5 Prompt Patterns for Customer Research and Analysis

Customer research generates mountains of qualitative data — interview transcripts, survey responses, support tickets, reviews — and most of it sits unanalyzed because synthesizing it takes time nobody has.

AI is particularly strong at processing qualitative data, finding patterns across multiple sources, and organizing unstructured feedback into themes. But "analyze this customer feedback" produces a surface-level summary that tells you nothing you didn't already know.

These five patterns turn AI into a genuine research assistant. Each one handles a different stage of customer research, from raw data processing to strategic synthesis.

Pattern 1: The Interview Analyzer

This pattern processes customer interview transcripts or notes and extracts structured insights. It's designed to handle the messy, conversational nature of real interviews.

You are a UX researcher analyzing a customer interview.

Interview context:
- Interviewee: [Role, company size, how long they've been a customer]
- Interview goal: [What we were trying to learn]
- Product/service discussed: [What we were asking about]

Interview notes or transcript:
[Paste here]

Analyze this interview and provide:

1. **Top 3 pain points**: Specific problems the customer described, ranked by emotional intensity. For each, include a direct quote (or close paraphrase) that captures the frustration.

2. **Unmet needs**: Things the customer wants but doesn't currently have — including needs they expressed indirectly or implied without stating explicitly.

3. **Current workarounds**: How they're solving problems today without our solution (or despite our solution). These reveal the highest-value opportunities.

4. **Surprising insights**: Anything that contradicts our assumptions or reveals something unexpected about their behavior or thinking.

5. **Sentiment signals**: Was the customer generally positive, neutral, or negative? Where did their energy or engagement peak?

Rules:
- Distinguish between what the customer explicitly said and what you're inferring. Label inferences with [INFERRED].
- If the customer contradicted themselves, note both statements.
- Don't summarize away nuance — preserve the specificity of their language.

Why it works: Ranking pain points by emotional intensity (not just frequency) surfaces what customers actually care about most. The "current workarounds" category reveals product opportunities that customers won't directly ask for. Labeling inferences prevents the AI from presenting interpretation as fact.

Example output snippet:

Pain point #1 (High intensity): Manual data reconciliation between systems

Quote: "I spend the first two hours of every Monday morning just making sure the numbers in our CRM match what's in the billing system. It's mind-numbing."

>

Current workaround: Built a personal spreadsheet that cross-references both systems. Updates it weekly. Acknowledged it "breaks constantly" when fields change.

>

Surprising insight: Despite being a power user of our platform, they were unaware of the API integration that would solve this. [INFERRED: Our documentation or onboarding may be failing advanced users.]

Pattern 2: The Persona Builder

This pattern creates research-backed customer personas from real data — not the fictional, stock-photo personas that nobody uses. It requires actual customer data as input.

You are a product strategist building customer personas from real data.

Data sources — I'll provide one or more of the following:
[Paste: interview notes, survey responses, support tickets, sales call notes, analytics data, or any other customer data]

Build a customer persona that includes:

1. **Identity**: Name (fictional), role, company type, and a one-sentence "day in the life" summary
2. **Goals**: What they're trying to accomplish (professional and personal motivations)
3. **Frustrations**: Their top 3 recurring pain points — use language from the actual data, not corporate jargon
4. **Behaviors**: How they currently solve problems (tools, habits, workarounds)
5. **Decision factors**: What influences their purchasing decisions (price sensitivity, feature requirements, social proof, risk tolerance)
6. **Objections**: Their likely hesitations about our product/solution and what would overcome each
7. **Preferred communication**: How they want to be marketed to (channels, content types, tone)

Rules:
- Base every element on evidence from the data provided. If a section requires guessing, mark it [INSUFFICIENT DATA].
- Include conflicting data points if they exist — real customers are contradictory.
- This persona should be specific enough that a product manager can use it to prioritize features, and a marketer can use it to write targeted copy.

Why it works: Requiring data-backed evidence prevents the AI from generating generic personas that could describe anyone. Including objections makes the persona immediately useful for sales and marketing. The "conflicting data points" instruction keeps the persona honest — real customer segments aren't monolithic.

Example output snippet:

"Operations Olivia"

Director of Operations at a mid-market B2B company (200-500 employees). Spends most of her day in spreadsheets and Slack, managing cross-team workflows that outgrew their current tools two years ago.

>

Frustrations:

1. "Every tool we try solves one problem but creates three integrations to manage"

2. Reports take hours to compile because data lives in 4+ systems

3. Gets resistance from her team when introducing new tools ("they've got change fatigue")

>

Objections:

- "How long until my team actually uses this?" → Overcome with concrete onboarding timeline and customer benchmarks

- "What happens to our existing workflows during migration?" → Overcome with parallel-run capability demonstration

Pattern 3: The Feedback Synthesizer

This pattern processes large volumes of customer feedback (reviews, NPS comments, support tickets) and distills them into actionable themes. It's designed for scale — handling dozens or hundreds of data points.

You are a customer insights analyst processing a batch of customer feedback.

Feedback source: [e.g., "NPS survey comments", "App Store reviews", "support tickets from March 2026"]
Total pieces of feedback: [approximate number]
Our product/service: [Brief description]

Feedback data:
[Paste all feedback — each piece on a new line or clearly separated]

Synthesize this feedback:

1. **Theme analysis**: Group feedback into 5-7 themes. For each theme:
   - Theme name (descriptive, not vague)
   - Number of mentions (or approximate frequency)
   - Representative quotes (2-3 per theme)
   - Sentiment breakdown within the theme (positive/negative/mixed)

2. **Priority matrix**: Rank themes by combining frequency (how many people mentioned it) with intensity (how strongly people feel about it). High frequency + high intensity = top priority.

3. **Hidden signals**: Patterns that aren't the loudest but may be important:
   - Issues mentioned by a small but specific segment
   - New topics that didn't appear in previous feedback cycles
   - Praise that reveals unexpected use cases

4. **One-sentence executive summary**: If leadership can only hear one thing from this data, what should it be?

Rules:
- Don't dilute strong negative feedback by softening the language
- If feedback is contradictory (some love feature X, others hate it), report both sides with counts
- Separate feature requests from complaints — they need different responses

Why it works: The priority matrix prevents the common mistake of treating all feedback equally. "Hidden signals" catches important minority feedback that gets drowned out in theme analysis. The executive summary forces the AI to commit to the single most important insight, which is what decision-makers need.

Example output snippet:

Theme #1: Slow load times (mentioned ~34 times, HIGH intensity)

- Sentiment: Strongly negative

- Quotes: "The dashboard takes so long to load that I've started making coffee while I wait" / "Page load is unusable on cellular" / "This was fast when we had 50 users, now with 500 it's painful"

>

Hidden signal: 5 customers mentioned using the product as a client-facing reporting tool — a use case we haven't designed for but which has retention implications.

>

Executive summary: Customers value the product's functionality but are increasingly frustrated by performance degradation at scale — this is the top risk to retention.

Pattern 4: The Journey Mapper

This pattern maps the customer journey from first awareness through long-term retention, identifying emotional highs and lows, friction points, and moments that matter most.

You are a CX strategist mapping the customer journey.

Based on the following customer data:
[Paste: interview excerpts, survey data, support tickets, onboarding metrics, churn reasons, or any combination]

Context:
- Product/service: [What we offer]
- Typical customer lifecycle: [How long from signup to full adoption]
- Known drop-off points: [Where we lose customers, if known]

Map the customer journey across these stages:
1. **Awareness**: How they discover us
2. **Evaluation**: What they compare us against and how they decide
3. **Onboarding**: First experience after purchase/signup
4. **Adoption**: Getting to regular, habitual usage
5. **Retention/Expansion**: Staying long-term, expanding usage

For each stage, document:
- **Customer actions**: What they're doing
- **Emotions**: How they feel (based on data, not assumptions)
- **Pain points**: Where friction exists
- **Moments of delight**: Where the experience exceeds expectations
- **Drop-off risk**: What causes people to leave at this stage
- **Opportunity**: One specific improvement we could make

Mark any stage where data is sparse with [LIMITED DATA].

Why it works: Combining emotional state with practical actions creates a journey map that's useful for both product and marketing teams. The "drop-off risk" at each stage turns the map into a retention tool. Requiring data-backed emotions prevents the common trap of journey maps filled with imagined customer feelings.

Example output snippet:

Stage 3: Onboarding

- Actions: Create account, import data, configure settings, invite team members

- Emotions: Initial excitement quickly shifts to frustration during data import

- Pain point: "The CSV import failed three times before I figured out the date format it expected" (mentioned by 8 customers)

- Moment of delight: First automated report generation — "I literally said 'wow' when it pulled everything together in 30 seconds"

- Drop-off risk: 40% of trial users never complete data import (from analytics data)

- Opportunity: Add a data preview/validation step before import to reduce failures

Pattern 5: The Competitive Positioning Analyst

This pattern analyzes what customers say about competitors to identify positioning opportunities — where you can win and where you need to defend.

You are a competitive intelligence analyst focused on customer perception.

Our product: [Brief description and key differentiators]
Competitors being analyzed: [List 2-4 competitors]

Customer data mentioning competitors:
[Paste: reviews comparing products, sales call notes where competitors came up, churn surveys mentioning alternatives, or any data where customers reference competitors]

Analyze competitive positioning from the customer's perspective:

1. **Perception map**: For each competitor, what do customers actually say about them? (Not what the competitor claims — what real users report.) Organize by strengths and weaknesses as perceived by customers.

2. **Switching triggers**: What causes customers to switch FROM competitors TO us? And what causes them to switch AWAY from us?

3. **Feature gaps that matter**: Competitor features customers actually mention and miss — not every feature they have, but the ones customers bring up unprompted.

4. **Positioning whitespace**: Based on this data, where is there an unoccupied position we could own? What do customers want that nobody is delivering well?

5. **Defensive priorities**: Where are we most vulnerable to competitive loss? What should we protect or improve first?

Rules:
- Base analysis only on what customers have said — don't supplement with your own knowledge of competitors
- If data is insufficient for a competitor, say so rather than speculating
- Distinguish between what power users want vs what the majority wants

Why it works: Analyzing competitor perception from the customer's perspective (not from marketing materials) reveals the real competitive landscape. "Switching triggers" are the most actionable data point in competitive research — they tell you exactly what wins and loses deals. The "whitespace" section identifies opportunities for differentiation.

Example output snippet:

Switching triggers — FROM Competitor A TO us:

- Price (mentioned 12 times) — Competitor A raised prices significantly in Q4

- Complexity (mentioned 8 times) — "I needed a PhD to configure their reporting"

- Support quality (mentioned 5 times) — "Their support tickets go into a black hole"

>

Positioning whitespace: No competitor is effectively serving mid-market companies (100-500 employees). Competitor A targets enterprise, Competitor B targets small business. Customers in the middle consistently describe feeling "too small for [A] and too complex for [B]."

Quick Tips for Customer Research Prompts

• Use real data, not summaries. Paste actual quotes, ticket text, and survey responses. Summaries strip out the nuance that makes research valuable.
• Process in batches. If you have 500 survey responses, process 50 at a time and then run a synthesis prompt across the batch analyses. It produces better results than feeding everything at once.
• Specify what you already know. Tell the AI "we already know customers want faster loading — look for insights beyond that" to push past obvious findings.
• Preserve customer language. Add "use the customer's actual words, not corporate paraphrases" to keep insights grounded and persuasive.
• Separate analysis from recommendations. Run one prompt to analyze the data and a second to generate recommendations. Combining both dilutes the analysis.

When to Use Templates vs. Write From Scratch

Use these patterns when:

• You do customer research on a regular cycle (quarterly surveys, monthly feedback reviews)
• You need consistent analysis format across different data sources
• You're handing off the analysis process to a team member

Write from scratch when:

• You're exploring a completely new market segment and need unconstrained analysis
• The research question is highly specific (e.g., "why did enterprise customers in healthcare churn in Q1?")
• You need to combine quantitative and qualitative data in a way these patterns don't cover

For teams running ongoing research programs, SurePrompts' Template Builder lets you save these patterns with your product context, persona definitions, and competitor information pre-loaded.