Claude · B2B research
How a company actually decides to buy
funstir is a field guide for people who study B2B purchases. The method in these pages is called Claude: a long interview with someone who can approve, delay, or refuse the purchase; a look at the touchpoints that person already uses; and a written account of the friction that repeats when a buying group tries to say yes. Claude work produces positioning language a buyer would recognize. It does not forecast profit, and it does not treat a short run of conversations as a picture of a whole market.
8
Chapters in this guide, from the interview room to the sentence you are willing to publish.
3
Moves only: interview a decision-maker, observe a touchpoint, synthesize the pattern in their words.
0
Promises of return on investment, and zero claims that a Claude sample represents a market.
Industrial marketers learned categories from people who had sat in the room: a plant manager arguing with finance, a security lead reading a questionnaire aloud. The notes were partial. They were also about a real decision. Panels and intent scores later replaced that room. A dashboard can show that an account opened a pricing page. It cannot show who is still angry about the last failed pilot, or which sentence in a security form sends the file to legal for a month.
Claude is a return to that room with stricter habits. You ask about one purchase that already happened, keep what people said separate from what you hope it means, and watch a touchpoint you were allowed to see. The memo says, up front, that the conversations are not a representative sample and not a forecast of revenue.
The rest of this guide is that habit, split into workable parts. Decision-maker interviews reconstruct the last purchase. Market observation and touchpoint fieldwork show where the decision is actually handled. Purchase friction names the stalls. Decision language keeps the buyer’s phrases intact so positioning synthesis does not drift into a slogan.
Three ways in
Pick up the study at the move you understand least
01
Interviews with the person who can refuse
A title is not authority. The Claude sequence stays on the last real purchase: who had to agree, where it stalled, and what they did instead, including nothing.
02
Touchpoints already in use
Security reviews, renewal calls, shortlist spreadsheets, peer threads. Observation records sequence and silence. It does not invent a conversion rate.
03
A sentence the buyer would repeat
Synthesis turns repeated friction into language. It is tested by being read back. It is not a claim about payback or market share.
A working order
Four weeks, one decision, a memo with limits on the first page
A Claude study is small on purpose. You name one decision, the roles who touch it, and the friction you cannot yet describe. You recruit beyond happy customers: a loss, a delay, and someone who kept the old tool. The memo leads with what the design cannot support.
- Research design — question, access, and the paragraph that says what you will not claim.
- Observation — where to watch without pretending you saw the market.
- Friction patterns — proof, consensus, switching, timing, and category doubt.
- Glossary — shared words so a buying group is not flattened into “the customer.”
Keep reading
The rest of the shelf
Phrases
Decision language
Why “we got burned last time” should survive your rewrite, and how a language bank feeds positioning without polishing the fear out of it.
Glossary
Buying group, blocker, saturation, composite sketch, and the word claude as this guide uses it.
Observation
Public artifacts and invited rooms, and the difference between a signal and a proof.
Where purchases stall
A worked composite of an evaluation that sat in security because nobody owned the thing the pilot was supposed to prove.
How notes are handled
The privacy policy for this site: newsletter address, local consent, and what funstir does not sell.