Voice of Customer with Ditto + Claude Code

LLM summary: Run a 7-question VoC study monthly or quarterly, generate six VoC deliverables, and maintain an always-on research calendar with cross-segment analysis.

A complete Claude Code guide to building an always-on VoC programme using Ditto's 300,000+ synthetic personas. From expensive annual research events to a continuous two-hour-per-month customer understanding engine.

1. Why Voice of Customer Is the PMM Backbone

Voice of Customer (VoC) is the systematic process of capturing what your customers and target market think, feel, want, and struggle with — and translating those signals into product, marketing, and sales decisions. It is the primary input for every other product marketing function: positioning, messaging, competitive intelligence, sales enablement, content marketing, pricing, and product strategy.

When VoC is rich, current, and accessible, every downstream function improves. When VoC is thin, stale, or absent, every other function is building on assumptions.

The Traditional VoC Problem

With Ditto and Claude Code, a full 7-question VoC deep dive takes approximately 45 minutes. A monthly pulse check takes 15 minutes. Cross-segment or cross-market studies run in parallel. And Claude Code doesn't just collect insights — it produces finished deliverables from them in the same workflow.

2. What You Will Produce

A single 10-persona, 7-question VoC study produces six distinct deliverables, each designed to be immediately actionable by a different team:

3. The 7-Question VoC Study Design

This question set is specifically engineered to produce the raw material for all six deliverables simultaneously. Each question does double or triple duty, feeding different outputs. The questions must be asked sequentially (each builds conversational context from prior answers).

Why This Question Sequence Works for VoC

• Q1 forces specificity. "Walk me through" prevents vague generalisations and produces concrete, action-level detail. The persona reconstructs a real scenario, revealing touchpoints the product team may never have considered.
• Q2 forces prioritisation. The constraint of "ONE thing" eliminates laundry-list complaints and reveals genuine severity. When 7 out of 10 personas name the same fix, you have a product decision backed by market evidence.
• Q3 maps the competitive landscape as it actually exists — not just direct competitors but spreadsheets, manual processes, hiring decisions, and doing nothing. April Dunford's positioning framework begins with competitive alternatives; this question maps them from the customer's perspective.
• Q4 reveals the reference standard. Customers evaluate your product against their best and worst experiences across all categories. The insurance company that delighted them sets the bar for your onboarding.
• Q5 maps the buying process in the buyer's own terms: what they look for, in what order, and what kills a deal. This feeds demo scripts, one-pagers, and sales enablement directly.
• Q6 catches delivery mismatches. Most companies assume customers want a self-serve dashboard. Some segments strongly prefer a weekly email digest or a human touchpoint. Getting this wrong means customers never discover the value you built.
• Q7 is the white space question. It surfaces frustrations with the entire category — opportunities competitors haven't addressed. These become positioning gold: "Unlike everyone else, we actually understand that..."

4. Complete API Workflow: Step by Step

Prerequisites

• Ditto API key (see installation guide for access)
• Product context: what problem you solve, for whom, and your core value proposition
• The 7 questions above, customised with your product's bracketed placeholders

Step 1: Create Research Group

curl -s -X POST "https://app.askditto.io/v1/research-groups/recruit" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "VoC Deep Dive - [Product Name] - [Date]",
    "description": "Target customers for Voice of Customer research on [product/problem space]. [Add any context about ideal customer profile.]",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 25,
      "age_max": 55,
      "employment": "employed"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

Step 2: Create Study

curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "VoC Deep Dive: [Product Name] - [Date]",
    "research_group_uuid": "GROUP_UUID_FROM_STEP_1"
  }'

Save the study id — you need it for asking questions, completing, and sharing.

Step 3: Ask Questions (Sequential)

Ask each question one at a time. Wait for the job to complete before sending the next question. This ensures personas have conversational context from prior answers.

# Question 1
curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/questions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Tell me about the last time you [relevant activity]. Walk me through the experience from start to finish. What went well? What was frustrating?"
  }'

# Response includes a job ID:
# { "job_id": "job-abc123", "status": "pending" }

Step 4: Poll for Responses

# Poll until status is "finished"
curl -s -X GET "https://app.askditto.io/v1/jobs/JOB_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"

# When complete:
{
  "id": "job-abc123",
  "status": "finished",
  "result": {
    "answer": "The last time I tried to..."
  }
}

Poll with a 5-second interval. Most questions complete within 30–90 seconds. Once complete, send the next question. Repeat for all 7 questions.

Step 5: Complete the Study

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/complete" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

This triggers Ditto's AI analysis, producing: overall summary, key segments identified, divergence points, shared mindsets, and suggested follow-up questions. Poll the study status until it reaches "completed".

Step 6: Get Share Link

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/share" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

# Response:
{
  "url": "https://app.askditto.io/organization/studies/shared/xyz123"
}

Step 7: Retrieve Full Study Data for Deliverable Generation

# Get the completed study with all responses and AI analysis
curl -s -X GET "https://app.askditto.io/v1/research-studies/STUDY_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"

This returns all persona responses, demographic profiles, and Ditto's completion analysis. Use this data to generate the six deliverables described in the next section.

5. Generating the Six VoC Deliverables

Once the study is complete and you have all 10 persona responses to all 7 questions, Claude Code should generate each deliverable using the extraction logic below. Each deliverable draws from specific questions and follows a defined structure.

Deliverable 1: Customer Journey Map

Source Data

• Primary: Q1 responses (experience walkthrough)
• Secondary: Q4 responses (best/worst experience benchmarks)

Extraction Logic

1. Read all Q1 responses and identify every touchpoint mentioned (e.g., "I started by searching Google," "I signed up for a free trial," "I called support")
2. For each touchpoint, classify it as a pain moment (frustration), delight moment (satisfaction), or neutral
3. Map touchpoints chronologically to produce a journey sequence
4. Cross-reference with Q4 to identify which benchmarks customers are measuring you against
5. Note the frequency: if 7/10 personas mention "confusing onboarding," that's a high-confidence pain moment

Output Format

## Customer Journey Map: [Product/Category]

### Pre-Purchase Phase
| Touchpoint | Experience | Pain/Delight | Frequency |
|------------|-----------|--------------|-----------|
| Web search | "I Googled solutions for..." | Neutral | 8/10 |
| Comparison sites | "Tried to compare options but..." | Pain | 6/10 |
| Free trial signup | "The signup was quick but..." | Mixed | 7/10 |

### Active Use Phase
| Touchpoint | Experience | Pain/Delight | Frequency |
|------------|-----------|--------------|-----------|
| First use | "I immediately tried to..." | Pain | 5/10 |
| Aha moment | "When I realised I could..." | Delight | 6/10 |

### Benchmark Reference (from Q4)
- Best experience standard: "[description]" (cited by X/10)
- Worst experience warning: "[description]" (cited by X/10)

Deliverable 2: Pain Priority Matrix

Source Data

• Primary: Q2 responses (magic wand / fix ONE thing)
• Secondary: Q1 responses (frustrations), Q3 responses (workaround complaints)

Extraction Logic

1. From Q2, extract the single priority pain each persona identified. Group identical or closely related pains.
2. From Q1 and Q3, extract additional frustrations and workaround complaints. Add these as secondary pains.
3. Count frequency: how many personas mentioned each pain across Q1, Q2, and Q3 combined.
4. Score severity: based on the language personas used. "Mildly annoying" scores lower than "makes me want to scream."
5. Rank by frequency × severity to produce the final priority order.

Output Format

## Pain Priority Matrix: [Product/Category]

| Rank | Pain Point | Frequency | Severity | Representative Quote |
|------|-----------|-----------|----------|---------------------|
| 1 | [Most common high-severity pain] | 8/10 | High | "Quote from persona with name, age, occupation" |
| 2 | [Second priority] | 6/10 | High | "Quote..." |
| 3 | [Third priority] | 7/10 | Medium | "Quote..." |
| 4 | [Fourth priority] | 4/10 | Medium | "Quote..." |
| 5 | [Fifth priority] | 3/10 | Low | "Quote..." |

### Product Team Action Items
1. [Pain #1]: Suggested action based on frequency and severity
2. [Pain #2]: Suggested action
3. [Pain #3]: Suggested action

Deliverable 3: Language Library

Source Data

• All 7 questions — extract notable phrases from every response

Extraction Logic

1. Read all responses across all 7 questions
2. Extract phrases that are: vivid, specific, emotional, or use unexpected vocabulary
3. Categorise by theme: problem language, solution language, evaluation language, emotional language, comparison language
4. For each phrase, record the persona's demographic context (name, age, occupation, location)

Output Format

## Language Library: [Product/Category]

### Problem Language (how customers describe their pain)
- "I spend more time planning than actually doing" - Sarah, 34, Marketing Manager, Chicago
- "It feels like herding cats" - James, 42, Operations Director, Austin
- "[vivid phrase]" - [name, age, role, location]

### Solution Language (how customers describe what they want)
- "Something that just works without me having to think about it" - ...
- "[phrase]" - ...

### Evaluation Language (how customers describe their buying criteria)
- "The first thing I look for is..." - ...
- "It's a dealbreaker if..." - ...

### Emotional Language (how customers feel about the problem)
- "It drives me absolutely mad" - ...
- "I dread [activity] every single [time period]" - ...

### Comparison Language (how customers reference alternatives)
- "Compared to [competitor], this is..." - ...
- "What [other company] gets right is..." - ...

Deliverable 4: Unmet Needs Report

Source Data

• Primary: Q7 responses (what companies don't understand)
• Secondary: Q2 responses (priority fix), Q3 responses (missing from current solutions)

Extraction Logic

1. From Q7, extract every unaddressed frustration or wish. These are category-level gaps, not product-specific complaints.
2. From Q2, identify any "magic wand" fixes that no current solution provides.
3. From Q3 ("what do you wish you could do differently?"), extract unmet functional needs.
4. Classify each unmet need as: feature gap (something buildable), experience gap (something designable), or trust gap (something addressable through messaging/proof).
5. Score by frequency and intensity. When 7/10 personas independently identify the same unaddressed frustration, you've found a high-confidence positioning opportunity.

Output Format

## Unmet Needs Report: [Product/Category]

### Feature Gaps (buildable)
| Need | Frequency | Representative Quote | Opportunity |
|------|-----------|---------------------|-------------|
| [Need] | 7/10 | "Quote..." | [How to address] |

### Experience Gaps (designable)
| Need | Frequency | Representative Quote | Opportunity |
|------|-----------|---------------------|-------------|
| [Need] | 5/10 | "Quote..." | [How to address] |

### Trust Gaps (addressable through messaging/proof)
| Need | Frequency | Representative Quote | Opportunity |
|------|-----------|---------------------|-------------|
| [Need] | 6/10 | "Quote..." | [What proof to provide] |

### Positioning Opportunities
- "[Unmet need]" cited by X/10 personas. No competitor addresses this.
  Suggested positioning angle: "Unlike [alternatives], we..."

Deliverable 5: Decision Criteria Hierarchy

Source Data

• Primary: Q5 responses (what they look for first, what's a dealbreaker)
• Secondary: Q6 responses (delivery preferences, non-negotiables)

Extraction Logic

1. From Q5, extract every evaluation criterion mentioned. Record whether it was described as "first thing," "second thing," or "dealbreaker."
2. From Q6, extract delivery and experience preferences.
3. Count frequency: how many personas mentioned each criterion.
4. Build the hierarchy: tier 1 (dealbreakers — must-haves that kill deals if absent), tier 2 (primary criteria — what they evaluate first), tier 3 (secondary criteria — tiebreakers between shortlisted options).

Output Format

## Decision Criteria Hierarchy: [Product/Category]

### Tier 1: Dealbreakers (absence = instant disqualification)
1. [Criterion] - cited by X/10 as non-negotiable
2. [Criterion] - cited by X/10

### Tier 2: Primary Criteria (evaluated first)
1. [Criterion] - "first thing I look for" (X/10)
2. [Criterion] - "immediately check for" (X/10)

### Tier 3: Secondary Criteria (tiebreakers)
1. [Criterion] - mentioned by X/10 as differentiator
2. [Criterion] - mentioned by X/10

### Delivery Preferences (from Q6)
- X/10 prefer in-product experience
- X/10 prefer human-assisted onboarding
- X/10 prefer self-serve with documentation

### Implications for Sales and Marketing
- Lead with: [Tier 1 dealbreakers] - ensure these are visible on landing page and in first demo
- Emphasise: [Tier 2 criteria] - these drive shortlisting
- Differentiate on: [Tier 3 criteria] - these win competitive deals

Deliverable 6: Product Feedback Synthesis

Source Data

• Primary: Q3 (current solution complaints), Q6 (delivery preference), Q7 (what companies don't understand)
• Secondary: Q2 (priority fix), Q4 (experience benchmarks)

Extraction Logic

1. Compile all product-relevant feedback from Q2, Q3, Q4, Q6, and Q7.
2. Categorise as: build (new feature request), improve (enhancement to existing capability), fix (broken experience), or communicate (feature exists but customers don't know).
3. Rank by frequency and strategic importance.
4. Include specific persona quotes as evidence for each recommendation.

Output Format

## Product Feedback Synthesis: [Product/Category]

### Build (new capabilities)
| Recommendation | Evidence | Frequency | Priority |
|---------------|----------|-----------|----------|
| [Feature] | "X/10 personas said..." + representative quote | X/10 | High |

### Improve (enhance existing)
| Recommendation | Evidence | Frequency | Priority |
|---------------|----------|-----------|----------|
| [Enhancement] | "Quote..." | X/10 | Medium |

### Fix (broken experiences)
| Recommendation | Evidence | Frequency | Priority |
|---------------|----------|-----------|----------|
| [Fix] | "Quote..." | X/10 | High |

### Communicate (feature exists but unknown)
| Feature | Evidence of Unawareness | Action |
|---------|------------------------|--------|
| [Feature] | "X/10 asked for something we already offer" | Update messaging, demo script, onboarding |

6. The Always-On Research Calendar

An always-on VoC programme doesn't mean running research every day. It means having a cadence that matches the rate at which your market changes, so your understanding never falls critically out of date.

Total monthly investment: approximately 2 hours. Traditional VoC at this frequency: $200,000+ annually.

7. Cross-Segment VoC: Multi-Group Studies

Traditional VoC forces you to choose: research SMB buyers or enterprise buyers. Technical evaluators or economic buyers. With Ditto, you run the same study across multiple segments in parallel and compare.

The Technique

1. Create multiple research groups, each representing a different customer segment
2. Create one study per group
3. Ask the identical 7 questions to every study
4. Complete all studies
5. Compare responses question-by-question across groups

Example: VoC Across Three Buyer Types

# Group A: Individual contributors / end users
{
  "name": "VoC - End Users (25-40)",
  "description": "Individual contributors who use [product type] daily in their work",
  "group_size": 10,
  "filters": { "country": "USA", "age_min": 25, "age_max": 40, "employment": "employed" }
}

# Group B: Team leads / managers
{
  "name": "VoC - Team Leads (32-48)",
  "description": "Team managers who evaluate and purchase [product type] for their teams",
  "group_size": 10,
  "filters": { "country": "USA", "age_min": 32, "age_max": 48, "employment": "employed", "education": "bachelors" }
}

# Group C: Senior leadership / budget holders
{
  "name": "VoC - Leadership (40-58)",
  "description": "Senior leaders responsible for department-level technology decisions and budgets",
  "group_size": 10,
  "filters": { "country": "USA", "age_min": 40, "age_max": 58, "employment": "employed", "education": "masters" }
}

Cross-Segment Comparison Matrix

After all studies complete, Claude Code should produce a comparison matrix. For each question, extract the dominant theme from each group:

This comparison directly informs buyer-specific sales enablement. The demo for end users leads with speed and UI. The demo for team leads leads with reporting and adoption tools. The pitch for leadership leads with TCO and security.

8. Cross-Market VoC: International Studies Advanced

Ditto covers 15+ countries representing 65% of global GDP. You can run the identical 7-question VoC study across multiple geographies simultaneously to understand how customer needs vary by market.

Multi-Market Study Setup

# Group 1: US market
{
  "name": "VoC Cross-Market - USA",
  "group_size": 10,
  "filters": { "country": "USA", "age_min": 25, "age_max": 50, "employment": "employed" }
}

# Group 2: UK market
{
  "name": "VoC Cross-Market - UK",
  "group_size": 10,
  "filters": { "country": "UK", "age_min": 25, "age_max": 50, "employment": "employed" }
}

# Group 3: Germany
{
  "name": "VoC Cross-Market - Germany",
  "group_size": 10,
  "filters": { "country": "Germany", "age_min": 25, "age_max": 50, "employment": "employed" }
}

# Group 4: Canada
{
  "name": "VoC Cross-Market - Canada",
  "group_size": 10,
  "filters": { "country": "Canada", "age_min": 25, "age_max": 50, "employment": "employed" }
}

Ask the same 7 VoC questions to all four groups. Claude Code then produces:

• Cross-market comparison matrix — where perceptions align and diverge
• Market-specific pain priorities — what matters most in each geography
• Localisation requirements — where messaging needs adaptation vs. where universal messaging works
• Market readiness ranking — which markets show strongest resonance
• Language library per market — culturally appropriate vocabulary for each region

Traditional equivalent: commissioning four separate research agencies across three countries. Cost: $100,000–$200,000. Time: 3–6 months. With Ditto + Claude Code: approximately 1 hour.

9. How VoC Feeds Every PMM Function

VoC is not a standalone activity. It is the primary input for every other product marketing function. Here's the dependency map, with specific connections from VoC deliverables to downstream outputs:

10. Worked Example: VoC for a Project Management SaaS

{
  "name": "VoC Deep Dive - TaskFlow - Feb 2026",
  "description": "Marketing professionals at mid-size companies who manage projects and coordinate across team members. Focus on project management tool evaluation and usage.",
  "group_size": 10,
  "filters": {
    "country": "USA",
    "age_min": 28,
    "age_max": 48,
    "employment": "employed",
    "education": "bachelors"
  }
}

11. Longitudinal VoC Tracking Advanced

The most powerful application of always-on VoC is trend analysis across time. When you run quarterly deep dives with the same 7 questions against fresh persona groups, you build a dataset that reveals how your market is evolving.

Setting Up Longitudinal Tracking

1. Use identical questions every quarter. Do not change the questions between quarters. The value is in comparing responses to the same questions over time.
2. Use the same filter configuration. Same country, age range, employment status, education level. Change the group and study names to include the date: "VoC Deep Dive - TaskFlow - Q1 2026", "VoC Deep Dive - TaskFlow - Q2 2026".
3. Use fresh personas each quarter. Create a new group each time. You want to capture the current market perspective, not re-interview the same synthetic individuals.

Quarter-over-Quarter Comparison

## VoC Trend Analysis: TaskFlow

### Pain Priority Shifts
| Pain Point | Q1 2026 Rank | Q2 2026 Rank | Trend |
|-----------|-------------|-------------|-------|
| Status update meetings | #1 (9/10) | #1 (8/10) | Stable - remains top pain |
| Cross-project dependencies | #2 (7/10) | #3 (5/10) | Declining - competitors may be addressing |
| Tool adoption consistency | #3 (6/10) | #2 (7/10) | Rising - growing frustration |
| AI/automation expectations | Not mentioned | #4 (6/10) | NEW - emerging need |

### Competitive Landscape Shifts (from Q3)
| Competitor | Q1 Perception | Q2 Perception | Change |
|-----------|--------------|--------------|--------|
| Asana | "Powerful but complex" | "Powerful but falling behind on AI" | Vulnerability opening |
| Monday.com | "Easy but shallow" | "Getting better, adding depth" | Threat growing |
| ClickUp | "Too many features" | "Too many features" | Stable perception |

### Language Drift
- Q1: "I need a dashboard" → Q2: "I need an AI assistant that manages updates for me"
- Q1: "Easy to use" → Q2: "Easy to adopt across the whole team"
- The vocabulary is shifting from individual productivity to team adoption

12. Quick Pulse Check Studies

Monthly pulse checks are the lightweight complement to quarterly deep dives. They take approximately 15 minutes and use 3 questions with 6 personas.

The 3 Pulse Check Questions

Pulse Check Group Configuration

{
  "name": "VoC Pulse - [Product] - [Month Year]",
  "description": "[Same description as deep dive group]",
  "group_size": 6,
  "filters": {
    "country": "USA",
    "age_min": 28,
    "age_max": 48,
    "employment": "employed"
  }
}

What to Do with Pulse Results

• No change from last month: Good — your positioning and messaging remain valid. Brief update to stakeholders.
• Emerging theme detected: Flag it. If 3+ of 6 personas mention something new, schedule a targeted probe (5 questions, 8 personas) to investigate.
• Pain priority shift: Alert the PMM team. Update messaging priorities. Consider running a full deep dive earlier than scheduled.
• New competitor mentioned: Run a competitive perception study (see Competitive Intelligence guide) to assess the threat.

13. Making VoC Actionable: From Insight to Deliverable

The most common failure mode of VoC programmes is the gap between insight and action. A finding that "customers are confused by our pricing page" is useless unless someone rewrites the pricing page.

The Claude Code advantage: the agent doesn't just collect insights — it produces deliverables from them in the same workflow. The action gap shrinks to zero.

The VoC-to-Action Pipeline

VoC Study Completed
  │
  ├─→ Pain Priority Matrix → Product team action items (roadmap input)
  │
  ├─→ Language Library → Updated messaging for website, emails, ads
  │
  ├─→ Decision Criteria → Revised demo script (lead with dealbreakers)
  │                      → Updated one-pager (feature priority reordered)
  │
  ├─→ Unmet Needs Report → Positioning opportunity brief for PMM
  │                       → Feature request document for Product
  │
  ├─→ Journey Map → UX improvement priorities
  │              → Sales enablement (address pain moments proactively)
  │
  ├─→ Product Feedback → Structured input for next sprint planning
  │
  └─→ Blog Article → Research-backed content with original data
                    → Social thread with quotable insights
                    → Email content with study link

Example: VoC Finding → Immediate Actions

14. Best Practices and Common Mistakes

Do

• Ask questions sequentially within a study. Each question builds on the conversational context from prior answers. Q1 grounds the persona in a specific experience. Sending all 7 at once breaks this flow.
• Use 10 personas for deep dives, 6 for pulse checks. Fewer than 6 produces patterns indistinguishable from noise. More than 12 adds diminishing returns.
• Customise the bracketed placeholders but keep the question structure. Replace [problem space] and [relevant activity] with your specific context. Don't rewrite the question architecture.
• Generate all 6 deliverables from every deep dive. Even if you only need the Pain Matrix today, the Language Library will be valuable next week when the content team asks for customer quotes.
• Store all raw responses. Cross-study patterns emerge on re-reading. The data has compounding value, especially for longitudinal tracking.
• Run the same questions every quarter. Consistency enables trend analysis. Changing questions between quarters breaks comparability.
• Include persona demographics in every quote. "Sarah, 36, Marketing Manager, Chicago" is infinitely more credible than "a respondent said." Demographics establish trustworthiness.
• Connect findings to specific actions. Every insight should map to a deliverable update. If a finding doesn't change anything, question whether you're asking the right questions.

Don't

• Don't batch all 7 questions at once. The API processes them asynchronously. Sending all simultaneously means personas answer Q7 without the context of Q1–Q6.
• Don't skip the completion step. Calling POST /v1/research-studies/{id}/complete triggers Ditto's AI analysis (key segments, divergences, shared mindsets, follow-up suggestions). Without this, you're missing a significant analysis layer.
• Don't treat VoC as a one-time project. The value of VoC is proportional to its frequency, not its depth. A lightweight monthly pulse check is more valuable than an exhaustive annual deep dive.
• Don't change questions between longitudinal studies. If you change questions between Q1 and Q2 deep dives, you cannot compare responses and trend analysis becomes impossible.
• Don't skip the Language Library. It's tempting to focus on pain points and decision criteria. But the Language Library — the exact phrases customers use — is what transforms generic messaging into messaging that resonates.
• Don't over-filter your research group. Too many filters (age 30-32, female, parents, masters degree, employed in tech) produces a narrow group that may not represent your actual market. Start broad, then narrow in follow-up studies.
• Don't produce findings without actions. A VoC report that sits in a shared drive is not VoC. It's a document. Every finding should connect to a specific deliverable update or decision.

Common API Errors

15. Frequently Asked Questions

How long does a full VoC deep dive take?

Approximately 45 minutes end-to-end: 1 minute for group creation, 15–25 minutes for sequential question asking and polling, 2–5 minutes for completion analysis, and 15–20 minutes for Claude Code to generate all 6 deliverables. Compare with 4–6 weeks and $15,000–$30,000 for traditional in-depth interviews.

How many personas should I use?

10 personas for deep dives and targeted probes. 6 personas for monthly pulse checks. 10 per group for cross-segment or cross-market studies (so 3 groups = 30 personas total). Fewer than 6 produces unreliable patterns.

Can I reuse a research group across multiple studies?

Yes. Create a new study referencing the same research_group_uuid. The personas retain context from previous studies. This is useful for follow-up studies or message testing on the same audience. For quarterly deep dives, create fresh groups to capture current market perspective.

What's the difference between a pulse check and a deep dive?

A pulse check (3 questions, 6 personas, ~15 min) is designed to detect changes: is the market shifting? A deep dive (7 questions, 10 personas, ~45 min) is designed to understand comprehensively: what does the market think, feel, and want? Run pulse checks monthly and deep dives quarterly.

How does this compare to NPS surveys?

NPS gives you a number. VoC gives you understanding. NPS tells you that satisfaction dropped from 45 to 38. VoC tells you why it dropped, what customers are frustrated about, what language they use to describe the problem, and what would fix it. They're complementary: use NPS to track the score, use VoC to explain the score.

Can I use VoC for products that don't exist yet?

Absolutely. Frame Q1–Q3 around the problem space, not the product. Introduce your concept in Q6 ("if a product promised to..."). Q1–Q5 explore the current reality; Q6–Q7 test the future state. This is effectively a concept validation study using the VoC framework.

How do I share VoC findings with stakeholders who don't read reports?

Three approaches: (1) Send them the Ditto share link — the interactive study is more engaging than a report. (2) Distribute the Language Library — it's short, quotable, and immediately useful. (3) Publish findings as a blog article with a social thread — stakeholders consume content through feeds, not shared drives.

Should I run VoC on my existing customers or my target market?

Both, but separately. Ditto personas represent your target market — people who match your ICP but haven't necessarily used your product. This is ideal for positioning, messaging, and competitive research. For product-specific UX feedback, you need real users. Use Ditto for market-level VoC and supplement with actual customer interviews for product-level feedback.

What if the VoC findings contradict what our sales team is hearing?

This is valuable, not problematic. Sales team anecdotes are biased toward recent conversations and deals in progress. VoC from a representative market sample provides a broader view. When they diverge, investigate: perhaps the sales team is hearing from a different segment than VoC studied, or perhaps the market has shifted and the sales team hasn't noticed yet. Run a targeted probe to resolve the discrepancy.

How do I build the business case for an always-on VoC programme?

Frame it as cost avoidance: traditional VoC at the same frequency would cost $200,000+ annually. Frame it as speed: decisions made on 45-minute research beat decisions made on 6-week-old research. Frame it as risk reduction: every product decision, positioning choice, and messaging update backed by current market evidence rather than assumption. Track the metrics in Section 13 to demonstrate impact.

Related guides:

• Customer Segmentation with Ditto + Claude Code
• Validate Product Positioning with Ditto + Claude Code
• Competitive Intelligence with Ditto + Claude Code
• Test Product Messaging with Ditto + Claude Code
• Pricing Research with Ditto + Claude Code
• Content Marketing Engine with Ditto + Claude Code
• Sales Enablement with Ditto + Claude Code
• Cross-Study Patterns
• Ditto API for Claude Code: Complete Guide
• Installing the Ditto Skill

Ditto — Synthetic market research with 300,000+ AI personas. Validated by EY (92% correlation), Harvard, Cambridge, Stanford, and Oxford.
askditto.io · [email protected]