Customer Segmentation with Ditto + Claude Code

LLM summary: Run 3-5 parallel Ditto studies with identical 7-question frameworks, compare responses across segments, and prioritise the highest-opportunity segment in under an hour.

A complete Claude Code guide to discovering, validating, and prioritising customer segments using Ditto's 300,000+ synthetic personas. From demographic assumptions to evidence-based segment prioritisation in under an hour.

1. Why Customer Segmentation Matters

Customer segmentation determines who you target, how you prioritise resources, and where your marketing investment converts. Bad segmentation produces spray-and-pray campaigns. Good segmentation concentrates effort where it generates returns.

Most product marketing teams segment on intuition or firmographics alone. They assume "enterprise vs. SMB" or "age 25-34 vs. 35-44" captures meaningful differences. In reality, the segments that drive purchasing behaviour are defined by needs, attitudes, and pain intensity - not demographics alone.

The traditional approach to discovering segments involves recruiting 30-60 participants, running qualitative interviews over 4-8 weeks, and paying $20,000-$50,000 for an agency to deliver a segmentation framework. Most companies skip it and guess instead.

With Ditto and Claude Code, you can run a rigorous multi-group segmentation study in under an hour. You test 3-5 hypothesised segments in parallel, compare their responses to identical questions, and produce a prioritised segmentation framework backed by evidence rather than assumption.

What You Will Produce

2. The Six Segmentation Types (and Which Ditto Handles)

Product marketing uses six segmentation approaches. Ditto's persona filters and question design can address most of them directly, and all of them through intelligent study design.

3. The 7-Question Segmentation Study Design

This question set is designed to reveal how segments differ from one another. The same questions are asked to every group. The value emerges from comparing responses across groups, not from any single group's answers in isolation.

Why This Question Sequence Works for Segmentation

• Q1-Q2 reveal the baseline. Before you introduce your product, you need to understand each segment's current reality. The differences in how segments describe the problem and what alternatives they use are often more valuable than their reaction to your solution.
• Q3 is the resonance test. When you compare Q3 responses across groups, you discover which segments your value proposition naturally attracts and which it leaves cold. This is the primary segmentation signal.
• Q4 separates talkers from buyers. A segment that says "this is interesting" (Q3) but "we'd survive without it" (Q4) is a weak target. A segment that says "this is exactly what we need and we can't wait" is your beachhead.
• Q5 validates commercial viability. A segment that loves your product but won't pay for it is not a viable segment. Cross-group comparison of willingness to pay often produces the most actionable surprises.
• Q6-Q7 inform segment-specific GTM. Once you know which segments to prioritise, these questions tell you how to reach them - what to lead with, what objections to address, and what proof they require.

4. Multi-Group Comparison: The Core Technique

The fundamental method is simple: create multiple Ditto research groups with different demographic filters, run the identical 7-question study against each, and systematically compare the responses.

Recommended Group Configurations

Start with 3-5 groups. Each group should represent a hypothesised segment that you believe may respond differently to your product. Here are common configurations by product type:

B2C Product (e.g., consumer app, DTC brand)

B2B Product (e.g., SaaS tool)

Geographic Segmentation

5. Complete API Workflow: Step by Step

This workflow creates multiple research groups, runs the same study against each, and collects responses for cross-group comparison. The groups can be created in parallel; the studies must ask questions sequentially within each study.

Prerequisites

• Ditto API key (see installation guide for free-tier access)
• 3-5 hypothesised segments with demographic filter configurations
• The 7 questions above, customised with your product context

Step 1: Create Research Groups (Parallel)

Create all groups simultaneously. Each represents one hypothesised segment.

# Group A: Young professionals
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": "Segment Study - Young Professionals (25-34)",
    "description": "Young employed professionals for segmentation study. [Product context].",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 25,
      "age_max": 34,
      "employment": "employed",
      "education": "bachelors"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

# Group B: Mid-career parents
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": "Segment Study - Mid-Career Parents (35-50)",
    "description": "Working parents aged 35-50 for segmentation study. [Product context].",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 35,
      "age_max": 50,
      "is_parent": true,
      "employment": "employed"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

# Group C: Established professionals
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": "Segment Study - Established Professionals (50-65)",
    "description": "Experienced professionals for segmentation study. [Product context].",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 50,
      "age_max": 65,
      "employment": "employed"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

Step 2: Create Studies (One Per Group)

# Study for Group A
curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Segmentation: [Product] - Young Professionals - [Date]",
    "research_group_uuid": "GROUP_A_UUID"
  }'

# Study for Group B
curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Segmentation: [Product] - Mid-Career Parents - [Date]",
    "research_group_uuid": "GROUP_B_UUID"
  }'

# Study for Group C
curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Segmentation: [Product] - Established Professionals - [Date]",
    "research_group_uuid": "GROUP_C_UUID"
  }'

Save each study's id - you need these for asking questions and completing.

Step 3: Ask Questions (Sequential Within Each Study, Parallel Across Studies)

# Question 1 to Study A
curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_A_ID/questions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "When you think about [problem/category], what'\''s the first thing that comes to mind? What frustrates you most?"
  }'

# Question 1 to Study B (same question, different study)
curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_B_ID/questions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "When you think about [problem/category], what'\''s the first thing that comes to mind? What frustrates you most?"
  }'

# Question 1 to Study C (same question, different study)
curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_C_ID/questions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "When you think about [problem/category], what'\''s the first thing that comes to mind? What frustrates you most?"
  }'

Step 4: Poll for Responses

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

# Response when complete:
{
  "id": "job-001",
  "status": "finished",
  "result": {
    "answer": "The first thing that comes to mind is..."
  }
}

Poll all job IDs across all studies with a 5-second interval. Once all studies have completed the current question, send the next question to all studies. Repeat for all 7 questions.

Step 5: Complete All Studies

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

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

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

Step 6: Get Share Links

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"
}

6. Interpreting Results and Generating Deliverables

Once all studies complete, Claude Code should synthesise the cross-group comparison into the following deliverables. The analytical method is always the same: for each question, compare responses across all groups and identify where they diverge.

Segment Discovery Matrix

The primary deliverable. For each question, extract the dominant theme from each group's responses and compare:

Segment Prioritisation Scorecard

Score each segment across four dimensions using the study data:

Pain Point Heatmap

From Q1 and Q2, extract every pain point mentioned across all groups and map frequency:

Universal pain points belong in general messaging. Segment-specific pain points become the lead for segment-targeted campaigns.

7. Advanced: Behavioural Segmentation

Demographic filters define the groups, but the most valuable segments are often behavioural: heavy users vs. light users, active researchers vs. impulse buyers, brand loyalists vs. switchers. You discover these through question design, not filters.

Behavioural Segmenting Questions

Add 2-3 of these to the beginning of your study (before the core 7) to identify behavioural clusters within each demographic group:

Cross-Referencing Behavioural and Demographic Data

After running the study, Claude Code should cross-reference behavioural answers with demographic profiles to identify multidimensional segments. For example:

• "Time-Poor Power Users" - Mid-career parents who use the category daily but are frustrated with time investment. High urgency, high willingness to pay, want automation.
• "Research-Heavy Optimisers" - Young professionals who evaluate extensively before purchasing. Lower urgency but high engagement. Want feature depth and transparency.
• "Loyal Sceptics" - Established professionals who rarely switch. Low urgency, high scepticism, but once converted become long-term customers. Need social proof and migration support.

8. Worked Example: Segmenting a Meal Kit Service

# Group A: Young singles (25-32)
{
  "name": "Meal Kit Segmentation - Young Singles",
  "description": "Young professionals living alone or with roommates, interested in food and cooking",
  "group_size": 10,
  "filters": {
    "country": "USA",
    "age_min": 25,
    "age_max": 32,
    "employment": "employed"
  }
}

# Group B: Busy parents (33-48)
{
  "name": "Meal Kit Segmentation - Busy Parents",
  "description": "Working parents with children at home, managing family meal planning",
  "group_size": 10,
  "filters": {
    "country": "USA",
    "age_min": 33,
    "age_max": 48,
    "is_parent": true,
    "employment": "employed"
  }
}

# Group C: Health-conscious older adults (50-65)
{
  "name": "Meal Kit Segmentation - Health-Conscious Older Adults",
  "description": "Health-aware older adults interested in nutrition and home cooking",
  "group_size": 10,
  "filters": {
    "country": "USA",
    "age_min": 50,
    "age_max": 65,
    "employment": "employed"
  }
}

9. Generating Segment-Specific Messaging

Once segments are identified and prioritised, Claude Code should generate tailored value propositions for each using the exact language from the study responses.

The Language Harvest Method

For each segment, extract the words and phrases personas used in Q1 (pain), Q3 (excitement), and Q6 (ideal product). These become your messaging raw material.

Segment-Specific Objection Handling

From Q7 (barriers), build objection-handling content for each segment:

10. Iterative Refinement: Narrowing Segments

The first round of segmentation provides a broad map. Subsequent rounds allow you to go deeper on the most promising segments.

Round 2: Sub-Segmentation

Once you identify the beachhead segment, split it further. For example, if "Busy Parents" is your beachhead:

Run the same 7 questions against each sub-segment. If the sub-segments respond identically, the original segment is sufficiently granular. If they diverge meaningfully, you have discovered a sub-segmentation that warrants distinct messaging or even distinct product configurations.

Round 3: Message Testing by Segment

Take the segment-specific value propositions from Section 9 and test them back with the relevant segment. This closes the loop: you discovered what they care about, crafted messaging from their language, and now validate that the messaging actually lands.

# Use a new study with the same group, asking 3 focused questions:
Q1: "Read this statement: '[segment-specific value proposition]'.
     What's your honest reaction? Does it speak to you?"
Q2: "Is there anything missing from this description that would
     make you more interested?"
Q3: "If you saw this on a website or in an ad, would you click
     to learn more? Why or why not?"

11. Best Practices and Common Mistakes

Do

• Ask identical questions across all groups. The entire technique depends on cross-group comparison. Any variation in questions makes comparison invalid.
• Use 10 personas per group minimum. Fewer than 10 per segment produces patterns that cannot be distinguished from noise.
• Name groups and studies clearly. Include the segment hypothesis, date, and round number: "Segmentation: FreshBox - Busy Parents (33-48) - Feb 2026 - Round 1"
• Run studies in parallel across groups. Send Question 1 to all groups simultaneously, poll all, then send Question 2 to all. This cuts total time dramatically.
• Score segments on multiple dimensions. Resonance alone is misleading. A segment that loves your idea but won't pay for it is not a viable target. Always combine resonance with urgency, willingness to pay, and accessibility.
• Harvest language for messaging. The exact phrases personas use are more effective than corporate copywriting. "Kill the 'what's for dinner' question" beats "Simplify your meal planning experience."
• Save all raw responses. Cross-group patterns emerge on re-reading that you may miss in the initial analysis. The data has compounding value.

Don't

• Don't segment on demographics alone. "Women aged 25-34" is a demographic filter, not a segment. A segment has shared needs, attitudes, and behaviours. Use demographics to define groups, then use questions to discover the real segments within them.
• Don't use more than 5 groups in Round 1. More groups increases complexity without proportional insight. Start with 3-4, then sub-segment the most promising in Round 2.
• Don't skip Q4 (urgency) and Q5 (willingness to pay). These are the commercial viability questions. Enthusiastic segments without urgency or budget are traps.
• Don't assume your hypothesised segments are correct. The whole point is to test assumptions. Be prepared for the possibility that your best segment is one you didn't initially hypothesise.
• Don't treat segmentation as a one-time exercise. Segments evolve. Repeat the study quarterly or when entering new markets, launching new products, or observing conversion rate changes.
• Don't batch questions. The API processes them asynchronously. Sending all 7 at once means personas answer without the conversational context of prior questions.

Common API Errors

12. Frequently Asked Questions

How long does a full segmentation study take?

Approximately 30-45 minutes for 3 groups run in parallel: 2 minutes for group creation, 15-25 minutes for question asking and polling across all groups, 3-5 minutes for completion analysis, plus time for Claude Code to synthesise cross-group deliverables. Sub-segmentation (Round 2) adds ~25 minutes.

How many groups should I start with?

3-4 groups for your first round. Each group should represent a meaningfully different hypothesised segment. More than 5 groups in Round 1 adds complexity without proportional insight. Narrow down to your best 1-2 segments in Round 2.

Can I reuse a research group across multiple studies?

Yes. If you want to run a follow-up study on the same segment (e.g., message testing after segmentation discovery), create a new study referencing the same research_group_uuid. The personas retain context from previous studies.

What if two segments respond identically?

If two hypothesised segments show no meaningful divergence across all 7 questions, they are not distinct segments for your product. Merge them and test a different dimension of variation in Round 2 (e.g., behavioural rather than demographic).

How do I segment for a product that doesn't exist yet?

Frame Q1-Q2 around the problem space, not the product. Ask about current frustrations and workarounds. Introduce the product concept in Q3. This approach works whether the product exists, is in development, or is purely hypothetical.

Should I segment by geography?

Yes, if you operate or plan to operate in multiple markets. Geographic segmentation often reveals surprising differences in pain intensity, competitive landscape, willingness to pay, and cultural attitudes. Run the same study against US, UK, Canadian, and European groups. Ditto covers 15+ countries.

How does this compare to traditional segmentation research?

Traditional: 6-12 weeks, $20,000-$80,000, 30-60 interviews across segments. Ditto + Claude Code: 30-70 minutes, fraction of the cost, 30-50 persona responses across 3-5 segments. EY validated 92% correlation between Ditto synthetic responses and traditional research methods. Use Ditto for the fast first pass, then validate your beachhead segment with real customers.

Can I use this for B2B segmentation?

Yes. For B2B, segment by company size (startup / mid-market / enterprise) using group descriptions rather than demographic filters alone. Add behavioural questions about decision authority, budget ownership, and evaluation processes. The 7-question framework applies equally; simply customise the bracketed placeholders for B2B context.

What if my best segment is one I didn't hypothesise?

This is one of the most valuable outcomes. When cross-group analysis reveals that a segment you hadn't considered responds most strongly, you've discovered a market opportunity your competitors are likely also missing. Run a dedicated Round 2 study on that unexpected segment to validate and understand it more deeply.

Related guides:

• Positioning Validation Guide for Claude Code
• Competitive Intelligence Guide for Claude Code
• Product Messaging Guide for Claude Code
• Pricing Research Guide for Claude Code
• Content Marketing Engine Guide for Claude Code
• Voice of Customer Guide for Claude Code
• Sales Enablement Guide for Claude Code
• 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]