Identify and Test Assumptions (Continuous Discovery Habits)

Goal

To extract explicit assumptions from insights and opportunities, categorize and prioritize them to identify "leap-of-faith" assumptions, and design a lightweight, iteratively-scaled testing plan that reduces risk across desirability, usability, feasibility, viability, and ethical dimensions.

When to Use

After creating opportunities using Create Opportunities
After synthesizing interviews using Synthesize Interview Snapshots
When preparing to generate or downselect solutions with Generate Solutions
Whenever a new idea is proposed and you need to surface and derisk its underlying assumptions

Input

Primary Sources:
- Prioritized opportunities from opportunities/
- Early solution sketches from solutions/
- Interview snapshots from user-interviews/snapshots/
- Synthesis documents from user-interviews/synthesis/
Optional Sources:
- Product analytics or behavioral data
Minimum Requirements:
- 1 target opportunity with supporting evidence, and
- 2–3 candidate solution ideas OR a single idea with key user journeys

Output

Format: Markdown (.md) Location: assumptions/[topic]/ Filename: assumptions-[opportunity-name]-v[version].md

Semantic Naming Guidelines:

opportunity-name: kebab-case opportunity name from opportunity document (e.g., newsletter-creation, user-onboarding)
version: auto-incrementing version number (v1, v2, v3...)
Example: assumptions-newsletter-creation-v1.md

Version Management:

Check existing files with same opportunity-name pattern before creating new assumptions
Auto-increment version number (v1 → v2 → v3...)
Never overwrite existing assumption files
Preserve all assumption versions for comparison
No date dependency required

AI Instructions for Assumption Identification

When Receiving Research and Opportunity Data

Validate Inputs: Confirm target opportunity, evidence strength, and affected segments.
Clarify Scope: Define which ideas/journeys will be story-mapped.
Note Constraints: Technical, legal, GTM, success metrics, time/resource limits.
Topic Extraction: Analyze opportunity content to extract main topic for semantic filename.
File Existence Check: MANDATORY - Check for existing files with pattern assumptions-[topic]-v*.md before creating new file.
Version Management: MANDATORY - Auto-increment version number based on existing files. Never overwrite existing files.

Assumption Generation Guidelines

Use Five Categories: Desirability, Usability, Feasibility, Viability, Ethical.
Phrase Positively and Specifically: State what must be true, in concrete, testable language.
Tie to Behavior: Prefer assumptions about what users will do over what they say.
Attach Evidence: Link each assumption to quotes, behaviors, or data when available.
Normalize Granularity: Split vague, compound assumptions into specific, testable statements.

Evidence Classification (Binary)

Strong Evidence:
- Direct user quotes or behaviors supporting assumption
- Quantitative data from multiple sources
- Previous successful tests of similar assumptions
Weak Evidence:
- No direct evidence
- Single source or anecdotal evidence
- Theoretical or assumed knowledge only

Importance Evaluation (Binary)

More Important:
- Core value proposition depends on this assumption
- Assumption failure would kill the solution
- Blocking other critical assumptions
Less Important:
- Nice-to-have features or optimizations
- Minor impact on solution success
- Non-blocking assumptions

Prioritization Guidelines (Assumption Mapping)

Place each assumption on a 2D grid:
- X-axis: Evidence known (left = strong evidence, right = weak evidence)
- Y-axis: Importance to idea success (bottom = less important, top = more important)
Leap of Faith (LoFA): Select ONLY assumptions in the top-right quadrant:
- Weak Evidence (right side of X-axis)
- More Important (top half of Y-axis)
- Maximum 3 LoFA per assumption document
- Visual indicator: Mark with circle or highlight

Testing Guidelines

Simulate an Experience, Evaluate Behavior: Design minimal simulations that let users behave in line with or against the assumption.
Define Success Upfront: Use numbers (e.g., "≥ 3 of 10 participants choose X"), not percentages.
Recruit the Right Audience: Screen by target opportunity and segment; select for variation.
Start Small, Then Scale: Begin with quick signals; escalate to larger tests only if warranted.
Triangulate: Combine small, different methods to reduce false positives/negatives.

Semantic File Naming Guidelines

Topic Extraction: Analyze opportunity content to identify main topic/theme
Filename Format: Use semantic naming pattern assumptions-[opportunity-name]-v[version].md
Version Management: Auto-increment version number based on existing files
Folder Organization: Create topic-specific subfolders for better organization
No Date Dependency: Remove all date-based filename requirements

Topic Extraction Process:

Analyze opportunity document for main theme and keywords
Identify the most relevant topic/theme from opportunity content
Convert to kebab-case format (e.g., "Newsletter Creation" → "newsletter-creation")
Ensure topic uniqueness across different assumption files
Use topic as primary identifier instead of date

Version Management Process:

MANDATORY STEP 1: Check existing files with pattern assumptions-[opportunity-name]-v*.md
MANDATORY STEP 2: Find the highest version number for the same opportunity
MANDATORY STEP 3: Auto-increment version number (v1 → v2 → v3...)
MANDATORY STEP 4: Generate new filename with incremented version
MANDATORY STEP 5: Verify no file with new filename exists before creation
CRITICAL: Never overwrite existing files - always create new version
No manual version tracking required

Smart Topic Detection:

Content Analysis: Extract main themes from opportunity document
Keyword Frequency: Use most frequent relevant keywords as topic
Kebab-Case Format: Convert spaces to hyphens, lowercase (e.g., "User Onboarding" → "user-onboarding")
Uniqueness Check: Ensure topic doesn't conflict with existing files
Fallback: Use generic topic name if extraction fails

Process

0) File Management (MANDATORY PRE-STEP)

Extract topic from opportunity document content
Check existing files with pattern assumptions-[topic]-v*.md in target directory
Find highest version number for same topic (e.g., if v1, v2 exist, next is v3)
Generate new filename with incremented version: assumptions-[topic]-v[version].md
Verify no file exists with new filename before proceeding
CRITICAL: Never overwrite existing files - always create new version

1) Prepare Context and Actors

Confirm target opportunity and desired outcome(s).
Identify key actors (end-user types, internal systems, third parties).

2) Story Map Candidate Ideas (or Key Journeys)

Assume the solution exists; map what users do to get value.
Sequence steps by actor; highlight moments critical to success.

3) Generate Assumptions (Five Categories)

For each pivotal step, enumerate assumptions across: Desirability, Usability, Feasibility, Viability, Ethical.

4) Pre-Mortem (Prospective Hindsight)

"It's six months later; launch failed. What went wrong?" Convert reasons into specific assumptions that must be true.

5) Walk OST Lines (Outcome ↔ Opportunity ↔ Solution)

Write why the solution addresses the opportunity and drives the outcome. Extract each inference as a testable assumption (esp. viability).

6) Normalize, Deduplicate, and Attach Evidence

Rewrite assumptions to be positive, specific, and single-concept. Link supporting quotes, behaviors, analytics.

7) Map and Prioritize (Assumption Mapping)

Plot all assumptions on 2D grid using binary classification
Identify top-right quadrant (Weak Evidence + More Important)
Select maximum 3 assumptions from this quadrant
If more than 3 in quadrant: Prioritize by impact, test complexity, dependencies
Mark selected LoFA with visual indicator

8) Define Test Cards for LoFA Assumptions

For each of the 3 selected LoFA, design the smallest simulation with clear success criteria, sample size, method, audience, and time window.

9) Run Tests → Record Results → Update the Map

Move assumptions leftward as evidence grows; iterate simulation quality or move to next riskiest item.

10) Decide and Proceed

Use accumulated evidence to: evolve the idea, change the opportunity focus, or scale the solution test.

Output Structure (assumptions-[opportunity-name]-v[version].md)

# Assumptions — [Opportunity Name]

**Topic:** [Extracted topic name]  
**Version:** [v1, v2, v3...]  
**Target Opportunity:** [Opportunity statement]  
**Related Documents:** [Snapshots/Synthesis/Opportunities/Solutions]

---

## Story Map Snapshot
- **Actors:** [End-user types, systems, partners]
- **Key Steps:**
  1. [Actor] — [Step]
  2. [Actor] — [Step]

---

## Assumption Log
| ID | Category | Assumption (positive, specific) | Evidence (link/quote/data) | Importance | Evidence Known | LoFA |
|----|----------|----------------------------------|-----------------------------|------------|----------------|------|
| A-01 | Desirability | [What must be true] | [Quote/analytics/ref] | More/Less Important | Strong/Weak | Yes/No |

---

## Assumption Map (Summary)
- **Top-right (LoFA):** [A-01, A-07, A-12] - Maximum 3 assumptions
- **Notable clusters:** [e.g., viability assumptions lacking data]
- **Visual Grid:** Use 2D grid with binary classification (Strong/Weak Evidence, More/Less Important)

---

## Test Cards (LoFA)

### Test Card: [A-01] — [Short name]
- **Assumption:** [Assumption statement]
- **Simulation:** [Prototype/mock experience/data query/concept test]
- **Method:** [Unmoderated test | 1-question survey | Customer letter technique|  Data Analysis | Concierge Test | Wizard of Oz | Usability Test | Live-data prototype | Fake door test | Landing Page Demand Test | Ealry Adopters | Longitudinal User Study | Qualitative Value Testing | Dogfood | Fishfood | Smoke Tests]
- **Audience:** [Screening criteria; segment]
- **Sample Size & Window:** [e.g., n=10 over 2 days]
- **Success Criteria:** [e.g., ≥ 3/10 do X]
- **Risks & Biases:** [Key concerns and mitigations]
- **Next Step if Pass/Fail:** [Scale test / iterate assumption / Experiment(e.g., Multivariate, A/B tests) / pivot idea]

*(Repeat per LoFA assumption)*

---

## Results and Decisions
- **Outcomes:** [Observed behaviors vs. criteria]
- **Map Update:** [Assumptions moved left; new LoFA]
- **Decisions:** [Proceed/iterate/stop; changes to opportunity or idea]

---

## Next Steps
- [ ] Run next LoFA test
- [ ] Evolve idea based on findings
- [ ] Share summary with stakeholders

Templates

Assumption Statement Pattern

Actor + Action + Context + Outcome expected
Example: "Prospective subscribers will select a live game from our home screen when browsing evening entertainment options."

Pre-Mortem Prompt

"It's six months after launch and this failed. What happened?"
Capture each reason → rewrite as a positive, specific assumption that must be true.

Story Map (Quick)

Actors: [User, Platform, Partner]
1) [User] does …
2) [Platform] shows …
3) [Partner] provides …

Assumption Mapping Template

                    Evidence Known
                Strong              Weak
More     [A-01]  [A-03]    [A-02]  [A-04] ← LoFA (3개)
Important [A-09]  [A-10]    [A-05]  [A-06] ← LoFA (3개)

Less     [A-15]  [A-14]    [A-18]  [A-17]
Important [A-12]  [A-11]    [A-13]  [A-16]

LoFA Selection Process

Plot all assumptions on 2D grid using binary classification
Identify top-right quadrant (Weak Evidence + More Important)
Select maximum 3 assumptions from this quadrant
If more than 3 in quadrant: Prioritize by:
- Impact on solution success
- Test complexity (easier first)
- Dependencies (blocking first)
Mark selected LoFA with visual indicator (circle or highlight)

Success Criteria Pattern

Define n participants and success threshold as an absolute number (not %).
Example: "≥ 4 of 10 choose sports content on the prototype home screen."

Quality Indicators

Strong

Specific & Positive: Testable statements tied to concrete behavior
Evidence-Linked: Quotes, analytics, or prior tests attached
Right LoFA: Maximum 3 from top-right quadrant (Weak Evidence + More Important)
Binary Classification: Consistent Strong/Weak, More/Less Important evaluation
Clear Criteria: Absolute numbers, audience defined, timeboxed
Iterative Cadence: Start small; scale only with positive signals

Weak

Too Many LoFA: More than 3 assumptions selected
Inconsistent Classification: Mixed evaluation criteria
Vague: Generic or compound statements
Opinion-Based: Future-intent questions; no behavior
Ambiguous Criteria: Percentages, no audience, no time window
Over-Building: Large experiments before early signals

Common Anti-Patterns and Guardrails

Too many LoFA → Select maximum 3 from top-right quadrant only
Inconsistent classification → Use binary classification (Strong/Weak, More/Less Important)
Not generating enough assumptions → Use story map + pre-mortem + OST lines
Negative phrasing → Rewrite as what must be true (positive)
Not specific enough → Add actor, context, behavior, and outcome
Favoring one category → Cover all five: Desirability/Usability/Feasibility/Viability/Ethical
Overly complex simulations → Design smallest viable simulation first
Using percentages for criteria → Use absolute counts (e.g., 3 of 10)
Missing evaluation details → Define audience, n, window, and behavior
Testing with wrong audience → Screen for the target opportunity
Designing for less than best-case → Start where passing is most likely; increase difficulty later

Error Handling

Insufficient Data: Request more snapshots/synthesis; reduce scope.
Weak Evidence: Mark evidence as "Weak" and prioritize accordingly.
Conflicting Results: Triangulate with another small method before decisions.
Ethical Concerns: Document potential harms; design mitigation and/or halt.

File Naming Issues

Topic Extraction Failure: Ensure clear topic identification from opportunity content
Version Conflicts: Always check existing files before creating new assumptions
Incorrect Format: Use kebab-case for opportunity-name, v[number] for version
Missing Topic: Extract main theme from opportunity document for filename
Overwriting Prevention: Never overwrite existing assumption files, always increment version
File Existence Check Failure: MANDATORY - Always verify file existence before creation
Version Detection Error: If version detection fails, start with v1 and add warning note

Process Flow

Individual Interviews → Create Snapshots → Synthesize Patterns → Create Opportunities → Generate Solutions → Identify & Test Assumptions
     ↓                    ↓                    ↓                    ↓                       ↓                        ↓
[Raw Data]        [Structured Stories]   [Shared Patterns]    [Problem Statements]    [Product Ideas]      [Risks & Tests]

Recommended Folder Structure

assumptions/
├── newsletter-creation/
│   ├── assumptions-newsletter-creation-v1.md
│   └── assumptions-newsletter-creation-v2.md
├── user-onboarding/
│   ├── assumptions-user-onboarding-v1.md
│   └── assumptions-user-onboarding-v2.md
└── payment-flow/
    └── assumptions-payment-flow-v1.md

Quality Assurance Checklist

Input Validation

Clear target opportunity statement with supporting evidence
Opportunity context and problem definition complete
Solution ideas or user journeys defined
Evidence strength assessed and documented

Assumption Quality

Assumptions cover all five categories (Desirability, Usability, Feasibility, Viability, Ethical)
Assumptions are positive, specific, and testable
Evidence linked to each assumption where available
Assumption mapping completed with LoFA identified

Testing Quality

Test cards designed for LoFA assumptions
Success criteria defined with absolute numbers
Audience screening criteria specified
Sample size and time window defined

Process Completion

All assumptions evaluated and mapped
Test results documented and analyzed
Decisions made based on evidence
Next steps are clear and actionable

File Naming Validation

MANDATORY: Extracted clear topic from opportunity content before creating filename
MANDATORY: Checked existing files with pattern assumptions-[topic]-v*.md before creation
MANDATORY: Verified no file with new filename exists before creation
Filename uses semantic naming format: assumptions-[opportunity-name]-v[version].md
Version number is correctly auto-incremented for same-opportunity assumptions
Opportunity name is descriptive and kebab-case formatted
CRITICAL: No overwriting of existing assumption files - always create new version
Topic extraction completed through content analysis before assumption creation
Version management process followed step-by-step

AI Implementation Checklist

Before Creating Any Assumption File:

Topic Extraction ✅
- Analyze opportunity document for main theme
- Extract kebab-case topic name (e.g., "newsletter-creation")
- Verify topic is descriptive and unique
File Existence Check ✅
- Search for existing files: assumptions-[topic]-v*.md
- List all found files with their version numbers
- Identify highest version number
Version Management ✅
- Calculate next version number (highest + 1)
- Generate new filename: assumptions-[topic]-v[version].md
- Verify new filename doesn't already exist
File Creation ✅
- Create file with new filename only
- Never overwrite existing files
- Add version number to document header

Error Handling:

If topic extraction fails: Use generic topic name and add warning
If version detection fails: Start with v1 and add warning note
If file already exists: Increment version and try again
If multiple topics found: Use most relevant one and document choice

identify-test-assumptions