# Customer Research

> Plans, runs, and synthesizes customer research — interviews, surveys, win-loss analysis, and message testing — into findings that change decisions. Use this when the customer is being guessed at, before positioning or a major launch, when churn or conversion has an unexplained cause, or when a team is arguing from anecdotes. Also use to turn raw interview notes into usable findings.

- Skill: `majiayu000/customer-research-2` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/customer-research-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/customer-research-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Marketing & Growth
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/customer-research-2

---


# Customer research

Research earns its cost only if a decision hangs on it. Name that decision before recruiting anyone.

## Choose the method by question

- **Why did this happen / what do they actually do** — interviews. Small n, deep.
- **How common is this** — survey. Only after interviews have told you what to ask; a survey written
  from assumptions returns your assumptions.
- **Why did we win or lose** — win-loss, with losses weighted higher. Losses are where the truth is
  and are systematically under-sampled.
- **Which framing works** — message testing against a real choice, not a preference rating.

## Interviewing

Ask about the past, not the future. "What did you do last time" is evidence; "would you use this"
is politeness.

- Open broad, follow the energy, and go quiet after their first answer — the second one is where the
  substance is.
- Chase specifics: when, what happened next, what did that cost you, who else was involved.
- Never describe your solution before you have their problem in their words. Once they know what you
  want to hear, the interview is over.
- Watch for the workaround. Anything someone has built a spreadsheet to survive is a validated
  problem.

## Synthesizing

Code the transcripts before forming a view — findings assembled to support an existing belief will
be. For each finding record: the claim, how many independently said it, a verbatim quote, and what
would change if it is true. A finding that changes nothing is trivia.

Separate what people **said**, what they **did**, and what you **inferred**. Conflating the three is
the most common way research misleads.

## Never

- Recruit only happy customers. The sample that answers your email is not your market.
- Lead: "How much would this feature help?" has one answer.
- Present a single interview as a finding.