Problem Framing Skill

Transform market signals into evidence-anchored problem statements.

Consumes

This skill assumes you have zero prior research. It is the starting point.

• No prior CFD- entries needed
• No prior PR D required
• Assumes: Founder/PM has observed market signals but hasn't validated them

Produces

This skill creates/updates:

• CFD-* entries (customer feedback) — 1-5 per problem dimension, with confidence scoring (see PRINCIPLES.md)
• PRD.md Why section — Evidence-anchored problem statement table
• MVP scope signal — Identifies which problem dimensions will drive MVP feature scope (handed to v0.3)

All CFD- entries should include:

• confidence: 1-3/5 (pre-product research, no usage data)
• Evidence source (competitive analysis, interviews, workarounds, etc.)
• Forward target: "Would move to 4/5 if we observe 10+ paying customers with this pain"

Workflow Overview

1. Assess gaps → Identify what evidence is missing before you can confidently state a problem
2. Anchor evidence → Create CFD- entries for each pain point dimension with confidence scoring
3. Extract dimensions → Pull multiple distinct problems from each source
4. Quantify costs → Add time/money/risk numbers to make pain concrete
5. Draft statement → Populate the problem table, tied to CFD- entries

Core Output Template

Populate this table for every problem statement:

Element	Definition	Evidence
Who is hurting?	Specific, findable, countable persona	Segment size
What pain exists?	Observable behavior or workflow friction	CFD-ID
Cost of problem	Time, money, or opportunity lost	Quantified
Why now?	Market trigger creating urgency	Trend/event
What's impossible?	Opportunity cost—what can't they do	User quote

See assets/problem-statement.md for copy-paste template.

Step 1: Gap Assessment

Before drafting, create this status table:

Element	Status	Source
Who is hurting?	⚠️ Hypothesis / ✅ Validated / ❌ Missing
What pain exists?	⚠️ / ✅ / ❌
Cost of problem	⚠️ / ✅ / ❌
Why now?	⚠️ / ✅ / ❌
What's impossible?	⚠️ / ✅ / ❌

Gate: Require ≥2 elements ✅ Validated before drafting. If ≥3 elements ❌ Missing, run deep research first. See references/research-prompts.md for research templates.

Step 2: Evidence Anchoring

Create CFD entries for each pain point with confidence scoring:

CFD-###: [Pain Point Name]
Source: [Where this evidence came from]
Tier: [1-5 evidence quality]
Confidence: [1-5]/5 (pre-product research)
Quote: "[Verbatim from source]"
Dimensions: [List distinct problems extracted from this source]
Next Target: "Would move to 3/5 if we interview X more customers"

Evidence Tier Hierarchy (strength of observation):

• Tier 1: Buying behavior (invoices, subscriptions, job budgets) — users spend money to solve this
• Tier 2: Active workarounds (spreadsheets, hired help, manual processes) — users invest labor
• Tier 3: Complaints with cost ("costs me X hours/week") — users quantify the pain
• Tier 4: General complaints ("this is annoying") — users acknowledge it but haven't quantified
• Tier 5: Speculation — REJECT ("users probably want...")

Confidence Scoring (pre-product, see PRINCIPLES.md):

• 1/5: PM assumption or single data point
• 2/5: Secondary research (competitive analysis, market reports)
• 3/5: Pre-product interviews (3-5 user conversations)
• 4/5: Beta cohort validation (observed behavior, not questions)
• 5/5: Production usage (reserved for post-launch)

Example entry with confidence:

CFD-001: Sales teams waste 5+ hours/week on spreadsheet workflow

Source: 3 customer interviews (SaaS sales director, SMB sales rep, enterprise sales manager)
Tier: 2-3 (workaround + cost quantification)
Confidence: 3/5 (source: 3-customer-interviews-jan-2026)
Quote: "I spend 5 hours every Friday reconciling our pipeline with the actual numbers in our CRM"
Dimensions:
  - Manual data reconciliation between systems (workaround)
  - Inventory work (scheduling impact)
  - Single source of truth fragmentation (data quality risk)
Next Target: "Would move to 4/5 if we validate with 5 more sales leaders or observe workflows directly"

Step 3: Pain Dimension Extraction

Extract multiple problems from each source. One quote often contains 3-4 distinct pain dimensions.

Example: "USB sticks removed for every update, no scheduling, screens don't communicate, priced for 100+ displays" → Sneakernet workflow, No dynamic scheduling, No centralization, Price mismatch

Step 4: Cost Quantification

Every problem needs a number:

Type	Calculation
Time	Hours/week × hourly rate
Money	Current spend on workaround
Opportunity	Revenue/outcomes missed
Risk	Penalty × probability

Step 5: Draft Problem Statement

Use the core output template. Reference CFD-IDs for every claim.

See references/examples.md for good/bad examples with explanations.

Quality Gates

Pass Checklist

• ≥1 Tier 1-2 evidence item
• Cost quantified (time, money, or risk)
• "Who" specific enough to build prospect list
• "Why now" has at least Tier 3 hypothesis

Testability Check

• Can find 10 people with this problem in 48 hours?
• Can observe the pain behavior?
• Can quantify cost without leading questions?

Anti-Patterns

Pattern	Example	Fix
Vague "Who"	"Small businesses"	→ "SMBs with 1-10 screens"
Feature-as-problem	"Need a dashboard"	→ "Can't see status"
Solution creep	"MVP must solve X"	→ Stay on problem (v0.4)
Missing cost	"This is annoying"	→ "Costs X hrs/week"
Speculation	"Users might want"	→ Find evidence or reject

Bundled Resources

• references/research-prompts.md — Deep research templates by gap type. Use when gap assessment shows ≥3 missing elements.
• references/examples.md — Good/bad problem statement examples with explanations.
• assets/problem-statement.md — Copy-paste template for problem tables and CFD entries.

Handoff

Problem statement complete when quality gates pass. Next: v0.2 Market Definition (segments, sizing, ICP).