ditto-vc-diligence

name	ditto-vc-diligence
description	Run customer diligence on startups using Ditto's synthetic persona platform (300K+ AI personas, 92% overlap with real focus groups). Validates whether a startup's stated problem actually exists, tests willingness to pay, and compares founder claims against what customers actually want. Uses a thesis-first workflow with over-recruit and curate methodology for high-quality niche panels. Designed for VCs, angel investors, accelerators, and founders validating their own thesis. Use when the user mentions customer diligence, startup validation, investor due diligence, portfolio company research, or market validation for investment decisions.
allowed-tools	Bash(curl *), Bash(python3 *), Read, Grep, WebFetch
argument-hint	[startup URL or research brief]

Ditto for VC Customer Diligence

Validate startup opportunities by asking real-world personas about their problems, spending habits, and ideal solutions — without ever revealing the startup or VC context. Directly from the terminal.

Full documentation: https://askditto.io/claude-code-guide

The Golden Rule

Never reveal the VC context or name the startup in study questions.

Frame everything as neutral market research. The personas should tell YOU what they want — not react to what the startup claims. If a persona's "magic wand" answer matches the startup's value prop unprompted, that is genuine validation. If you lead them there, it is worthless.

This single constraint shapes every question in this skill.

What Ditto Does

Ditto maintains 300,000+ AI-powered synthetic personas calibrated to census data across USA, UK, Germany, and Canada. You ask them open-ended questions and get qualitative responses with the specificity of real interviews.

• 92% overlap with traditional focus groups (EY Americas validation)
• 95% correlation with traditional research
• Harvard/Cambridge/Stanford/Oxford peer-reviewed methodology
• A diligence study completes in 15-18 minutes (over-recruit + curate)
• Traditional customer diligence: 3-6 weeks, $15,000-40,000

Quick Start (Free Tier)

Get a free API key — no credit card, no sales call:

curl -sL https://app.askditto.io/scripts/free-tier-auth.sh | bash

Free keys (rk_free_): ~12 shared personas, no demographic filtering. Paid keys (rk_live_): custom groups, demographic filtering, over-recruit.

API Essentials

Base URL: https://app.askditto.io Auth header: Authorization: Bearer YOUR_API_KEY Content-Type: application/json

The Diligence Workflow (9 Steps)

This is NOT a standard study. The thesis-first approach and curation steps are what make diligence studies reliable. Do not skip steps.

Step 1: Research the Startup

Before touching the API, research:

• What the startup does (product, features, pricing)
• The pain point hypothesis (who has this problem?)
• The target customer (job title, industry, daily workflow)
• The competitive landscape (what exists today?)
• The market context (growing? shrinking? regulatory?)

Use web research, the startup's website, pitch deck, and any public info.

Step 2: Build a Thesis

Structure your thesis:

• Who experiences the pain: specific role, industry, company size
• The proposed solution: what the startup claims to fix
• Why it might work: evidence supporting the opportunity
• Why it might not: risks, competitive moats, adoption barriers
• Key assumptions to validate: 3-5 specific hypotheses for the study

Step 3: Get User Approval

STOP and present the thesis to the user before creating any study. The user must approve the direction, target audience, and key questions. This checkpoint prevents wasted studies and ensures alignment.

Step 4: Over-Recruit Your Panel

Recruit 15-20 personas with broader filters than the ideal target:

curl -s -X POST "https://app.askditto.io/v1/research-groups/recruit" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Auto Parts Professionals",
    "group_size": 15,
    "filters": {
      "country": "USA",
      "industry": ["Automotive Manufacturing"],
      "age_min": 28,
      "age_max": 60
    }
  }'

Save the uuid. Use group_size (not size), group uuid (not id).

Why over-recruit: Niche audiences often include tangential personas. Recruiting 15-20 and screening with Q1 produces a tighter panel than recruiting exactly 10.

Step 5: Create Study

curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "[Startup Name] Customer Validation Study",
    "objective": "Validate the [painpoint] for [target customers]",
    "research_group_uuid": "UUID_FROM_STEP_4"
  }'

Save the study id. Response nests under data.study — access via response["study"]["id"], NOT response["id"].

Step 6: Ask Q1 (Screening Question)

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/questions" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"question": "In your current or past work, how often do you need to [do the task]? Walk me through what that process typically looks like."}'

Poll until complete (45-50s first poll, then 20s intervals — see Polling Strategy).

Step 7: Curate the Panel

Retrieve Q1 responses:

curl -s "https://app.askditto.io/v1/research-studies/STUDY_ID/questions" \
  -H "Authorization: Bearer $DITTO_API_KEY"

Score each persona using Q1 response + agent_occupation + agent_summary:

Score	Criteria	Action
HIGH	Job title directly matches target customer. Q1 shows specific, detailed experience.	Keep
MEDIUM	Related role or adjacent industry. Q1 mentions relevant experience.	Keep
LOW	Tangentially related. Q1 is vague or generic about the task.	Remove
NONE	Completely unrelated role. No relevant experience in Q1.	Remove

Remove LOW and NONE personas from the study (not the group):

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/agents/remove" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_ids": [123, 456, 789]}'

CRITICAL:

• agent_ids must be list[int] — never strings or UUIDs
• Use study-level removal (/v1/research-studies/{id}/agents/remove), NOT group-level
• Agents stay in the group for potential reuse in follow-up studies
• Keep minimum 8 personas (never below 6)
• Ideal curated panel: 8-12 personas with demographic diversity and 2-3 different roles

Step 8: Ask Q2-Q7 (Curated Panel)

Ask remaining questions one at a time. Poll to completion between each. Only the curated personas will respond.

Step 9: Complete and Extract Insights

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/complete" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"force": false}'

Use "force": true to re-run analysis on an already-completed study (avoids 409).

Get share link (only if user requests it):

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/share" \
  -H "Authorization: Bearer $DITTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"enabled": true}'

The Non-Leading Question Framework

These 7 questions are specifically designed for investment diligence. The startup is NEVER mentioned. Adapt the bracketed terms to the specific pain point and task being validated.

Q1 — Screening / Relevance (asked before curation)

"In your current or past work, how often do you need to [do the task]? Walk me through what that process typically looks like."

Purpose: Filter irrelevant personas + establish baseline behaviour. After responses: CURATE the panel before Q2.

Q2 — Pain Identification

"What is the most frustrating part of [task]? Tell me about a time when it was particularly painful."

Purpose: Capture emotional pain and specific stories, not abstract ratings.

Q3 — Quantify Impact

"Roughly how much time per week do you or your team spend on [task]? What is the cost of that time to your business?"

Purpose: Hard numbers for ROI calculations — this is what VCs want.

Q4 — Current Solutions

"What tools, websites, or methods do you currently use to [solve problem]? What works well? What does not?"

Purpose: Reveals competitive landscape and switching costs.

Q5 — Past Attempts

"Have you ever tried a new tool or system specifically to make [task] easier? What happened? Why did you stick with it or abandon it?"

Purpose: Shows if this is an ACTIVE problem (they have tried to solve it).

Q6 — Magic Wand (the most valuable question)

"If you could wave a magic wand and fix ONE thing about how you [do task], what would it be?"

Purpose: Let THEM tell you what features matter. Compare their answer to the startup's actual value prop. This is where divergence appears.

Q7 — Adoption Barriers

"What would make you hesitant to switch to a new [solution type], even if it promised to save you time and money?"

Purpose: Surfaces sales objections, product requirements, and dealbreakers.

Insight Extraction for VCs

Structure your findings around these five areas:

1. Problem Validation

Do they experience this pain? How severe? Quantified in time/money?

• Strength: "8/10 described inventory management as a daily frustration"
• Risk: "Only 3/10 actively looking for a new solution"

2. #1 Pain Point

What frustration came up most consistently across participants?

• Look for emotional language and specific stories, not just complaints

3. Competitive Landscape

What have they tried? Why did current solutions fail?

• Map every tool/method mentioned in Q4 and Q5
• Note: "good enough" incumbents are harder to displace than bad ones

4. Magic Wand Consensus

What do they actually want? (Often diverges from the startup's pitch.)

• In 7/10 diligence studies, the founder's stated value prop diverged significantly from Q6 answers. When this happens, flag it prominently — it is the single most actionable finding for investors.

5. Adoption Barriers

What would stop them from switching? Dealbreakers?

• Rank by frequency across respondents
• Common pattern: "hassle" and "time" cited more than "money"

For each area, note:

• Strengths — evidence supporting the opportunity
• Risks — concerns or red flags
• Questions for Founders — what the investor should ask in the next meeting

Industry Proxy Mapping

Ditto's panel may not have exact industry matches. Use these validated proxies:

Target Audience	Industry Filter	Description Refinement
Auto mechanics / repair shops	Automotive Manufacturing	"Auto repair shop owners and mechanics"
Restaurant owners	Food & Beverages	"Independent restaurant owners and managers"
Pet store owners	Retail	"Pet retail store owners and pet care professionals"
Fitness studio owners	Health, Wellness & Fitness	"Gym and fitness studio owners"
Real estate agents	Real Estate	"Licensed real estate agents"
E-commerce sellers	Retail, Internet	"Online store operators and e-commerce managers"
Freelance designers	Design, Graphic Design	"Freelance graphic and web designers"
Content creators	Online Media, Marketing	"Full-time content creators and influencers"
Construction contractors	Construction	"General contractors and construction managers"
Veterinarians	Veterinary	"Veterinary practice owners and veterinarians"
Accountants / CPAs	Accounting	"CPA firm partners and accounting managers"
Insurance agents	Insurance	"Independent insurance agents and brokers"

Tip: When the exact industry isn't available, use the closest proxy and add a specific description to narrow recruitment. The description field is used by Ditto's matching algorithm and can compensate for imprecise industry filters.

Recruitment Failure Recovery

If recruitment returns fewer agents than expected:

1. Append to the same group with relaxed filters (don't create a new group):

   curl -s -X POST "https://app.askditto.io/v1/research-groups/GROUP_UUID/append" \
     -H "Authorization: Bearer $DITTO_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"group_size": 10, "filters": {"country": "USA", "age_min": 22, "age_max": 65}}'
   

   ⚠️ Note: append uses group_uuid (not group_id).

2. Relaxation order: Widen age → broaden industry → remove state → remove gender. Do NOT relax filters critical to the research question.

3. Maximum 10 attempts. Hard stop regardless of outcome.

4. Minimum viable group: 6 for reliable patterns. Below 6 = anecdotal, not systematic. Do NOT proceed with 0-3 agents.

Complete API Reference

Research Groups

Method	Endpoint	Purpose
POST	/v1/research-groups/recruit	Recruit group with demographic filters
POST	/v1/research-groups/create	Create group from explicit agent IDs
POST	/v1/research-groups/interview	AI-assisted recruitment from objective
GET	/v1/research-groups	List all groups (?limit=N&offset=N)
GET	/v1/research-groups/{id}	Get group details + agent profiles
POST	/v1/research-groups/{id}/update	Update name/description/dedupe
DELETE	/v1/research-groups/{id}	Archive group
POST	/v1/research-groups/{id}/agents/add	Add agents by ID
POST	/v1/research-groups/{id}/agents/remove	Remove agents permanently from group
POST	/v1/research-groups/{uuid}/append	Recruit more agents into existing group

Research Studies

Method	Endpoint	Purpose
POST	/v1/research-studies	Create study (research_group_uuid required)
GET	/v1/research-studies	List studies (?limit=N&offset=N)
GET	/v1/research-studies/{id}	Get study details
POST	/v1/research-studies/{id}/complete	Trigger AI analysis
POST	/v1/research-studies/{id}/agents/remove	Remove agents from study (curation)

Questions

Method	Endpoint	Purpose
POST	/v1/research-studies/{id}/questions	Ask question in study (returns job_ids)
GET	/v1/research-studies/{id}/questions	Get all Q&A data + agent profiles
POST	/v1/research-agents/{id}/questions	Quick question to one agent
POST	/v1/research-groups/{id}/questions	Quick question to entire group

Jobs, Sharing, Media

Method	Endpoint	Purpose
GET	/v1/jobs/{job_id}	Poll async job status
POST	/v1/research-studies/{id}/share	Enable/disable sharing
GET	/v1/research-studies/{id}/share	Check current share state
POST	/v1/media-assets	Upload image/PDF for question attachments

Agents

Method	Endpoint	Purpose
GET	/v1/agents/find	Find one matching persona
GET	/v1/agents/search	Search personas by demographics

Natural Language Requests

Method	Endpoint	Purpose
POST	/v1/research-study-requests	Create study from plain-text brief
POST	/v1/research-group-requests	Create group from description
GET	/v1/research-group-requests/{id}	Get group request status

Free Tier

Method	Endpoint	Purpose
POST	/v1/free/questions	Ask question to shared free-tier group

Demographic Filters

Filters go inside the filters dict when recruiting:

Filter	Type	Examples	Notes
country	string	"USA", "UK", "Canada", "Germany"	Required. Only these 4
state	string	"TX", "MI", "CA"	2-letter codes ONLY
city	string	"Austin", "Detroit"	Supported but narrows pool
age_min	integer	25, 30, 45	Recommended
age_max	integer	45, 55, 65	Recommended
gender	string	"male", "female", "non_binary"	Optional
is_parent	boolean	true, false	Optional
education	string	"high_school", "bachelors", "masters", "phd"	Optional
industry	array	["Healthcare", "Technology"]	Optional

NOT supported: income, employment, ethnicity, political_affiliation.

Polling Strategy

Study size	First poll delay	Subsequent interval	Strategy
15-20 personas (over-recruit)	60 seconds	20 seconds	Poll ONE job_id as proxy
8-12 personas (curated)	45-50 seconds	20 seconds	Poll ONE job_id as proxy

All jobs from the same question finish together — poll one as proxy. Status: queued → started → finished (or failed). If a job returns failed, report partial failure. Do NOT auto-retry.

Response Structure

Study creation: response["study"]["id"] (nested under study key)

Question responses (from GET /v1/research-studies/{id}/questions):

• response_text — the persona's answer (may contain HTML)
• agent_name, agent_age, agent_city, agent_state, agent_country
• agent_occupation, agent_summary

Share link: Prefer share_link field. If absent, use share_url.

Multi-Phase Diligence

For high-stakes investments, run iterative studies on the same group:

Phase	Purpose	Questions
1. Pain Discovery	Validate the problem exists	7 open-ended (standard framework)
2. Deep Dive	Understand requirements and trust boundaries	7 targeted (informed by Phase 1)
3. Concept Test	Describe the solution neutrally, test reactions	7 structured (still non-leading)

Reuse the same curated group. ~45 minutes total for all three phases. In Phase 3, describe the concept generically — "imagine a tool that does X" rather than naming the startup.

Cross-Study Patterns (from 50+ Studies)

These patterns appear consistently across diligence studies:

• Magic Wand Divergence: In 7/10 studies, founder's value prop diverged from Q6 answers. Flag this prominently when it occurs.
• Switching Inertia > Switching Cost: Personas cite "hassle" and "time" more than "money" as switching barriers.
• Trust by Industry: Healthcare = credentials, Finance = security certs, SaaS = free trial, CPG = ingredient transparency.
• Price ≠ Value Objection: When personas reject pricing, it's almost never the dollar amount — it's unclear value mapping.
• Authenticity Premium: Personas consistently reward brands that "feel real" over polished messaging.

Common Mistakes

• Mentioning the startup or VC in questions (violates the golden rule)
• Asking leading questions about the startup's features
• Skipping the thesis step (user must approve direction first)
• Using group-level removal instead of study-level for curation
• Passing agent_ids as strings instead of list[int]
• Dropping below 6 personas after curation
• Polling too aggressively (use 45-50s first, then 20s)
• Using size instead of group_size in recruitment
• Using full state names instead of 2-letter codes
• Including income filter (not supported — screen via Q1 instead)
• Asking yes/no pricing questions ("Would you pay $X?")
• Proceeding with 0-3 agents (results will not be meaningful)
• Not checking agent_summary alongside Q1 response during curation

Limitations

Ditto personas have NOT used the specific startup's product. For:

• Actual product UX feedback → use real beta testers
• Technical due diligence → use domain experts
• Financial model validation → use real unit economics
• Legal/regulatory assessment → use human specialists

Recommended hybrid: Ditto for customer demand validation (80% of insight in 18 minutes), then targeted human interviews for the remaining 20% requiring real product experience.

Further Reading

• Full API guide: https://askditto.io/claude-code-guide
• Question design: @question-playbook.md
• Over-recruit strategy: @advanced/over-recruit-and-curate.md