Deep Customer Research

---
name: deep-customer-research
version: 1.1.0
description: "Use this skill to conduct deep qualitative customer research on any topic. Triggers include: requests to understand customer pain points, unmet needs, motivations, or emotions around a product, category, brand, or market. Also use when validating product-market fit, exploring how an audience feels about a trend, or gathering insights to inform product strategy, positioning, or roadmap decisions. Uses OriginalVoices Digital Twins (ask_twins) to ask 10-12 comprehensive questions per audience in a single pass covering behaviour, pain points, emotions, priorities, decision-making, unmet needs, trust signals, and willingness to pay."
---

# Deep Customer Research Skill

## Overview

Deeply understand what your customers really think and feel about any topic. This workflow uses OriginalVoices Digital Twins to conduct comprehensive qualitative research in a single pass — asking 10-12 carefully designed questions per audience that cover the full landscape of motivations, emotions, behaviours, and unmet needs.

## Workflow Steps

### Step 1: Define the Research Topic & Audience

Gather from the user:
- **Research topic or question**: What do they want to understand? (e.g. "How do parents feel about screen time management apps?")
- **Target audience**: Who should we hear from? Be as specific as possible — age, location, interests, lifestyle, values. (e.g. "Parents aged 28-45 in the US with children under 12 who are concerned about their kids' screen time")

### Step 2: Design the Question Set

Craft 10-12 open-ended questions that cover the full research landscape in a single round. This count is PER AUDIENCE: when comparing segments (e.g. Gen Z vs Gen X), each cohort gets its own full set of 10-12 comparable questions — never split one question budget across cohorts. The questions should progress from broad context through to specific opinions and forward-looking needs.

**The #1 quality goal: maximise the *range* of answers.** A good question surfaces spread, tension, and surprise across respondents. A weak question is *narrow* or *convergent* — it points everyone at the same obvious answer, so the twins echo each other and you learn nothing. Before finalising any question, ask: *"Could I predict most of the answers before asking? Would 15 different people answer this 15 different ways?"* If the answers would cluster, rewrite the question.

**Rules — apply every one of these to every question you write. These are requirements, not suggestions. The examples below are illustrations of the rules, not templates to copy verbatim — adapt the wording to the specific topic and audience.**

1. **Ask about a specific moment or behaviour, not an abstract definition.** Never ask "How do you define / think about [concept]?" — it collapses to textbook answers. Ask "Walk me through the last time…" / "Tell me about a time…".
2. **Keep the valence open.** Never presuppose the experience was good or bad. Banned phrasings: "the time [X] annoyed/frustrated/disappointed you", "where does it fall short", "the last time [brand] let you down". Instead ask "how did it go?" or make it two-sided: "what worked well and what didn't?".
3. **One idea per question.** If a question contains "and" joining two different topics (cost *and* quality), split it into two questions.
4. **No socially-correct or yes/no answers.** If almost everyone would answer the same way ("Is quality important?" → yes), reframe until there's no obvious "right" reply — usually by adding a trade-off ("when is quality *not* worth it?").
5. **Stay neutral — never lead.** The question must not contain the answer or the sentiment. No "Don't you find…", no "How frustrating is…".
6. **Build in room to diverge.** Prefer questions with a fork, trade-off, or spectrum ("what would make you — and what would hold you back?") over flat single-answer prompts.
7. **Open-ended only.** Every question starts with or invites "How / What / Why / Walk me through / Tell me about". No closed questions.

Run the two final tests in Step 3 (Range + Neutrality) on each question before sending. If a question fails any rule above, rewrite it.

**How the rules play out (illustrative — do not copy verbatim):**

- **Anchor in lived experience, not abstract definitions.** Questions that ask people to *define* a concept ("How do you define long-term savings?") collapse to near-identical textbook answers. Instead ask for a specific moment, story, or behaviour ("Tell me about the last time you thought about your long-term savings — what prompted it and how did it feel?"). Concrete beats conceptual, every time.
- **Ask for the specific over the general.** "Walk me through the last time…", "Describe a moment when…", "Tell me about your most recent experience with…". Recalled episodes produce vivid, divergent, quotable answers; general opinions produce bland consensus.
- **Keep the valence open — this is what separates a strong open question from a leading one.** Don't presuppose the experience was good or bad. "The last time [topic] annoyed you…" assumes it did; "Tell me about your last experience with [topic] — how did it go?" lets the respondent supply the emotion, and range comes *from* that freedom. When you want to probe frustration or delight, make it two-sided: "…what worked well, and what didn't?" rather than only asking for the negative.
- **Create room for divergence.** Build in a fork ("What would make you do X — and what would hold you back?"), a trade-off ("What would you give up to get Y?"), or a spectrum ("Where do you sit between A and B, and why?"). Questions with a built-in tension pull answers apart without loading the answer.
- **Surface the exceptions, neutrally.** "What's something about [topic] you think most people get wrong?" or "When does [category] work well for you, and when does it fall short?" pull out the tail of the distribution — but phrase them so a happy respondent and an unhappy one can both answer honestly.
- **One idea per question.** Don't bundle ("How do you feel about cost, quality, and convenience?") — split them. Bundled questions get partial, uneven answers.
- **Avoid questions with a socially-correct answer.** "Is saving money important to you?" → everyone says yes. Reframe so there's no obvious "right" reply ("When is saving *not* worth it for you?").
- **Stay neutral and non-leading.** Don't smuggle in the answer or the sentiment ("Don't you find X frustrating?", "where does it fall short?"). Ask "What's your experience with X?" and let them tell you whether it's good, bad, or mixed.

**Question design framework — cover all of these areas:**

Each example below is framed to pull answers *apart* — anchored in a specific moment, choice, or tension rather than an abstract definition.

| Area | Purpose | Example (open, experience-anchored) |
|------|---------|---------|
| Current behaviour | How they deal with the problem today | "Walk me through the last time you dealt with [problem] — what did you actually do, step by step?" |
| Pain points | What works and what doesn't | "Thinking about your recent experiences with [topic], what has gone well and what has been more of a struggle?" |
| Emotional drivers | What feelings are at play | "Tell me about a time [topic] stirred up a strong reaction for you — what was going on, and how did you feel?" |
| Priorities | What matters most (with trade-off) | "If you could only get one thing right about [category], what would it be — and what would you let slide?" |
| Decision-making | How they choose | "Tell me about the last time you chose a [product/solution] in this space — what tipped the decision?" |
| Current solutions | What they use now and why | "What are you using for [problem] right now, and how well is it working for you?" |
| Unmet needs | Gaps in what's available | "Is there anything you wish [products] in this space could do that none of them seem to?" |
| Ideal outcome | What great looks like | "Imagine you've just had the perfect experience with [solution] — describe what happened." |
| Triggers | What would make them act | "What would make you switch from your current approach — and what's kept you with it so far?" |
| Trust & credibility | What builds confidence | "Has a brand in this space ever earned or lost your trust? What did they do, and what would it take to earn it?" |
| Social influence | Role of others | "When you make decisions about [category], what part do other people — friends, reviews, social media — play, if any?" |
| Willingness to pay / trade-offs | Value perception | "What's the most you'd realistically pay for something that solved [problem] — and what would make it feel worth it or not?" |

### Step 3: Ask All Questions in a Single Call

Send all 10-12 questions to the Digital Twins in one `ask_twins` call per audience (one call per cohort when comparing segments, each with the full question set).

```
ask_twins(
  audience: "[detailed audience description]",
  questions: [
    "Walk me through the last time you dealt with [problem] — what did you actually do, step by step?",
    "Thinking about your recent experiences with [topic], what has gone well and what has been more of a struggle?",
    "Tell me about a time [topic] stirred up a strong reaction for you — what was going on, and how did you feel?",
    "If you could only get one thing right about [category], what would it be — and what would you let slide?",
    "Tell me about the last time you chose a [product/solution] in this space — what tipped the decision?",
    "What are you using for [problem] right now, and how well is it working for you?",
    "Is there anything you wish [products] in this space could do that none of them seem to?",
    "Imagine you've just had the perfect experience with [solution] — describe what happened.",
    "What would make you switch from your current approach — and what's kept you with it so far?",
    "Has a brand in this space ever earned or lost your trust? What did they do, and what would it take to earn it?",
    "When you make decisions about [category], what part do other people — friends, reviews, social media — play, if any?",
    "What's the most you'd realistically pay for something that solved [problem] — and what would make it feel worth it or not?"
  ]
)
```

**Before sending, run each question through this quick check:**

| Weak question | Why it fails | Strong (open + wide range) |
|---|---|---|
| "How do you define long-term savings?" | Definitional → everyone gives the same textbook answer | "Tell me about the last time you thought about your long-term savings — what set it off and how did it feel?" |
| "Is quality important when you choose [product]?" | Socially-correct answer → everyone says yes | "When is quality worth paying up for in [category], and when isn't it?" |
| "Do you like [category]?" | Yes/no, no spread | "What do you like about [category], and what would you change if you could?" |
| "The last time [product] let you down, what happened?" | **Leading** — presupposes it let them down; forces a negative | "Tell me about your most recent experience with [product] — how did it go?" |
| "How do you feel about cost, quality and service?" | Bundled → partial, uneven answers | Split into three separate questions, one idea each |

Two tests before you send: (1) **Range** — could most respondents answer this the same way, or is there an obvious "right" answer? If so, anchor it in a specific moment or add a trade-off. (2) **Neutrality** — does the question assume how they feel (annoyed, disappointed, delighted)? If so, open the valence so a happy, unhappy, or indifferent respondent can each answer it honestly. A question should pass *both* — wide range, no leading.

### Step 4: Analyse & Synthesise

Review all Digital Twin responses and compile findings:

- **Key themes**: Patterns across respondents and across questions
- **Emotional landscape**: Dominant feelings (frustration, anxiety, hope, guilt, indifference)
- **Current behaviour map**: How the audience navigates this space today
- **Pain point ranking**: Most intense and most common frustrations
- **Unmet needs**: Clearest gaps between what's available and what's wanted
- **Decision drivers**: What influences choice and action
- **Trust signals**: What builds or breaks confidence
- **Surprising insights**: Anything unexpected or counterintuitive

### Step 5: Deliver the Research Report

Format the output using proper markdown headers: use `##` for the report title, `###` for major sections, `####` for individual themes/subsections, `>` blockquotes for Digital Twin quotes, and `---` horizontal rules between major sections. Never use plain text or bold-only text as section headers — always use proper markdown headers. Where findings have clear quantitative patterns (e.g. "7 out of 15 twins mentioned cost"), include a chart to visualize the distribution.

### Report sections:

1. **Executive Summary** — Top-line findings in 2-3 sentences
2. **Key Themes** — 3-5 most significant patterns, with supporting quotes from Digital Twins. Include a chart showing theme prevalence.
3. **Emotional Landscape** — What emotions drive behaviour in this space. Include a chart of dominant emotions.
4. **Current Behaviour & Pain Points** — How the audience deals with the problem today and what's broken
5. **Unmet Needs & Ideal Outcomes** — Gaps and what "great" looks like to the audience
6. **Decision Drivers & Trust Signals** — What influences choice and earns confidence. Include a chart ranking decision factors.
7. **Opportunities** — Actionable insights for product, marketing, or strategy
8. **Recommendations** — Specific next steps based on the research

## Tips for Best Results

- **Be specific with audiences**: "Women aged 25-35 in the UK interested in fitness who have tried meal planning apps" yields far better results than "women interested in health"
- **Optimise for range, not coverage**: The best questions produce *spread* — 15 people answering 15 different ways. If you can predict the answers in advance, the question is too narrow. Rewrite it (see the weak→strong check in Step 3).
- **Prefer stories to definitions**: "Tell me about the last time…" and "Walk me through…" beat "How do you define…" or "What do you think about…" — recalled episodes are vivid, specific, and divergent; abstractions converge.
- **Build in a tension or trade-off**: A fork ("what would make you — and what would stop you?") or a trade-off ("what would you give up for…?") pulls answers apart and surfaces the interesting minority views.
- **Ask open-ended questions**: Start with "How", "What", "Why", "Walk me through..."
- **Adapt the question template**: The 12 questions above are a framework — tailor them to the specific topic
- **Let the Twins speak**: Include direct quotes from Digital Twins in findings — they carry authenticity and emotional weight
- **Don't lead**: Ask "What matters to you about X?" not "Don't you think X is important?"
- **Look for tension**: The most valuable insights often sit in contradictions

## Related Skills

- **icp-discovery**: If you need to find the right audience before researching them
- **feature-concept-testing**: If you want to validate a specific feature idea
- **brand-messaging-validation**: If you want to test how your brand resonates
- **creative-testing**: If you want to test specific creative concepts