---
# === IDENTITY ===
id: business/startup/customer-discovery-playbook/2026
canonical_question: "How do I actually run customer discovery — recruit subjects, conduct Mom Test interviews, synthesize findings, and make a pivot-or-persevere decision?"
aliases:
  - "customer discovery interview execution guide"
  - "how to run Mom Test interviews for startup validation"
  - "customer interview sample size and saturation methodology"
  - "startup customer discovery synthesis and pivot criteria"
  - "lean customer development interview-to-decision pipeline"
entity_type: execution_recipe
domain: business > startup > Customer Discovery Playbook
region: global
jurisdiction: global
temporal_scope: 2024-2026

# === VERIFICATION ===
last_verified: 2026-03-11
confidence: 0.92
version: 2.0
first_published: 2026-03-10

# === TEMPORAL VALIDITY ===
temporal_validity:
  status: evolving
  last_breaking_change: "AI transcription tools (Otter.ai, tl;dv, Dovetail AI) reduced interview synthesis time by 70% in 2025; async video interview platforms added as viable alternative to live calls"
  next_review: 2026-09-07
  change_sensitivity: high

# === CONSTRAINTS ===
constraints:
  - "Minimum 12-20 interviews per customer segment to reach thematic saturation — 92% of themes identified within first 12 interviews, but meaning saturation requires 16-24"
  - "Never ask 'Would you use this?' or 'Would you pay for this?' — these produce false positives because people are polite, not honest"
  - "Discovery is not selling — the moment you pitch, the interview becomes a sales call and all subsequent answers are contaminated"
  - "Synthesize within 48 hours of each interview — delayed synthesis loses nuance and emotional context"
  - "Budget 150 outreach contacts to secure 30 interviews — expect 20% conversion rate on cold outreach, 50%+ on warm introductions"

# === SKIP CONDITIONS ===
skip_this_unit_if:
  - condition: "User already has product-market fit and is optimizing, not discovering"
    use_instead: "business/startup/pricing-experimentation-playbook/2026"
  - condition: "User needs a strategic framework, not execution steps"
    use_instead: "business/startup/startup-launch-checklist/2026"
  - condition: "User needs quantitative survey design, not qualitative interviews"
    use_instead: "Search knowledgelib.io for quantitative survey design — no dedicated unit yet"

# === AGENT HINTS ===
inputs_needed:
  - key: discovery_stage
    question: "What stage of discovery are you in?"
    type: choice
    options: ["problem discovery (do they have this pain?)", "solution discovery (does our solution fit?)", "business model discovery (will they pay?)"]
  - key: technical_skill
    question: "What is the user's technical skill level?"
    type: choice
    options: ["non-technical (no-code only)", "semi-technical (can edit code)", "developer (can write code)"]
  - key: budget_for_tools
    question: "What's the tool/incentive budget?"
    type: choice
    options: ["free tier only ($0)", "lean ($100-$500)", "standard ($500-$2,000)"]
  - key: access_to_customers
    question: "How easy is it to reach target customers?"
    type: choice
    options: ["easy (B2C, online communities, personal network)", "moderate (B2B, requires introductions)", "difficult (enterprise, regulated, niche)"]

# === EXECUTION METADATA ===
execution:
  required_inputs:
    - name: "Problem hypothesis"
      source: "Founder brainstorming or market observation"
      format: "document"
    - name: "Target customer profile"
      source: "Founder domain knowledge or initial research"
      format: "structured data"
  outputs:
    - name: "Interview synthesis report"
      format: "spreadsheet"
      description: "Theme frequency analysis, pain point rankings, customer segment map, key quotes, and saturation assessment across all interviews"
    - name: "Pivot-or-persevere decision document"
      format: "document"
      description: "Evidence-backed go/pivot/stop decision with hypothesis scorecard, validated ICP, and next-step recommendations"
    - name: "Customer language bank"
      format: "spreadsheet"
      description: "Exact words, phrases, and descriptions customers use to describe their problems — feeds directly into marketing copy and positioning"
  tools_required:
    - name: "Calendly / Cal.com"
      purpose: "Interview scheduling with automated reminders"
      tier: "free"
      cost: "$0 (free tier: unlimited 1-on-1 meetings)"
      alternatives: ["Cal.com (open source)", "Google Calendar direct"]
    - name: "Zoom / Google Meet"
      purpose: "Video interviews with recording"
      tier: "free"
      cost: "$0 (free tier: 40-60 min meetings)"
      alternatives: ["Riverside.fm", "Around", "Microsoft Teams"]
    - name: "Otter.ai / tl;dv"
      purpose: "AI transcription and highlight extraction"
      tier: "free"
      cost: "$0-$16.99/mo"
      alternatives: ["Rev ($1.50/min)", "Dovetail ($29/mo)", "manual notes"]
    - name: "Google Sheets / Notion"
      purpose: "Interview tracking, theme coding, synthesis"
      tier: "free"
      cost: "$0"
      alternatives: ["Airtable", "Dovetail ($29/mo)", "Miro"]
    - name: "Respondent.io / User Interviews"
      purpose: "Paid participant recruitment (if cold outreach fails)"
      tier: "paid"
      cost: "$50-$150/participant"
      alternatives: ["Prolific", "UserTesting", "personal network (free)"]
  credentials_needed:
    - service: "Calendly"
      type: "OAuth"
      where_to_get: "https://calendly.com/signup"
      free_tier_limits: "Unlimited 1-on-1 meetings, 1 event type"
    - service: "Otter.ai"
      type: "OAuth"
      where_to_get: "https://otter.ai/signup"
      free_tier_limits: "300 minutes/month transcription, 30 min per conversation"
  estimated_duration: "4-8 weeks (20 interviews standard)"
  estimated_cost: "$0 (free tier) to $2,000 (paid recruitment + incentives)"

# === DISTRIBUTION ===
canonical_source: "https://knowledgelib.io/business/startup/customer-discovery-playbook/2026"
suggested_citation: "Source: knowledgelib.io — AI Knowledge Library (verified 2026-03-11)"

# === RELATED UNITS ===
related_kos:
  depends_on: []
  feeds_into:
    - id: "business/startup/idea-validation-playbook/2026"
      label: "Discovery insights feed validation experiments"
    - id: "business/startup/go-to-market-launch-playbook/2026"
      label: "Validated ICP feeds GTM channel selection"
    - id: "business/startup/brand-identity-creation-playbook/2026"
      label: "Customer language bank informs brand messaging"
  related_to:
    - id: "business/startup/pricing-experimentation-playbook/2026"
      label: "Pricing research extends discovery into willingness-to-pay"
    - id: "business/startup/mvp-scoping-framework/2026"
      label: "MVP feature scoping uses validated pain points"
  alternative_to: []

# === SOURCES ===
sources:
  - id: src1
    title: "How to Do Customer Discovery to Build a Startup: The Definitive Guide"
    author: Lean B2B
    url: https://leanb2bbook.com/blog/customer-discovery-startup-guide/
    type: industry_report
    published: 2024-08-15
    reliability: authoritative
  - id: src2
    title: "The Mom Test: Everything You Need to Know (2026)"
    author: tl;dv
    url: https://tldv.io/blog/the-mom-test/
    type: industry_report
    published: 2026-01-15
    reliability: high
  - id: src3
    title: "How Many Participants for a UX Interview?"
    author: Nielsen Norman Group
    url: https://www.nngroup.com/articles/interview-sample-size/
    type: industry_report
    published: 2024-06-01
    reliability: authoritative
  - id: src4
    title: "A Simple Method to Assess and Report Thematic Saturation in Qualitative Research"
    author: Guest, Namey & Chen (PLOS ONE)
    url: https://pmc.ncbi.nlm.nih.gov/articles/PMC7200005/
    type: academic_paper
    published: 2020-05-01
    reliability: authoritative
  - id: src5
    title: "Customer Discovery Interviews: A Secret of Successful Startups"
    author: GrowthRamp
    url: https://www.growthramp.io/articles/customer-discovery-interviews
    type: industry_report
    published: 2025-03-01
    reliability: high
  - id: src6
    title: "Customer Discovery: How Product Teams Lead the Charge"
    author: Product School
    url: https://productschool.com/blog/user-experience/customer-discovery
    type: industry_report
    published: 2025-02-10
    reliability: high
  - id: src7
    title: "How to Talk to Users"
    author: Y Combinator (Gustaf Alstromer)
    url: https://www.ycombinator.com/library/Iq-how-to-talk-to-users
    type: official_docs
    published: 2024-01-01
    reliability: authoritative
  - id: src8
    title: "Steve Blank on Customer Discovery"
    author: Steve Blank
    url: https://steveblank.com/tag/customer-discovery/
    type: official_docs
    published: 2024-01-01
    reliability: authoritative
---

# Customer Discovery Recipe: Interviews to Pivot-or-Persevere Decision

## Purpose

This recipe produces a validated pivot-or-persevere decision backed by 15-25 customer discovery interviews, a theme-coded synthesis database, and a customer language bank — within 4-8 weeks at $0-$2,000. It executes the full discovery pipeline: hypothesis formulation, subject recruitment, Mom Test interviews, thematic saturation monitoring, affinity-mapped synthesis, and an evidence-scored decision. The output feeds directly into idea validation, MVP scoping, and go-to-market planning. [src1]

## Prerequisites
<!-- Agents: verify ALL prerequisites before executing. Missing prerequisites = failed execution. -->

- [ ] **Problem hypothesis** — "I believe [customer segment] experiences [specific problem] when [trigger], which costs them [time/money/frustration]" (can be rough, will be refined)
- [ ] **Target customer profile** — At least one segment defined with specific criteria (role, company size, industry, behavior) — specific enough to recruit against
- [ ] **150+ reachable prospects** identified across LinkedIn, communities, personal network — budget 150 contacts to get 30 interviews (20% cold conversion rate) [src5]
- [ ] **Calendly account** — Free tier at [calendly.com](https://calendly.com/signup) for scheduling with automated reminders
- [ ] **Video conferencing** — Zoom or Google Meet with recording capability (free tier sufficient)
- [ ] **Transcription tool** — Otter.ai free tier (300 min/month) or tl;dv for automated notes
- [ ] **Tracking spreadsheet** — Google Sheets template with columns: Name, Date, Segment, Pain (1-10), Frequency, Current Solution, Spend, Key Quotes, New Themes
- [ ] **8-15 hours/week available** over 4-8 weeks for interviews, synthesis, and analysis

## Constraints
<!-- Hard rules. Agents: enforce throughout execution. Violating these = broken output or legal risk. -->

- Never ask "Would you use this?" or "Would you pay for this?" — these produce false positives because humans are polite, not honest. In one study, 80% who said "I would buy this" did not buy when given the chance. [src2]
- Minimum 12 interviews per segment for code saturation (identifying the range of themes); 16-24 for meaning saturation (nuanced understanding). Research shows 92% of themes appear in the first 12 interviews. [src4]
- Never pitch your solution during problem discovery — "You aren't allowed to tell them what their problem is and in return they aren't allowed to tell you what to build." [src2]
- Synthesize within 48 hours of each interview — delayed synthesis loses emotional context and non-verbal observations that recordings alone cannot capture. [src1]
- Two-person interview team recommended (interviewer + note-taker) — solo interviews lose 40-60% of observational data. If solo, recording with consent is mandatory. [src1]
- Budget 150 outreach contacts to secure 30 scheduled interviews — expect 20% conversion on cold outreach, 50%+ on warm introductions. [src5]

## Tool Selection Decision

<!-- Agent selects the right tool path based on user inputs.
     Each path leads to a different execution flow below. -->

```
Which path?
├── User has $0 budget AND easy customer access
│   └── PATH A: Free Manual — Calendly + Google Meet + Google Sheets + manual notes
├── User has $0 budget AND moderate/difficult access
│   └── PATH B: Free + AI Assist — Cal.com + Zoom + Otter.ai Free + Google Sheets
├── User has $100-$500 budget
│   └── PATH C: Lean — Calendly + Zoom + Otter.ai Pro + Notion/Airtable + gift cards
└── User has $500-$2,000 budget
    └── PATH D: Full Stack — Calendly + Zoom + Dovetail + Respondent.io + gift cards
```

| Path | Tools | Cost | Speed | Output Quality |
|------|-------|------|-------|---------------|
| A: Free Manual | Calendly, Meet, Sheets, manual notes | $0 | 6-8 weeks | Good — requires disciplined manual coding |
| B: Free + AI Assist | Cal.com, Zoom, Otter.ai, Sheets | $0 | 5-7 weeks | Good — AI transcription improves accuracy |
| C: Lean | Calendly, Zoom, Otter.ai Pro, Notion | $100-$500 | 4-6 weeks | High — incentives improve recruitment speed |
| D: Full Stack | Calendly, Zoom, Dovetail, Respondent.io | $500-$2,000 | 4-5 weeks | Excellent — professional recruitment + analysis |

## Execution Flow

### Step 1: Formulate Falsifiable Hypotheses

**Duration**: 2-4 hours
**Tool**: Google Sheets or Notion

Write 3-5 core problem hypotheses using Steve Blank's format: "We believe [customer segment] experiences [specific problem] when [trigger/context], which costs them [time/money/frustration]." Rank by risk — which ones, if wrong, kill the business model? Convert the top 3-5 into testable hypotheses with explicit pass/fail criteria. [src8]

```
Hypothesis template:
"We believe [customer segment] experiences [problem] at [frequency].
We'll know this is TRUE when [X] of [Y] interviewees describe this pain unprompted.
We'll know this is FALSE when fewer than [Z] of [Y] mention it at all."

Example:
"We believe SaaS founders with 10-50 employees spend 5+ hours/week
on manual customer onboarding workflows.
TRUE: 10 of 15 interviewees describe manual onboarding as a top-3 pain.
FALSE: Fewer than 4 of 15 mention onboarding pain without prompting."

Validation thresholds (from Lean B2B):
- Validated: 60%+ unprompted mention with emotional intensity
- Partially validated: 30-60% mention
- Invalidated: Under 30% or directly contradicted
```

Identify the "riskiest assumption" — the single hypothesis that, if wrong, makes the entire business model impossible. This gets tested first. [src8]

**Verify**: 3-5 hypotheses documented with specific pass/fail thresholds; riskiest assumption identified
**If failed**: If you cannot articulate 3 testable assumptions, spend 2-4 hours on competitor research and customer community lurking first — read forums, Reddit, Hacker News, industry Slack channels

### Step 2: Design the Interview Guide (Mom Test Format)

**Duration**: 2-3 hours
**Tool**: Google Docs or Notion

Write 8-12 open-ended questions following the Mom Test's three core rules: (1) talk about their life instead of your idea, (2) ask about specifics in the past instead of hypotheticals about the future, (3) talk less and listen more. [src2]

```
Interview structure (30-40 minutes):

WARM-UP (2-3 min):
"Thanks for taking the time. I'm researching how [role] teams handle
[domain area]. No right or wrong answers — I just want to learn."

CONTEXT (5 min):
1. "Can you walk me through your current workflow for [domain]?"
2. "What does a typical week look like for you regarding [area]?"

PROBLEM EXPLORATION (15-20 min):
3. "What's the hardest part about [domain]?"
4. "Tell me about the last time you dealt with [problem area]."
   Follow-up: "What happened? How did it end?"
5. "What have you tried to solve it?"
   Follow-up: "How did that work out? What broke?"
6. "Why was that hard?" (dig for root cause, not surface complaint)
7. "How much time/money do you spend on this today?"

CURRENT SOLUTIONS (5-10 min):
8. "What tools or workarounds do you use right now?"
9. "If you could wave a magic wand, what would change?"
10. "What would happen if this problem stayed unsolved for another year?"

WRAP-UP (3 min):
11. "Is there anything else about [domain] that frustrates you that
    I haven't asked about?"
12. "Who else should I talk to about this?"
    (Referral chain — doubles recruitment rate)

NEVER ASK:
- "Would you use a product that does X?"
- "How much would you pay for this?"
- "Is this a good idea?"
- "What features would you want?"
```

Eliminate three types of bad data: (1) compliments — generic praise without specifics, (2) hypothetical fluff — language using "might," "could," "would," (3) wishlists — feature requests without behavioral backing. Redirect fluff with: "When did it last happen? Walk me through the experience." [src2]

**Verify**: Interview guide reviewed by at least one other team member; no leading or hypothetical questions present
**If failed**: Remove any question that starts with "Would you..." or "Do you think..." — these violate Mom Test rules and contaminate every answer that follows

### Step 3: Recruit Interview Subjects (Target: 25-35 Scheduled)

**Duration**: 5-10 days
**Tool**: LinkedIn, Slack/Discord communities, Reddit, personal network, email

Identify 100-150 people matching the target customer profile across multiple channels. Send short, personalized outreach. End every email with a yes/no question — this increases reply rates from 4.8% to 12.8%. [src5]

```
Cold outreach template (B2B, keep under 150 words):
Subject: Quick question about [domain] (under 40 characters)

Hi [Name],

I'm researching how [role] teams handle [problem area].
Would you have 20 minutes this week for a quick call?

No pitch — I'm just trying to understand workflows.
Happy to share findings afterward.

Would [Tuesday/Thursday] work?

Best,
[Name]

Warm intro template (via mutual connection):
"[Mutual connection] suggested I reach out. I'm studying how
[role] teams handle [problem] and your experience would be
incredibly valuable. 20 minutes — no pitch, just learning.
Would you be open to a quick call?"

Target response rates:
- Cold LinkedIn: 10-15% response, 5-8% scheduled
- Cold email: 10-20% response (with follow-up)
- Warm intro: 40-60% response
- Community post: 3-5% of viewers respond
```

Over-recruit by 30-40% to account for no-shows (20-30% no-show rate is normal). Send a reminder 24 hours before and 1 hour before each interview. Max 3-4 interviews per day to prevent interviewer fatigue. [src1]

For B2B: offer $25-$100 gift cards to improve response rates. For B2C: $10-$50 or a coffee voucher. Incentives improve response rates from ~10% to ~25%. [src5]

**Verify**: 20+ interviews scheduled with screened participants within 10 days; screening confirms segment match
**If failed**: If fewer than 15 scheduled after 7 days, try these recovery actions in order: (1) broaden customer profile slightly, (2) try paid platforms like Respondent.io at $50-$150/participant, (3) post in 3+ new communities, (4) ask every interviewee for 2 referrals. If you still cannot recruit 15, the segment may be too narrow or hard to reach — this is itself a validation signal about market size. [src5]

### Step 4: Conduct Interviews and Track in Real-Time

**Duration**: 2-4 weeks (3-4 interviews/week recommended pace)
**Tool**: Zoom/Google Meet (recording), Otter.ai (transcription), Google Sheets (tracking)
**Rate limit**: Max 3-4 interviews per day. Space at least 30 minutes between sessions for debrief.

Follow the Mom Test methodology: listen 80%, talk 20%. Never pitch, never lead, never suggest solutions. Watch for three types of bad data: (1) compliments ("That's a great idea!"), (2) hypothetical fluff ("I would definitely..."), (3) wishlists ("It would be nice if..."). Redirect fluff with: "When did it last happen? Walk me through the experience." [src2]

```
Real-time tracking spreadsheet columns:

| # | Name | Date | Segment | Pain Score (1-10) | Frequency |
|   | Current Solution | Monthly Spend | Key Quotes (verbatim) |
|   | Emotional Intensity (low/med/high) | New Themes Found |
|   | Workarounds Described | Would Refer Others? (Y/N) |

After each interview, capture immediately:
1. Top 3 insights (what surprised you?)
2. Hypothesis evidence: which hypotheses gained/lost support?
3. New themes: anything you hadn't heard before?
4. Customer language: exact words they used for the problem
5. Behavioral signals: workarounds, spending, time investment

Red flags (interview may be contaminated):
- You talked more than 30% of the time
- You described your solution before minute 25
- Interviewee gave only compliments and no specifics
- You asked "Would you use/buy...?" at any point
```

Conduct a 5-minute debrief immediately after each interview: interviewer + note-taker compare observations, note non-verbal cues, and flag quotes worth preserving verbatim. After every 5 interviews, review notes for emerging themes. Add questions exploring unexpected patterns. Remove questions that consistently produce uninformative answers. Do not change core hypothesis questions mid-stream. [src1]

**Verify**: After every 5 interviews, check: Are new themes still emerging? Is pain score converging? Do at least 3 interviewees describe the same problem unprompted?
**If failed**: If all interviewees are polite but describe no real pain, the problem hypothesis is likely wrong. Pivot the hypothesis: ask "what IS painful about [domain]?" and let them lead. Update your hypothesis based on what actually emerges. [src6]

### Step 5: Monitor Thematic Saturation

**Duration**: Ongoing from interview 10 onward (assessed every 3 interviews)
**Tool**: Google Sheets (saturation tracker)

Track how many new themes emerge per interview using Guest, Namey & Chen's methodology. Thematic saturation is reached when 3 consecutive interviews produce zero or one new themes at the code level. [src4]

```
Saturation tracking method:

Base size: 6 interviews (initial analysis batch)
Run length: 3 interviews (consecutive check window)
New information threshold: <=5% new codes = saturated

Research benchmarks:
- First 6 interviews: ~78-79% of all themes identified
- First 10-12 interviews: ~92% of themes identified
- Code saturation (theme range): typically 9-12 interviews
- Meaning saturation (nuanced understanding): 16-24 interviews

Saturation log:
| Interviews Completed | New Themes This Batch | Total Themes | Saturation? |
| 6                    | N/A (baseline)        | 12           | No          |
| 9                    | 3                     | 15           | No          |
| 12                   | 1                     | 16           | Approaching |
| 15                   | 0                     | 16           | YES         |

B2B problems: saturation typically at 15-20 interviews
B2C problems: may require 20-30 due to population diversity
Multiple segments: 12-15 per segment minimum
```

Do not stop before 12 interviews even if patterns seem clear — premature pattern-matching is the most common discovery bias. At the same time, do not interview indefinitely — if saturation is reached at 15, conducting 30 more adds cost without insight. [src4]

**Verify**: Saturation log shows <=1 new theme in last 3 interviews; minimum 12 interviews completed
**If failed**: If still discovering major new themes at interview 20, the segment is likely too broad. Sub-segment the population and continue with a tighter group. If no convergence after 25 interviews across a tight segment, the problem may be too heterogeneous to address with one product. [src3]

### Step 6: Synthesize Findings with Affinity Mapping

**Duration**: 2-4 days
**Tool**: Google Sheets (theme coding), Miro or FigJam (affinity map), or Dovetail

Transform raw interview data into structured insights using thematic coding and affinity mapping. This is where data becomes decisions. [src6]

```
Synthesis process:

STEP A — Theme coding (2-3 hours):
1. Review all interview notes and transcripts
2. Tag each observation with a theme code
3. Group codes into 5-8 parent themes
4. Count frequency: how many participants mentioned each theme?
5. Weight unprompted mentions 3x over prompted responses

STEP B — Affinity mapping (2-3 hours):
1. Write each key quote/observation on a virtual sticky note
2. Cluster related notes without predefined categories
3. Name each cluster — these are your validated themes
4. Rank clusters by: frequency x intensity x willingness to spend
5. Identify the "hair on fire" problem: highest frequency +
   highest emotional intensity + existing workaround spending

STEP C — Customer language bank (1-2 hours):
1. Extract exact words customers used for the problem
2. Document metaphors, analogies, and emotional phrases
3. These become marketing copy — "The precise words customers
   use should be in your marketing material" [src5]

STEP D — Hypothesis scorecard (1-2 hours):
| Hypothesis | Unprompted Mentions | Pain Intensity | Status      |
| H1: [text] | 12 of 18 (67%)      | 8.2/10 avg     | VALIDATED   |
| H2: [text] | 5 of 18 (28%)       | 4.1/10 avg     | INVALIDATED |
| H3: [text] | 8 of 18 (44%)       | 6.8/10 avg     | PARTIAL     |
| H4 (NEW):  | 9 of 18 (50%)       | 7.5/10 avg     | EMERGED     |

Validation thresholds:
- Validated: 60%+ unprompted mention with intensity 7+/10
- Partially validated: 30-60% mention OR intensity 5-7/10
- Invalidated: Under 30% mention or intensity below 5/10
```

**Verify**: Synthesis complete with theme frequency table, affinity map, customer language bank, and hypothesis scorecard; at least one hypothesis has clear validated/invalidated status
**If failed**: If no clear pattern emerges from synthesis, bring in a second analyst for independent coding — compare results. Disagreements often reveal hidden assumptions.

### Step 7: Make Pivot-or-Persevere Decision

**Duration**: 1-2 days
**Tool**: Google Sheets (decision matrix)

Apply evidence from all prior steps to make a definitive go/pivot/stop decision. Do not cherry-pick positive signals. Founders who love their original idea ignore contradicting evidence in 60%+ of cases. [src1]

```
Decision matrix:
+---------------------------------+----------+----------+----------+
| Signal                          | Stop     | Pivot    | Go       |
+---------------------------------+----------+----------+----------+
| Problem pain score (avg)        | <5/10    | 5-7/10   | 7+/10    |
| Problem pattern clarity         | No pattern| Emerging | 10+ same |
| Unprompted mention rate         | <30%     | 30-60%   | >60%     |
| Existing workaround spending    | $0       | $1-$50/mo| >$50/mo  |
| Current solution satisfaction   | High     | Moderate | Low      |
| Referral willingness            | 0-1      | 2-4      | 5+       |
| "Hair on fire" problem found    | No       | Maybe    | Clear    |
| Customer segments identified    | 0        | 1        | 2+       |
+---------------------------------+----------+----------+----------+

Scoring: 5+ "Go" signals = proceed to validation/MVP
         3+ "Pivot" signals = adjust hypothesis and re-interview
         3+ "Stop" signals = abandon this problem space

If pivoting, define the specific pivot type:
- Customer segment pivot (same problem, different audience)
- Problem pivot (same audience, different problem)
- Solution pivot (same problem, different approach)
- Channel pivot (same everything, different distribution)
```

Set a hard deadline: make the decision within 2 weeks of completing the last interview. Discovery without a forcing function becomes academic research, not startup strategy. Perfect information does not exist — the goal is sufficient evidence to reduce risk, not certainty. [src1]

**Output files**:
- `discovery_synthesis.csv` — Full theme-coded interview database with frequencies, intensity scores, and quotes
- `hypothesis_scorecard.md` — Each hypothesis scored as validated/partial/invalidated with evidence
- `customer_language_bank.csv` — Exact customer phrases organized by theme (for marketing copy)
- `decision_document.md` — Go/pivot/stop decision with supporting evidence summary and next steps

**Verify**: Decision document completed with supporting evidence; strongest validated hypothesis and customer segment identified; next action defined
**If failed**: If evidence is genuinely ambiguous after 20+ interviews and thorough synthesis, extend by 5-8 interviews with a tighter segment. If still ambiguous after 30 total, treat as "stop" — ambiguity at this scale means the signal is too weak to build a business on. [src1]

## Output Schema

```json
{
  "output_type": "customer_discovery_package",
  "format": "document collection",
  "columns": [
    {"name": "decision", "type": "string", "description": "Go, Pivot, or Stop with reasoning", "required": true},
    {"name": "core_hypothesis_validated", "type": "boolean", "description": "Whether the riskiest assumption was confirmed", "required": true},
    {"name": "hair_on_fire_problem", "type": "string", "description": "Highest-frequency, highest-intensity validated pain point", "required": true},
    {"name": "target_segment", "type": "string", "description": "Best-fit customer segment with specific profile criteria", "required": true},
    {"name": "unprompted_mention_rate", "type": "number", "description": "Percentage of interviewees who described the core problem unprompted", "required": true},
    {"name": "average_pain_intensity", "type": "number", "description": "Mean pain score (1-10) for the core problem", "required": true},
    {"name": "interviews_completed", "type": "number", "description": "Total interviews conducted", "required": true},
    {"name": "saturation_reached", "type": "boolean", "description": "Whether thematic saturation was achieved", "required": true},
    {"name": "themes_identified", "type": "number", "description": "Total unique themes coded from interviews", "required": true},
    {"name": "customer_language_entries", "type": "number", "description": "Number of verbatim customer phrases captured", "required": false},
    {"name": "pivot_type", "type": "string", "description": "If pivoting: segment, problem, solution, or channel pivot", "required": false}
  ],
  "expected_row_count": "1 (single discovery decision)",
  "sort_order": "N/A",
  "deduplication_key": "core_hypothesis + target_segment"
}
```

## Quality Benchmarks

| Quality Metric | Minimum Acceptable | Good | Excellent |
|---------------|-------------------|------|-----------|
| Interviews completed per segment | 12 | 18-20 | 25+ |
| Thematic saturation reached | Approaching (1-2 new themes) | Reached (0-1 new in last 3) | Confirmed (0 new in last 5) |
| Unprompted mention rate (core problem) | >30% | >50% | >70% |
| Average pain intensity (core problem) | >5/10 | >7/10 | >8/10 |
| Hypothesis scorecard completeness | All scored | All scored + evidence | All scored + quotes + data |
| Customer language bank entries | 20+ phrases | 50+ phrases | 100+ phrases |
| Synthesis completed within deadline | Within 2 weeks | Within 1 week | Within 3 days |

**If below minimum**: Re-run Steps 3-4 with broader segment or different channels. If interviews are below 12 per segment, recruit more before synthesizing — premature pattern-matching with small samples is the #1 discovery failure mode. [src4]

## Error Handling

| Error | Likely Cause | Recovery Action |
|-------|-------------|----------------|
| Cannot recruit 15 interviewees in 10 days | Target segment too narrow or wrong channels | Broaden profile; try Respondent.io ($50-$150/participant); ask every interviewee for 2 referrals; post in 3+ communities |
| All interviewees polite but describe no real pain | Problem hypothesis is wrong | Pivot: ask "what IS painful about [domain]?" and let them lead; update hypothesis based on what emerges |
| Interviewees describe pain but it's a different problem | Hypothesis targeted wrong pain | Document the actual pain as a new hypothesis (H-new); this is a valuable pivot signal, not a failure |
| High pain scores but interviewees spend $0 on solutions | Problem is painful but not worth paying to solve | Test whether the pain is "vitamin" (nice to have) vs "painkiller" (must solve); pivot to higher-stakes version of problem |
| New themes still emerging at interview 20 | Segment too heterogeneous | Sub-segment: split into 2-3 narrower profiles and continue; each sub-segment needs 12+ interviews |
| Interviewer dominated conversation (>30% talk time) | Insufficient Mom Test training | Discard contaminated interview; review recording; practice with a teammate before next session |
| Recording/transcription failed | Tool configuration issue | Always take manual backup notes; switch to alternative tool (Otter.ai to tl;dv or manual); re-interview if critical |
| Team disagrees on synthesis interpretation | Different analytical frames | Independent coding: each person codes themes separately, then compare; resolve disagreements with evidence counts |

## Cost Breakdown

| Component | Free ($0) | Lean ($500) | Standard ($2,000) |
|-----------|-----------|-------------|-------------------|
| Interview incentives | $0 (goodwill) | $200 (20 x $10 gift cards) | $750 (25 x $30 gift cards) |
| Transcription tools | $0 (Otter.ai free) | $0 (Otter.ai free) | $60 (Otter.ai Pro, 2 months) |
| Recruiting platforms | $0 (DIY) | $0 (DIY) | $500 (Respondent.io, 10 participants) |
| Analysis tools | $0 (Google Sheets) | $0 (Google Sheets) | $100 (Dovetail, 2 months) |
| Scheduling | $0 (Calendly free) | $0 (Calendly free) | $0 (Calendly free) |
| Video conferencing | $0 (Zoom/Meet free) | $0 (Zoom free) | $30 (Zoom Pro, 2 months) |
| Contingency | $0 | $300 | $560 |
| **Total** | **$0** | **$500** | **$2,000** |

**Constraint**: Zero-budget discovery is possible but slower — recruiting without incentives reduces response rates by 50-70% and skews toward people who enjoy talking rather than representative customers. [src1]

## Anti-Patterns

### Wrong: Asking "Would you use a product that does X?"
Future-prediction questions generate false positives. Customers agree to avoid social awkwardness. In one study, 80% of people who said "I would buy this" did not buy when given the chance. The Mom Test exists precisely because your mom will never tell you your idea is bad. [src2]

### Correct: Ask about past behavior with specifics
"Tell me about the last time you dealt with [problem]. What happened? How did it end?" Past behavior is the best predictor of future behavior. If customers do not currently spend time or money on workarounds, they are unlikely to buy your solution. [src2]

### Wrong: Conducting 5-8 interviews and declaring "we've validated the problem"
Five interviews in a heterogeneous population produce unreliable patterns. Confirmation bias means you will find what you are looking for in almost any small sample. Research shows you need 12+ interviews for code saturation and 16-24 for meaning saturation. [src4]

### Correct: Track saturation explicitly with a running log
Monitor new theme emergence per interview batch. Only declare saturation when 3 consecutive interviews produce zero or one new themes. Use the base-size-6, run-length-3 methodology from Guest et al. [src4]

### Wrong: Running discovery indefinitely without a decision point
Some teams interview for months without synthesizing or deciding. Discovery without a forcing function becomes academic research, not startup strategy. [src1]

### Correct: Set a hard deadline for the pivot-or-persevere decision
Commit to making a decision within 2 weeks of completing the last interview. Perfect information does not exist — the goal is sufficient evidence to reduce risk, not certainty. [src1]

### Wrong: Letting the interviewer talk more than 30% of the time
When founders talk, they pitch. When they pitch, interviewees switch from honest mode to polite mode. Three types of bad data follow: compliments, hypothetical fluff, and wishlists. [src2]

### Correct: Listen 80%, talk 20% — use silence as a tool
Ask a question and wait. Silence is uncomfortable but productive — interviewees fill it with the details you actually need. Your best insights come from follow-up probes ("Why?", "Tell me more"), not from your prepared questions. [src7]

## When This Matters

Use when a founder or agent needs to execute a complete customer discovery cycle — recruit subjects, run Mom Test interviews, code themes, monitor saturation, synthesize findings, and make an evidence-backed pivot-or-persevere decision. Not a document about discovery methodology, but the actual execution steps with tools, templates, and quality gates. Requires a problem hypothesis as input; produces a validated decision, customer language bank, and synthesis database as output.

## Related Units

- [Idea Validation Playbook](/business/startup/idea-validation-playbook/2026) — validation experiments after discovery confirms the problem
- [Go-to-Market Launch Playbook](/business/startup/go-to-market-launch-playbook/2026) — validated ICP from discovery feeds GTM channel selection
- [Brand Identity Creation Playbook](/business/startup/brand-identity-creation-playbook/2026) — customer language bank informs brand messaging
- [Pricing Experimentation Playbook](/business/startup/pricing-experimentation-playbook/2026) — extends discovery into willingness-to-pay testing
- [MVP Scoping Framework](/business/startup/mvp-scoping-framework/2026) — validated pain points feed feature prioritization