heath-gtm Customer Research
What this does
Takes a research goal and builds the whole study: who to interview and why they are the right people, a non-leading interview guide, the questions that get past the polite answer to the real one, and a method for turning the transcripts into positioning and messaging you can use. It designs and structures the research. It does not sit in the interviews for you.
What you'll need
You do not need to connect anything to get value today. Bring your goal and the skill runs now. Connect the tools below and it helps you find the right people and analyze what they said.
- Works today with: what you are trying to learn (why they bought, why they churned, how they say it in their words), and a rough sense of who your customers are. Paste it in.
- More powerful connected to a CRM: it helps you pick a representative slate by segment, stage, and outcome instead of just the friendly accounts.
- Sharper with a transcripts or notes source: paste interview transcripts and the skill pulls themes, quotes, and language patterns back out.
- Sharper with a survey tool: pairs the qualitative interviews with quantitative validation of what you heard.
How this runs at your connection level
This skill is never reliant on a connector. It runs on the goal you give it today and gets more powerful as you connect tools. It never invents a quote or a finding it did not hear. A theme is only a theme if the transcripts support it.
- Bring your data: paste the goal and who your customers are. The skill builds the recruit criteria, the interview guide, and the analysis method today. No connection required.
- Connect your tools: the same skill helps build a representative slate from the CRM and pulls themes and verbatim quotes from transcripts you paste.
- Just exploring: no study yet? Get the framework, the exact inputs it reads, and a worked example guide, so you can see what good looks like before you book a single call.
Every run ends with the one thing that would make the next run sharper, a segment to include or a tool to connect.
Customize this for yourself
This was built for a B2B team running win, loss, and positioning interviews with their own customers. Set these to your stack:
| Set this | What it is | Default / Example |
|---|---|---|
| GOAL | what the study must answer | why they buy, why they churn, how they describe us |
| SEGMENTS | who you sample across | by size, industry, use case, outcome |
| SAMPLE_SIZE | how many interviews | 8 to 15 for qualitative saturation |
| RECRUIT_SOURCE | where the slate comes from | CRM, customer list, warm intros |
| GUIDE_LENGTH | interview length | 30 to 45 minutes |
| TRANSCRIPT_SOURCE | where recordings or notes live | your notes tool or call recorder |
| VALIDATION | how you confirm the themes | a follow-up survey, optional |
The point of the study is your customers' words. The skill will never invent a quote or attribute a finding to someone who did not say it.
The method
Who to talk to (representative, not convenient)
Pick a slate that can actually answer the GOAL, sampled across the SEGMENTS that matter. If you only interview happy, engaged customers, you learn why happy people are happy. Include the ones who churned, the ones who almost did not buy, the quiet ones. Name why each type is in the sample.
The interview guide (do not lead the witness)
Write open questions that let the customer tell their story before you steer it. Start broad ("walk me through how you ended up looking for something like this"), then follow their thread. The rule of the guide is that it collects their language, not confirmation of yours.
Questions that avoid leading
The craft is in the wording. "What did you love about the product?" plants the answer. "What happened right before you decided to buy?" lets them tell you. The skill flags leading questions, double-barreled questions, and yes-or-no traps, and rewrites them to open the story. Ask about behavior and specific moments, not opinions and hypotheticals.
Run and capture
Keep the interviewer quiet and curious: ask, then let silence do the work. Capture verbatim language, the exact words for the problem, the alternatives, the moment of decision. Record or take clean notes, because the quotes are the deliverable, not your paraphrase.
Transcripts into positioning and messaging
Turn the raw transcripts into inputs: recurring problems in the customer's words, the language they use (which becomes your copy), the alternatives they weighed (which becomes your competitive frame), and the moment and trigger that made them buy (which becomes your positioning). Every theme is backed by real quotes, counted across interviews, never a single loud voice generalized into a finding.
Quality gates
- The sample includes more than the happy, convenient accounts. Churned and near-miss voices are in it.
- Every question in the guide is open and non-leading. The skill flags and rewrites any that plant an answer.
- Every theme is backed by verbatim quotes and a count of how many interviews support it.
- No invented quotes, no findings attributed to people who did not say them.
Output (example)
CUSTOMER RESEARCH DESIGN · positioning study (illustrative)
Goal: how do buyers describe the problem in their own words, and why us over the rest.
Slate (12 interviews):
4 recent buyers · 3 long-time customers · 3 churned · 2 near-misses (chose a competitor)
Guide (35 min), opening questions:
1. Walk me through what was going on right before you started looking.
2. What had you tried before this? (not: "why didn't other tools work?")
3. Tell me about the day you decided. What tipped it?
Flagged + rewritten: "What do you love about us?" -> "When has it clearly paid off?"
Analysis output (from transcripts):
Theme Support Verbatim
"manual + slow" 9/12 "I was doing it by hand every Friday"
Trigger = new hire 5/12 "we grew and it just broke"
Chose us for X 7/12 their words, pulled straight from the call
Positioning inputs: lead with the Friday-by-hand pain, in their language, not ours.
Next move: connect the CRM to build the churned + near-miss slate you cannot get from memory.
Where the inputs come from
The goal and the customer picture come from you. A representative slate by segment and outcome comes from a CRM once connected. Themes, counts, and verbatim quotes come from transcripts or notes you paste. Optional quantitative confirmation comes from a survey tool. Every finding is traceable to a real quote, and the skill will not manufacture one.
Example chat
Here is the skill running on a real positioning study.
You: We want to know why buyers pick us over the alternatives, in their own words. Design the study.
Claude: Study's designed, and the sample is the part most teams get wrong.
Slate (12 interviews):
4 recent buyers · 3 long-time customers · 3 churned · 2 near-misses
Guide opening question:
"Walk me through what was going on right before you started
looking." (not: "why did you choose us?")
Interviewing only happy customers only teaches you why happy people are happy, the churned and near-miss voices are what actually explain the decision. The opening question is deliberately broad so the buyer tells their own story before anyone steers it toward a compliment. One sharpener: connect your CRM so the near-miss and churned slate is pulled from real accounts, not a guess at who those people even were.
You: I ran 6 interviews. Here are the transcripts, pull the themes.
Claude: Pulled, and one theme is stronger than your current homepage copy.
Theme Support Verbatim
"manual + slow" 5/6 "I was doing it by hand every Friday"
Trigger = new hire 3/6 "we grew and it just broke"
"Manual and slow" showing up in 5 of 6 interviews, in nearly identical language, is a strong enough pattern to lead your positioning with, not bury in a feature list. That phrase, "by hand every Friday," is more specific and more human than anything currently on the site. One sharpener: run 4 to 6 more interviews to hit full saturation, right now this reads as a strong lead, not yet a locked finding.
Go further
The read is step one. Here's where an operator takes it once the manual version proves out.
- Build the slate without the manual pull. Connect Salesforce so every future study auto-generates a representative sample across segment and outcome, not just whoever's easiest to reach.
- Turn transcripts into copy on a cadence. Point a scheduled Claude task at your notes tool monthly and refresh the theme list as new interviews land, so messaging tracks live language instead of a study that ages out.
- Validate the theme at scale. Push the top qualitative themes into a quick survey tool to confirm the pattern holds beyond 6 interviews before it becomes the homepage headline.
You built the read once; now it runs itself.
Make it yours
Fork it. Change the segments, the guide, the number of interviews, the way you code themes. The point is not to run a generic study. It is to hear your customers in their own words and turn those words into positioning that lands. Built by an operator. Customize it, break it, make it better.