Mapping current alternatives (including non-digital/analog) and “why now / why digital”
• Building an evidence + assumptions log to drive learning
• Defining success metrics + guardrails and clear scope boundaries

When to use

• “Write a problem statement for…”
• “We need to define the problem space / JTBD.”
• “We keep jumping to solutions; help us get clear on the real problem.”
• “Pressure to ‘do AI’ — verify there’s a real pain point first.”
• “Before we write a PRD, align on what problem we’re solving.”

When NOT to use

• You already have an approved problem definition and need a delivery-ready PRD (use writing-prds)
• You need roadmap prioritization across many competing initiatives (use prioritizing-roadmap)
• You need to set company-level strategy/vision (use defining-product-vision)
• You need a competitive landscape analysis as the primary output (use competitive-analysis); this skill references alternatives only to frame the problem
• You’re doing deep research execution (recruiting, interviews, analysis); use this to frame *what to learn*, not as a substitute for conducting-user-interviews or designing-surveys
• You need to analyze existing user feedback data (use analyzing-user-feedback); this skill frames the problem, not the evidence pipeline

Inputs

Minimum required

• Product/context + target user (or segment hypotheses)
• The triggering signal (customer quotes, data trend, stakeholder request, competitor move)
• The decision to make (e.g., invest now vs later; explore vs stop) + timeline
• Known constraints (tech/legal/privacy/compliance/capacity)

Missing-info strategy

• Ask up to 5 questions from references/INTAKE.md.
• If still missing, proceed with clearly labeled assumptions and list Open questions that would change the decision.

Outputs (deliverables)

Produce a Problem Definition Pack in Markdown (in-chat; or as files if the user requests):

1. Context snapshot (product, user, trigger, decision, constraints)
2. Problem statement (1-liner + expanded) + why now
3. JTBD (primary job + key sub-jobs) + target segment notes
4. Current alternatives (including analog/non-digital) + gaps + switching costs
5. Evidence & assumptions log (what we know vs what we’re guessing)
6. Success criteria (outcome metric(s), leading indicators) + guardrails
7. Scope boundaries (in/out, non-goals, dependencies)
8. Prototype / learning plan (fast prototype + tests to de-risk)
9. Risks / Open questions / Next steps (always included)

Templates: references/TEMPLATES.md Expanded heuristics: references/WORKFLOW.md

Workflow (8 steps)

1) Intake + decision framing

• Inputs: User context; references/INTAKE.md.
• Actions: Clarify the decision, time horizon, stakeholders, and constraints. Capture the trigger signal (data/quotes/event).
• Outputs: Context snapshot.
• Checks: You can state the decision in one sentence (“We are deciding whether to… by ”).

2) Define the target user + situation (segment + context)

• Inputs: Context snapshot.
• Actions: Specify who experiences the problem, when it happens, frequency, and what’s at stake. If multiple segments, pick a primary and list others as secondary.
• Outputs: Target user + context bullets.
• Checks: The segment is specific enough that a researcher could recruit for it.

3) Write the problem statement (1-liner + expanded)

• Inputs: Target user + trigger signal.
• Actions: Draft a crisp 1-liner, then expand with symptoms, root causes (hypotheses), and impact. Include why now.
• Outputs: Problem statement section (using references/TEMPLATES.md).
• Checks: Statement describes the problem without implying a specific solution or technology.

4) Map current alternatives (including non-digital) + “why use this”

• Inputs: Problem statement.
• Actions: List how users solve this today (manual workarounds, spreadsheets, incumbents, doing nothing). Include at least one analog/non-digital alternative when relevant.
• Outputs: Alternatives table + gaps + switching costs.
• Checks: You can answer: “Why would a user give this the time of day vs their current way?”

5) Separate problem from solution (avoid the shiny object trap)

• Inputs: Alternatives + early solution ideas (if any).
• Actions: Capture solution ideas as hypotheses, not commitments. If “AI” (or any tech) is proposed, state the user pain point first and treat tech choice as an implementation detail.
• Outputs: Evidence & assumptions log (with test ideas).
• Checks: Each assumption has a proposed test and a confidence level.

6) Define success criteria + guardrails

• Inputs: Problem statement + evidence.
• Actions: Define measurable outcomes, leading indicators, and guardrails (quality, trust, cost, latency, support load, etc.).
• Outputs: Success metrics + guardrails section.
• Checks: Metrics are unambiguous and tied to the user’s desired outcome.

7) Visualize the end state + prototype a path to clarity

• Inputs: Success criteria + scope constraints.
• Actions: Describe what “done” looks like (user-visible end state). Create a fast prototype/experiment plan to validate the hardest assumptions before building.
• Outputs: End-state description + prototype/learning plan.
• Checks: The team can “see the end” and name the 1–3 biggest unknowns being tested.

8) Quality gate + finalize the pack

• Inputs: Full draft pack.
• Actions: Run references/CHECKLISTS.md and score with references/RUBRIC.md. Add Risks/Open questions/Next steps.
• Outputs: Final Problem Definition Pack.
• Checks: A stakeholder can review async and decide “proceed / pause / stop” without a meeting.

Quality gate (required)

• Use references/CHECKLISTS.md and references/RUBRIC.md.
• Always include: Risks, Open questions, Next steps.

Anti-patterns (common failure modes)

1. Solution-first framing — The problem statement embeds a specific technology or feature (“We need an AI chatbot for support”) instead of describing the user pain. Strip the solution; restate as a user struggle with evidence.
2. Segment hand-waving — “Our users” without specifying who, when, and how often. A problem definition that applies to everyone applies to no one. Require a recruitable segment description.
3. Evidence-free conviction — The pack lists assumptions but marks them all “high confidence” without citing any data, quotes, or signals. Insist on at least one concrete evidence artifact per top assumption.
4. Metric theater — Success metrics exist but are unmeasurable within the stated timeline or are vanity metrics (e.g., “user happiness”). Every metric needs a measurement method and a realistic baseline.
5. Scope creep via JTBD — Sub-jobs multiply until the problem definition covers the entire product. Cap sub-jobs at 3-5 and explicitly mark anything beyond as out-of-scope.

Examples

Example 1 (B2B SaaS): “Define the problem for improving onboarding activation in our analytics product.” Expected: a pack with a tight segment, current onboarding alternatives/workarounds, measurable activation outcomes, and a prototype plan to test the most uncertain hypothesis.

Example 2 (Consumer): “Users abandon checkout on mobile; define the problem space and JTBD before proposing fixes.” Expected: a problem statement grounded in evidence, an alternatives map (including ‘do nothing’), and guardrails (fraud/chargebacks/support load).

Boundary example (redirect to writing-prds): “Write a PRD for building an AI assistant; we don’t know what problem it solves.” Response: push back; run this skill to define the user pain point and success metrics first, then hand off to writing-prds.

Boundary example (redirect to conducting-user-interviews): “We think onboarding is broken but have zero user data. Define the problem.” Response: flag that the problem definition will be assumption-heavy; recommend running conducting-user-interviews first to gather evidence, then return here to frame the problem with real data.