persona-cohort-analyzer

# Persona Cohort Analyzer

This skill transforms raw qualitative data from synthetic persona interactions into structured insights, statistical patterns, and actionable research findings.

## Purpose

Qualitative research generates rich data but extracting patterns across multiple interviews is time-consuming and subjective. This skill provides systematic frameworks for:

- Identifying themes across 10-1000+ persona responses
- Quantifying qualitative findings (sentiment, intent, preference)
- Comparing segments to surface meaningful differences
- Detecting consensus, divergence, and outlier perspectives
- Generating executive-ready research reports
- Tracking longitudinal changes in cohort sentiment

## When to Use This Skill

Use this skill when:

- Multiple personas have responded to the same questions/concepts
- Interview transcripts need systematic theme extraction
- Stakeholders need quantified findings from qualitative data
- Segment comparisons are needed for strategic decisions
- Research findings need to be synthesized into reports
- Pattern detection is required across large response sets

## Analysis Frameworks

### 1. Thematic Analysis

Systematically identify, organize, and report patterns (themes) across qualitative data.

**Process:**

```
THEMATIC ANALYSIS STEPS:

1. FAMILIARIZATION
   Read through all responses to get holistic sense
   Note initial impressions and recurring ideas
   
2. INITIAL CODING
   Identify discrete units of meaning
   Apply descriptive labels to segments
   Remain close to the data (in-vivo codes when possible)
   
3. THEME GENERATION
   Group related codes into candidate themes
   Check themes against coded data
   Ensure themes are distinct and coherent
   
4. THEME REFINEMENT
   Merge overlapping themes
   Split themes that are too broad
   Name themes to capture essence
   
5. THEME VALIDATION
   Verify themes work across the full dataset
   Check for data that contradicts themes
   Assess theme saturation
```

**Output Format:**

```
THEMATIC ANALYSIS REPORT

Research Question: [What were we trying to learn?]
Cohort: [N personas, segment breakdown]
Data Source: [Interviews, surveys, concept tests]

IDENTIFIED THEMES:

Theme 1: [Theme Name]
Definition: [What this theme captures]
Prevalence: [X of N personas, Y%]
Representative Quotes:
  - "[Quote 1]" - Persona A
  - "[Quote 2]" - Persona B
Segment Variation: [How this differs across segments]

Theme 2: [Theme Name]
[Same structure...]

THEME RELATIONSHIPS:
[How themes connect, conflict, or build on each other]

OUTLIER PERSPECTIVES:
[Views that didn't fit major themes but may be significant]
```

### 2. Sentiment Distribution Analysis

Quantify emotional responses across the cohort.

**Framework:**

```
SENTIMENT CODING SCALE:

+2 = Strongly Positive (enthusiastic, excited, delighted)
+1 = Moderately Positive (pleased, satisfied, interested)
 0 = Neutral (indifferent, ambivalent, mixed)
-1 = Moderately Negative (disappointed, frustrated, skeptical)
-2 = Strongly Negative (angry, hostile, dismissive)

CODING RULES:
- Code based on emotional tone, not just words
- Watch for sarcasm and qualified positivity
- Note intensity markers ("really," "absolutely," "kind of")
- Capture mixed responses as 0 with notation
```

**Output Format:**

```
SENTIMENT DISTRIBUTION REPORT

Concept/Topic: [What was being evaluated]
Cohort: [N personas]

OVERALL DISTRIBUTION:
Strongly Positive (+2): XX% (n=X)
Moderately Positive (+1): XX% (n=X)
Neutral (0): XX% (n=X)
Moderately Negative (-1): XX% (n=X)
Strongly Negative (-2): XX% (n=X)

Mean Sentiment Score: X.XX
Standard Deviation: X.XX

SEGMENT BREAKDOWN:
| Segment | Mean | Distribution Skew |
|---------|------|-------------------|
| [Seg A] | X.XX | [Left/Center/Right] |
| [Seg B] | X.XX | [Left/Center/Right] |

SENTIMENT DRIVERS:
Positive Drivers: [What creates positive sentiment]
Negative Drivers: [What creates negative sentiment]
Polarization Factors: [What splits the cohort]
```

### 3. Purchase Intent Analysis

Evaluate likelihood to buy/adopt across the cohort.

**Framework:**

```
PURCHASE INTENT SCALE:

5 = Definite ("I would absolutely buy this")
4 = Probable ("I would likely buy this")
3 = Possible ("I might buy this")
2 = Unlikely ("I probably wouldn't buy this")
1 = No Intent ("I definitely wouldn't buy this")

INTENT QUALIFIERS:
- Conditional Intent: "I would if [condition]"
- Timing Dependency: "Not now, but maybe later"
- Price Contingent: "Depends on the price"
- Trial First: "I'd need to try it first"
```

**Output Format:**

```
PURCHASE INTENT ANALYSIS

Product/Concept: [What was tested]
Cohort: [N personas]

INTENT DISTRIBUTION:
Definite (5): XX% (n=X)
Probable (4): XX% (n=X)
Possible (3): XX% (n=X)
Unlikely (2): XX% (n=X)
No Intent (1): XX% (n=X)

Top-2 Box (4+5): XX%
Bottom-2 Box (1+2): XX%

CONDITIONAL INTENT BREAKDOWN:
"Would buy if..." conditions:
- [Condition 1]: XX% mentioned
- [Condition 2]: XX% mentioned

BARRIERS TO INTENT:
- [Barrier 1]: XX% cited
- [Barrier 2]: XX% cited

SEGMENT COMPARISON:
| Segment | Top-2 Box | Primary Barrier |
|---------|-----------|-----------------|
| [Seg A] | XX%       | [Barrier]       |
| [Seg B] | XX%       | [Barrier]       |
```

### 4. Segment Comparison Analysis

Identify meaningful differences between defined segments.

**Framework:**

```
COMPARISON DIMENSIONS:

BEHAVIORAL DIFFERENCES:
- How segments currently solve the problem
- Purchase patterns and preferences
- Channel and touchpoint preferences

ATTITUDINAL DIFFERENCES:
- How segments feel about the category
- Values and priorities that differ
- Risk tolerance and openness to new

RESPONSE DIFFERENCES:
- How segments react to concepts/stimuli
- Language and framing preferences
- Objection patterns

STATISTICAL SIGNIFICANCE:
For quantitative measures, note when differences
exceed threshold for meaningful distinction
(rule of thumb: >10 percentage points for qualitative)
```

**Output Format:**

```
SEGMENT COMPARISON REPORT

Segments Compared: [Segment A] vs [Segment B] vs [Segment C]
Sample Sizes: A=X, B=X, C=X

KEY DIFFERENCES SUMMARY:

Dimension 1: [Dimension Name]
| Measure | Seg A | Seg B | Seg C | Notable |
|---------|-------|-------|-------|---------|
| [Metric]| XX%   | XX%   | XX%   | [Note]  |

Dimension 2: [Dimension Name]
[Same structure...]

IMPLICATIONS BY SEGMENT:

Segment A Insights:
- [What's unique about this segment]
- [Strategic implications]

Segment B Insights:
- [What's unique about this segment]
- [Strategic implications]

COMMON GROUND:
[What all segments agree on]

POINTS OF DIVERGENCE:
[Where segments fundamentally differ]
```

### 5. Consensus vs. Divergence Mapping

Identify where the cohort agrees and where opinions split.

**Framework:**

```
CONSENSUS CLASSIFICATION:

STRONG CONSENSUS (>80% alignment):
Universal or near-universal agreement
Safe to treat as cohort-wide truth

MODERATE CONSENSUS (60-80% alignment):
Majority view with meaningful minority
May require nuanced positioning

SPLIT OPINION (40-60% alignment):
No clear majority
Segment-based strategies likely needed

POLARIZED (<40% in middle):
Strong opinions on opposite ends
May indicate distinct personas to serve differently
```

**Output Format:**

```
CONSENSUS MAP

Topic: [Research focus]
Cohort: [N personas]

STRONG CONSENSUS AREAS:
✓ [Topic 1]: XX% agree that [finding]
✓ [Topic 2]: XX% agree that [finding]

MODERATE CONSENSUS AREAS:
~ [Topic 3]: XX% majority, XX% minority view
  Majority: [position]
  Minority: [position]

SPLIT OPINION AREAS:
÷ [Topic 4]: XX% vs XX% vs XX%
  View A: [position]
  View B: [position]
  View C: [position]

POLARIZED AREAS:
⚡ [Topic 5]: XX% strongly for, XX% strongly against
   Pro Argument: [position]
   Con Argument: [position]
   Neutral: XX% (often uncomfortable/undecided)
```

### 6. Longitudinal Trend Analysis

Track how cohort sentiment/behavior changes over time.

**Framework:**

```
LONGITUDINAL TRACKING:

Time Points: [T1, T2, T3, etc.]
Cohort: [Same personas tracked across time]

METRICS TO TRACK:
- Overall sentiment score
- Purchase intent
- Key theme prevalence
- Barrier mentions
- Competitive mentions

CHANGE CLASSIFICATION:
Significant Improvement: >0.5 point increase (on 5-point scale)
Moderate Improvement: 0.2-0.5 point increase
Stable: <0.2 point change
Moderate Decline: 0.2-0.5 point decrease
Significant Decline: >0.5 point decrease
```

**Output Format:**

```
LONGITUDINAL ANALYSIS REPORT

Tracking Period: [Start] to [End]
Cohort: [N personas]
Touchpoints: [T1=X, T2=X, T3=X]

TREND SUMMARY:
| Metric | T1 | T2 | T3 | Trend |
|--------|-----|-----|-----|-------|
| Sentiment | X.X | X.X | X.X | [↑↓→] |
| Intent | XX% | XX% | XX% | [↑↓→] |

KEY SHIFTS:

Positive Shifts:
- [What improved and why]

Negative Shifts:
- [What declined and why]

INFLECTION POINTS:
[Events or interventions that caused notable change]

LEADING INDICATORS:
[Early signals that predicted later changes]
```

## Quantification Methods

### Converting Qualitative to Quantitative

```
CODING APPROACHES:

FREQUENCY COUNTING:
Count how many personas mention a theme/topic
Express as percentage of cohort
Example: "15 of 20 personas (75%) mentioned price as a concern"

INTENSITY SCORING:
Rate strength of expressed opinion (1-5 or sentiment scale)
Calculate mean and distribution
Example: "Mean concern intensity: 3.2/5.0"

BINARY CODING:
Yes/No for specific behaviors or attributes
Enables cross-tabulation
Example: "Would recommend: Yes 65%, No 35%"

RANKING EXTRACTION:
When personas rank or compare options
Calculate Borda counts or preference shares
Example: "Feature A ranked first by 40%, Feature B by 35%"
```

### Statistical Considerations

```
SAMPLE SIZE GUIDANCE:

For qualitative theme identification:
- 5-8 personas: Adequate for initial exploration
- 10-15 personas: Good for theme saturation
- 20-30 personas: Reliable pattern identification
- 50+ personas: Enables meaningful quantification

For segment comparisons:
- Minimum 5 per segment for directional findings
- 10+ per segment for confident comparisons
- 20+ per segment for statistical reliability

CONFIDENCE QUALIFIERS:
Always note sample sizes with findings
Use language like "directionally suggests" for small n
Reserve "clearly demonstrates" for robust samples
```

## Report Generation

### Executive Summary Format

```
EXECUTIVE SUMMARY TEMPLATE

RESEARCH OVERVIEW
Objective: [1-2 sentences on what we wanted to learn]
Method: [Brief description of approach]
Sample: [N personas, key segment breakdown]

KEY FINDINGS (Top 3-5)
1. [Most important finding with key data point]
2. [Second finding with support]
3. [Third finding with support]

STRATEGIC IMPLICATIONS
- [What this means for business decisions]
- [Recommended actions]

CONFIDENCE ASSESSMENT
What we're confident about: [High-confidence findings]
What needs validation: [Findings requiring human research]
What we don't know: [Gaps identified]
```

### Full Research Report Format

```
FULL REPORT STRUCTURE

1. EXECUTIVE SUMMARY (1 page)
   [As above]

2. RESEARCH DESIGN (1-2 pages)
   - Objectives and research questions
   - Methodology and approach
   - Cohort composition and segments
   - Limitations and caveats

3. DETAILED FINDINGS (5-10 pages)
   - Thematic analysis results
   - Quantitative metrics
   - Segment comparisons
   - Key quotes and examples

4. SYNTHESIS AND IMPLICATIONS (2-3 pages)
   - Pattern interpretation
   - Strategic recommendations
   - Risks and considerations

5. APPENDIX
   - Full persona profiles
   - Complete data tables
   - Methodology details
```

## Examples

### Example 1: Product Concept Analysis

**Input Data:**
- 25 synthetic personas evaluated a meal kit subscription concept
- Each persona provided: initial reaction, purchase intent, concerns, suggestions
- Segments: Singles (8), Couples (9), Families (8)

**Analysis Output:**

```
MEAL KIT CONCEPT TEST ANALYSIS

COHORT: 25 personas (Singles: 8, Couples: 9, Families: 8)

SENTIMENT DISTRIBUTION:
Strongly Positive: 16% (4)
Moderately Positive: 36% (9)
Neutral: 24% (6)
Moderately Negative: 20% (5)
Strongly Negative: 4% (1)

Mean Sentiment: +0.4 (slightly positive)

PURCHASE INTENT:
Top-2 Box (Definite + Probable): 44%
Bottom-2 Box (Unlikely + No Intent): 28%
Middle (Possible): 28%

THEMATIC ANALYSIS:

Theme 1: Convenience vs. Control Trade-off
Prevalence: 72% (18 of 25)
Personas want convenience but fear losing control over
ingredients, portions, and variety. This creates hesitation
even among those positively disposed to the concept.

Key Quote: "I love the idea of not planning meals, but what
if I don't like what they send? Then I've paid for food I
won't eat." - Persona 12 (Family segment)

Theme 2: Subscription Fatigue
Prevalence: 48% (12 of 25)
Many personas expressed weariness with subscription models,
citing forgotten payments and guilt about unused services.

Key Quote: "Another subscription? I'm already paying for
three things I barely use." - Persona 7 (Singles segment)

Theme 3: Environmental Concerns
Prevalence: 36% (9 of 25)
Packaging waste emerged as unexpected barrier, particularly
among younger personas and parents.

Key Quote: "All those little packets and containers... where
does that all go?" - Persona 19 (Couples segment)

SEGMENT COMPARISON:

| Metric | Singles | Couples | Families |
|--------|---------|---------|----------|
| Top-2 Intent | 38% | 56% | 38% |
| Price Concern | 63% | 33% | 75% |
| Convenience Driver | 75% | 67% | 88% |

Key Insight: Couples show highest intent, driven by
"date night at home" use case. Singles and families
both struggle with value perception but for different
reasons (singles = portion waste, families = not enough food).

CONSENSUS MAP:
✓ Strong Consensus: Quality ingredients matter (92%)
✓ Strong Consensus: Flexibility to skip weeks essential (88%)
÷ Split: Weekly vs. bi-weekly preference (48% vs. 44%)
⚡ Polarized: Willingness to pay premium (36% yes, 40% no)

RECOMMENDATIONS:
1. Lead with flexibility messaging to address subscription fatigue
2. Develop distinct value props for couples vs. singles/families
3. Address packaging sustainability prominently
4. Consider non-subscription purchase option for hesitant segments
```

---

### Example 2: Message Testing Analysis

**Input Data:**
- 30 personas evaluated 4 taglines for financial planning app
- Each rated appeal (1-5) and explained their reaction
- Segments: Risk-Averse (15), Risk-Tolerant (15)

**Analysis Output:**

```
TAGLINE TESTING ANALYSIS

TAGLINES TESTED:
A) "Your Money, Your Future, Your Control"
B) "Smart Investing Made Simple"
C) "Build Wealth While You Sleep"
D) "Financial Freedom Starts Here"

APPEAL RATINGS (Mean, 1-5 scale):

| Tagline | Overall | Risk-Averse | Risk-Tolerant | Gap |
|---------|---------|-------------|---------------|-----|
| A | 3.8 | 4.2 | 3.4 | 0.8 |
| B | 3.6 | 3.9 | 3.3 | 0.6 |
| C | 2.9 | 2.1 | 3.7 | 1.6 |
| D | 3.4 | 3.5 | 3.3 | 0.2 |

KEY FINDING: Tagline C is highly polarizing—loved by risk-tolerant,
distrusted by risk-averse. Taglines A and D have broader appeal
but generate less enthusiasm overall.

THEMATIC RESPONSES BY TAGLINE:

Tagline A: "Your Money, Your Future, Your Control"
+ Control resonates strongly with risk-averse
+ Empowerment language appeals broadly
- Some found it generic/"could be anything"
- Risk-tolerant wanted more ambition

Tagline B: "Smart Investing Made Simple"
+ "Simple" appreciated by tech-hesitant
+ "Smart" validates the user
- "Made Simple" felt patronizing to some
- Risk-tolerant felt it undersold sophistication

Tagline C: "Build Wealth While You Sleep"
+ Passive income dream resonates with risk-tolerant
+ Memorable and distinctive
- "Too good to be true" alarm for risk-averse
- Sounds like scam/scheme to skeptical personas

Tagline D: "Financial Freedom Starts Here"
+ "Freedom" is universal positive
+ "Starts Here" creates entry point
- Generic; could apply to any financial product
- Doesn't differentiate from competitors

RECOMMENDATION:
Primary: Tagline A for broad market launch
Secondary: Tagline C for risk-tolerant segment targeting
Avoid: Tagline C in risk-averse segment communications

Consider testing hybrid: "Your Path to Financial Freedom, On Your Terms"
(combines A's control with D's freedom aspiration)
```

---

### Example 3: Longitudinal NPS Tracking

**Input Data:**
- 20 personas tracked over simulated 90-day period
- Check-ins at Day 1, Day 30, Day 60, Day 90
- Measuring satisfaction, NPS, and key theme evolution

**Analysis Output:**

```
LONGITUDINAL SATISFACTION ANALYSIS

Cohort: 20 personas (B2B software users)
Period: 90-day simulated onboarding journey

NPS TREND:
| Timepoint | Promoters | Passives | Detractors | NPS |
|-----------|-----------|----------|------------|-----|
| Day 1 | 45% (9) | 40% (8) | 15% (3) | +30 |
| Day 30 | 35% (7) | 35% (7) | 30% (6) | +5 |
| Day 60 | 40% (8) | 40% (8) | 20% (4) | +20 |
| Day 90 | 50% (10) | 35% (7) | 15% (3) | +35 |

TREND INTERPRETATION:
- Day 1 honeymoon effect (high initial optimism)
- Day 30 reality check (complexity hits, NPS drops)
- Day 60 recovery begins (users find their groove)
- Day 90 stabilization at higher baseline

THEME EVOLUTION:

"Ease of Use" Mentions:
Day 1: 70% positive
Day 30: 40% positive (complexity discovered)
Day 60: 55% positive (adaptation occurring)
Day 90: 75% positive (proficiency achieved)

"Value for Money" Mentions:
Day 1: 50% positive
Day 30: 35% positive (haven't seen ROI yet)
Day 60: 60% positive (starting to see benefits)
Day 90: 80% positive (clear ROI realized)

"Support Quality" Mentions:
Day 1: N/A (not yet needed)
Day 30: 60% positive (critical period)
Day 60: 70% positive
Day 90: 75% positive

INFLECTION POINT ANALYSIS:

Day 30 Dip Root Causes:
1. Report builder complexity (mentioned by 8 personas)
2. Integration challenges (mentioned by 5 personas)
3. Unmet feature expectations (mentioned by 4 personas)

Day 60 Recovery Drivers:
1. Training webinar attendance (12 personas)
2. Success manager check-in (15 personas)
3. First "aha moment" with advanced features (10 personas)

RECOMMENDATIONS:
1. Proactive outreach at Day 14 to prevent Day 30 dip
2. Mandatory report builder training in first week
3. Set realistic expectations in sales process
4. Celebrate "aha moments" to accelerate recovery
```

## Integration with Research Workflow

This skill works best as part of a connected research system:

```
RECOMMENDED WORKFLOW:

1. Create personas → [synthetic-persona-creator]
2. Conduct interviews → [synthetic-interview-conductor]
3. Analyze patterns → [persona-cohort-analyzer] ← YOU ARE HERE
4. Generate actions → [research-to-action-bridge]
```

The analysis outputs feed directly into action generation, closing the loop from insight to execution.

## Quality and Limitations

### Confidence Calibration

```
HIGH CONFIDENCE (report with conviction):
- Strong consensus findings (>80%)
- Consistent patterns across segments
- Supported by multiple data points
- Aligns with known market behavior

MODERATE CONFIDENCE (report with caveats):
- Moderate consensus (60-80%)
- Some segment variation
- Limited supporting data
- Novel or unexpected findings

LOW CONFIDENCE (flag for validation):
- Split or polarized findings
- Small sample sizes
- Contradictory data points
- Findings that challenge strong assumptions
```

### When Human Validation is Essential

Always recommend human research validation when:

- Findings will drive >$100K decisions
- Results are surprisingly positive (sycophancy check)
- Cultural or demographic nuances are critical
- Novel product categories with no precedent
- Findings contradict established market knowledge