Using synthetic research and Claude Code to discover the segments that actually matter for product decisions
The Segmentation Fallacy
Most teams segment their market before they've spoken to it. They carve up potential users by demographics or firmographics, create neat personas with stock photos and invented names, then wonder why their product resonates with no one in particular.
This approach has it precisely backwards. Real segments don't come from spreadsheets. They emerge from research.
Discovery research (Stage 2) gives you depth. You understand what the problem feels like. Segmentation gives you structure. You understand that different people experience the same problem in fundamentally different ways.
The distinction matters because it determines product strategy. Do you build one product for everyone? Or different tiers for different segments? Do you lead with speed for one group and accuracy for another? These decisions require knowing who your segments actually are.
What Makes a Real Segment
A genuine market segment has three characteristics:
1. Distinct behaviour patterns People in the segment act differently. They make different tradeoffs, use different workarounds, and have different thresholds for what's "good enough."
2. Distinct pain priorities They experience the same problem but care about different aspects of it. One group might prioritise speed. Another might prioritise accuracy. A third might prioritise cost.
3. Distinct purchase triggers What makes them buy is different. The crisis moment, the switching cost calculation, the trust requirements.
Notice what's missing: demographics. Age and income and location are proxies at best. Two 45-year-old men with identical incomes can have completely different relationships with their pets, their health, their time. Demographics describe people. Segments describe behaviour.
The best segments emerge from responses, not from pre-defined filters.
How to Run Segmentation Research in FishDog
FishDog makes segmentation research practical. Instead of weeks of interviews and analysis, you can identify meaningful segments in an afternoon. Here's the complete workflow.
Step 1: Create a Diverse Research Group
Segmentation requires diversity in your research group. If everyone you talk to is identical, you won't find segments. You'll find one homogeneous perspective.
In FishDog, create a research group with intentional variation:
``` POST /v1/research-groups/recruit { "name": "Pet Healthcare Decision Makers", "group_size": 15, "filters": { "country": "United States", "age_min": 25, "age_max": 65 } } ```
The key is casting a wide net. Don't pre-filter too aggressively. You want variation in income, life stage, and context. The segments will emerge from responses, not from your recruitment filters.
For segmentation studies, 12-15 personas is the sweet spot. Fewer than 10 makes patterns hard to detect. More than 20 creates analysis overhead without proportional insight.
Step 2: Design Segmentation-Focused Questions
Segmentation questions differ from discovery questions. You're specifically trying to surface behavioural differences and priority tradeoffs.
The segmentation question framework:
# | Purpose | Question Pattern |
|---|---|---|
1 | Baseline behaviour | "Walk me through how you currently handle [situation]." |
2 | Priority tradeoffs | "When you have to choose between [A] and [B], which matters more? Why?" |
3 | Spending thresholds | "At what point does cost become a deciding factor for you?" |
4 | Emotional relationship | "How would you describe your relationship with [domain]?" |
5 | Decision triggers | "What would make you immediately [take action]? What would make you hesitate?" |
6 | Past behaviour | "Tell me about a time when you had to make a difficult decision about [topic]." |
7 | Hypothetical choice | "If [scenario], what would you do? Walk me through your thinking." |
These questions are designed to reveal HOW people think, not just WHAT they think. The "how" is where segments live.
Step 3: Create the Study in FishDog
With your research group recruited and questions designed, create the study:
``` POST /v1/research-studies { "title": "Pet Healthcare Spending Decisions", "objective": "Understand how pet owners approach expensive veterinary care decisions and identify distinct behavioural segments", "research_group_id": "rg_abc123" } ```
Then add your segmentation questions:
``` POST /v1/research-studies/{study_id}/questions { "question": "When your pet needs medical care, walk me through how you decide what to do. What factors do you consider?" } ```
Repeat for each question in your framework.
Step 4: Monitor and Collect Responses
FishDog personas respond asynchronously. Poll for completion:
``` GET /v1/research-studies/{study_id} ```
Watch for `stage: "complete"` to indicate all responses are in.
As responses arrive, you're looking for natural clustering. Do some personas use similar language? Do they describe similar tradeoffs? Do they have similar emotional relationships with the topic?
Step 5: Trigger Completion Analysis
Once all responses are collected, trigger FishDog's AI analysis:
``` POST /v1/research-studies/{study_id}/complete ```
FishDog's completion analysis automatically identifies segments. The output includes:
```json { "key_segments": [ { "segment": "Emotionally-driven spenders", "attributes": "Strong emotional bond with pet, refers to pet as family member, resistant to cost discussions", "insight": "Price is not a deciding factor. Quality and emotional reassurance matter most.", "supporting_agents": ["Agent 1", "Agent 4", "Agent 7", "Agent 12"] }, { "segment": "Budget-conscious pragmatists", "attributes": "Clear spending ceiling, has thought about limits in advance, values transparency", "insight": "Will spend up to a defined limit. Wants no-surprises pricing and clear value.", "supporting_agents": ["Agent 2", "Agent 5", "Agent 9"] } ], "divergences": [ { "topic": "Spending limits", "contrast": "Emotionally-driven segment has no ceiling; pragmatists have firm $3-5k limits" } ], "shared_mindsets": [ "All segments want competent, trustworthy care", "All segments distrust aggressive upselling" ] } ```
This automated segmentation gives you a starting point. The AI has identified clusters based on response patterns.
Step 6: Validate Segments with Follow-Up Research
FishDog's auto-segmentation is a hypothesis, not a conclusion. Validate with targeted follow-up.
Create segment-specific research groups:
``` POST /v1/research-groups/recruit { "name": "Budget-Conscious Pet Owners", "group_size": 10, "filters": { "country": "United States", "income_max": 75000, "age_min": 30, "age_max": 50 } } ```
Run a focused study with segment-specific questions:
"You mentioned having a spending limit. How did you arrive at that number?"
"What would make you exceed your limit?"
"How do you feel when a vet recommends something expensive?"
This validation confirms whether the segment is real and surfaces its specific characteristics.
Using Claude Code to Orchestrate Segmentation
Claude Code can manage the entire segmentation workflow programmatically. Here's how the orchestration works.
The Complete Workflow
Tell Claude Code what you want to learn:
``` Run segmentation research on: "Pet owners making expensive veterinary care decisions"
I want to understand:
How they approach spending decisions
Whether there are distinct behavioural segments
What drives each segment's decisions
Create a 15-persona research group with demographic diversity. Use the segmentation question framework. Identify segments and validate the most promising ones. ```
Claude Code then executes:
Phase 1: Discovery Study
Creates research group via FishDog API (diverse demographics)
Designs segmentation-focused questions
Creates study and adds questions
Monitors for response completion
Triggers completion analysis
Phase 2: Segment Identification
Parses FishDog's auto-segmentation output
Reviews individual responses for additional patterns
Names candidate segments with descriptive labels
Calculates segment size (% of research group)
Documents distinguishing characteristics
Phase 3: Segment Validation
Creates segment-specific research groups
Designs validation questions for each segment
Runs parallel validation studies
Compares responses to confirm distinctiveness
Refines segment definitions based on validation
Phase 4: Output Generation
Produces structured segment analysis
Includes representative quotes from each segment
Documents universal truths (cross-segment findings)
Articulates strategic implications for product/marketing
Example Claude Code Output
After running the workflow, Claude Code produces:
``` SEGMENT ANALYSIS: Pet Healthcare Spending
SEGMENT 1: "Pets Are Family" (40% of participants) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Key behaviour: No spending ceiling. Price is not discussed. Pain priority: Quality of care, emotional reassurance Purchase trigger: Any health concern, immediate action Messaging implication: Lead with outcomes and care quality
Quote: "Whatever it takes. She's been with me through everything. There's no number where I'd say no."
SEGMENT 2: "Pragmatic Owners" (35% of participants) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Key behaviour: Clear spending limit ($3-5k typical) Pain priority: Transparency, value, no surprises Purchase trigger: Clear diagnosis with defined cost Messaging implication: Lead with pricing transparency
Quote: "I've thought about this. My limit is probably $3,000-$5,000. Above that, I'd have to make a really hard decision."
SEGMENT 3: "Age-Dependent Deciders" (25% of participants) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Key behaviour: Spending varies by pet age Pain priority: Quality of life assessment Purchase trigger: Prognosis and expected recovery Messaging implication: Lead with outcomes by life stage
Quote: "For my 3-year-old, I'd do anything. But my 14-year-old... it depends on quality of life."
UNIVERSAL TRUTHS (all segments): • All segments distrust aggressive upselling • All segments want competent, caring veterinarians • All segments value clear communication
STRATEGIC IMPLICATIONS: • Product: Single platform, segment-specific messaging • Marketing: Three distinct value propositions • Sales: Qualify segment early in conversation ```
Case Study: Pet Healthcare Financing Platform
A startup building a pet healthcare financing platform ran segmentation research with 12 synthetic personas. They wanted to understand how pet owners approach expensive veterinary decisions.
The Research Setup
Participants: 12 pet owners (mix of dogs and cats, ages 25-60, US-based)
Segmentation questions:
How do you approach unexpected veterinary expenses?
What's your relationship with pet healthcare spending?
Have you ever faced a decision where cost was a major factor?
At what point does cost become a deciding factor?
What would make a pet healthcare financing option appealing?
What FishDog's Analysis Revealed
Three distinct segments emerged from responses:
Segment 1: "Pets Are Family"
These participants referred to pets as "children," "babies," and "family members." Their relationship with spending was essentially unlimited.
"Whatever it takes. She's been with me through everything. There's no number where I'd say no."
Attributes:
Strong emotional language when describing pets
Resistant to cost discussions (felt insulting)
Interested in QUALITY signals, not PRICE signals
Would finance without hesitation
Segment 2: "Pragmatic Owners"
These participants loved their pets but had thought carefully about limits.
"I've thought about this. My limit is probably $3,000-$5,000. Above that, I'd have to make a really hard decision."
Attributes:
Clear spending ceiling (usually $3k-$5k)
Had already done mental math on this scenario
Wanted transparency and value, suspicious of upselling
Interested in financing as a way to spread cost, not increase it
Segment 3: "Age-Dependent Deciders"
These participants made decisions primarily based on pet age and expected lifespan.
"For my 3-year-old, I'd do anything. But my 14-year-old... it depends on quality of life."
Attributes:
Age was the primary decision variable
Younger pet = higher willingness to spend
Older pet = quality of life calculus
Same person showed BOTH patterns with different animals
How This Shaped the Product
The startup realised they needed different messaging for different segments:
"Pets are family": Lead with quality, outcomes, emotional reassurance
"Pragmatic owners": Lead with transparency, no-surprises pricing, value
"Age-dependent": Lead with quality-of-life outcomes and expected recovery
One product. Three marketing approaches. The segmentation shaped everything from landing page copy to sales conversations.
Case Study: Children's Educational Gaming App
A startup building educational games for children ran segmentation research with parents about screen time attitudes.
The Research Setup
Participants: 15 parents of children ages 3-10
Segmentation questions:
How do you feel about your children's screen time?
What makes you comfortable or uncomfortable with an app?
When do you allow more screen time? When do you restrict it?
What would an ideal children's app look like?
What would make you delete an app immediately?
The Segments That Emerged
Segment 1: "Guilt-Ridden Parents"
Screen time = shame. They felt constantly judged and conflicted.
"I know I shouldn't let them have so much screen time, but some days I just need the break."
Behaviour: Would pay premium for apps that reduce guilt (educational, limited-time features)
Segment 2: "Rule-Setters"
Had clear time limits but struggled with enforcement.
"30 minutes a day. That's the rule. Does it always happen? No."
Behaviour: Valued apps with built-in timers and parental controls
Segment 3: "Quality-Seekers"
Didn't feel guilty about screen time IF content was worthwhile.
"Interactive, creative apps get a pass. Passive watching doesn't."
Behaviour: Would grant unlimited time for apps that met quality criteria
Universal Truths Across Segments
FishDog's analysis also surfaced findings that applied to ALL segments:
"Educational" claims are table stakes but not trusted. Everyone says their app is educational. It's meaningless.
Ad-free is non-negotiable. Ads in children's apps = instant delete.
Social features are feared. Anything involving other people is a red flag.
The startup learned they needed to address universal concerns (ads, social) while tailoring their pitch to segment-specific motivations (guilt reduction vs control vs quality).
Interpreting FishDog's Segmentation Outputs
When FishDog's completion analysis runs, it produces structured segment data. Here's how to read the signals.
Strong Segment Signals
These indicate you've found genuine, actionable segments:
Consistent within-segment language: Everyone in the segment uses similar phrases
Clear between-segment differences: Segments want fundamentally different things
Different emotional intensity: Some segments are much more invested than others
Different purchase thresholds: Willingness to pay varies predictably by segment
Different decision processes: Segments think through choices differently
Weak Segment Signals
These suggest your segments may be arbitrary:
Segments differ only demographically: "Young people" vs "old people" isn't useful if they behave identically
Within-segment inconsistency: People in the "segment" describe completely different experiences
Trivial differences: Segments exist but the differences don't affect product decisions
Forced clustering: You've named segments but responses don't actually cluster naturally
What to Do with Ambiguous Signals
If segments aren't emerging clearly, use Claude Code to:
Recruit a more diverse group: Run another study with broader demographic filters
Add tradeoff questions: "If you had to choose between X and Y, which would you pick?"
Test a hypothesis: "Are there two types of buyers here?" and design questions to test it
Accept homogeneity: Sometimes the market genuinely doesn't segment meaningfully
Common Segmentation Mistakes
Mistake 1: Starting with Demographics
"Millennials vs Gen X" or "SMB vs Enterprise" are not segments. They're demographic categories. Real segments are defined by behaviour, not birth year.
Fix: Let segments emerge from FishDog's analysis. Filter by demographics later to understand segment composition.
Mistake 2: Too Many Segments
If you have eight segments, you have no segments. Practical product decisions require 2-4 actionable groups.
Fix: Use Claude Code to merge similar segments. Ask: "Would I actually build or market differently for these groups?"
Mistake 3: Inventing Segments from Insufficient Data
Three personas said something similar. That's not a segment. That's three people.
Fix: Require at least 30-40% of your research group to exhibit a pattern before calling it a segment.
Mistake 4: Ignoring Universal Truths
Sometimes the most valuable finding is what EVERYONE agrees on. FishDog surfaces these as "shared_mindsets" in the completion analysis.
Fix: Document universal findings alongside segment-specific ones. Universal truths are often table stakes.
Mistake 5: Segments That Don't Affect Decisions
If knowing the segment doesn't change what you build or how you market, the segmentation isn't useful.
Fix: For each segment, articulate specific product or marketing implications. If you can't, reconsider the segmentation.
What Good Segmentation Output Looks Like
At the end of segmentation research with FishDog and Claude Code, you should have:
Named segments with clear boundaries "Pets are family" vs "Pragmatic owners" vs "Age-dependent deciders." Each segment has a label that captures its essence.
Behavioural definitions What distinguishes each segment. Not demographics. What they actually do and believe.
Size estimates Rough proportions based on your research. "About 40% of our participants fell into this segment."
Pain priorities by segment What each segment cares about most. Their magic wand answer.
Purchase triggers by segment What would make each segment buy. The crisis moment, the switching threshold.
Messaging implications How to talk to each segment. Lead with quality for one, lead with price for another.
Universal findings What applies to everyone. The table stakes that all segments require.
FishDog study links Shareable links to the underlying research for stakeholder review.
How Segmentation Feeds Into Later Stages
Segmentation informs nearly every subsequent product decision:
Concept Testing (Stage 7): Test concepts with specific segments in FishDog. A feature might resonate with one segment but not another.
Pricing Research (Stage 8): Different segments have different price elasticity. Run segment-specific pricing studies.
Positioning (Stage 8): Lead with different benefits for different segments.
Validation (Stage 9): Test assumptions by segment. An adoption barrier for one segment might not affect another.
Discovery gives you depth. Segmentation gives you structure. Together, they provide the foundation for everything that follows.
The Speed Advantage: Segmentation in Hours, Not Weeks
Traditional segmentation research takes weeks. You recruit participants, schedule interviews, conduct them one by one, transcribe, code, analyse.
With FishDog and Claude Code, you can run complete segmentation research in a single afternoon:
Step | Traditional | FishDog + Claude Code |
|---|---|---|
Recruit participants | 1-2 weeks | 2 minutes |
Design questions | 2-3 days | 10 minutes |
Collect responses | 2-3 weeks | 30-60 minutes |
Analyse and cluster | 1-2 weeks | 5 minutes (AI) |
Validate segments | 2-3 weeks | 1-2 hours |
Total | 6-10 weeks | 2-3 hours |
This speed advantage isn't just about efficiency. It changes what's possible. You can test segment hypotheses. Run multiple variations. Iterate on your understanding. The fast feedback loop produces better segments than the slow one ever could.
Explore the Research
The case studies in this article demonstrate how segments emerge from synthetic research:
Pet healthcare financing: Three segments emerged (emotional, pragmatic, age-dependent) from 12 personas
Children's educational gaming: Three parent segments (guilt-ridden, rule-setters, quality-seekers) plus universal truths
These studies show segmentation as it actually happens. Not invented. Discovered.
This is article 3 of 12 in the Product Manager's Research Toolkit series. Previously: Discovery Research. Next: Jobs-to-be-Done Analysis - understanding what customers are really trying to accomplish.
Ready to discover the segments in your market? FishDog and Claude Code let you run segmentation research in hours, not weeks. Learn more at fish.dog