Interview the person who can refuse the purchase

Name the buying group before you book the hour

B2B purchases are argued by a group, even when one signature lands on the contract. Before you recruit, write the roles you need to hear, not the logos you wish you had. An economic buyer controls budget or can stop it. A user lives with the tool after the kickoff cake. A champion spends internal political capital. A blocker — security, legal, procurement, finance — can hold the file without ever becoming a “detractor” in a survey. A technical evaluator writes the shortlist and then loses the room.

A vice president title does not prove any of those roles. Ask, in the first five minutes, who else had to agree the last time they bought something in this neighborhood of problem, and who could have killed it without a meeting. If your sample is only champions who already like you, you will hear the speech they use on friendly vendors. The glossary keeps these roles from being flattened into “the customer.” How you get access without only calling winners is part of research design.

A sequence that stays on the last real purchase

Plan forty-five to seventy minutes. One decision. No hypothetical roadmap. The Claude sequence below is a path, not a script to recite. Follow their chronology when they offer it, then come back for the gaps.

1. What they are accountable for if a purchase of this kind fails in public.
2. The last time they bought, renewed, or refused something aimed at the same job. Dates, rough sequence, and who started it.
3. The moment a complaint became a project: a ticket spike, a lost customer, an audit, a new executive, a contract coming due.
4. Who was in the room, who was only on the thread, and who appeared late.
5. Where the file sat still, and what artifact was open while it sat: a security questionnaire, a redline, a pilot scorecard, a reference call.
6. The words they used internally when they argued for it, and the words used against it. Ask them to say the sentence again.
7. What would have made them walk away, including a condition that never quite arrived.
8. What they did instead if they did not buy: the incumbent, a spreadsheet, an agency, or nothing.

Do not ask what return on investment they require. That question invites a slogan and teaches you their slide template, not their fear. The stalls you are hearing have names. They are sorted in purchase friction. The sentences belong in a language bank, marked as verbatim.

What you write down, and what you keep in the margin

During the call, capture short verbatim lines in quotation marks and mark everything else as paraphrase. Your hunch — “this is really about a failed implementation two years ago” — goes in a margin labeled interpretation. If the hunch is doing the work of evidence, you do not yet have a finding. After the call, send back facts you might have wrong: dates, titles, sequence. Do not send your conclusion for approval. A buyer who corrects your story is a source. A buyer who edits your strategy is a stakeholder you have accidentally hired.

“We did not lose a quarter arguing about features. We lost it because nobody would sign the sentence that said the pilot had worked.”

Composite sketch, built from the shape of several stalled evaluations. It is not a quote from a named account, and it is not a rate.

Consent, and the limit printed on the memo

Say what the conversation is for, who will read the notes, and whether you are recording. Offer a path to strike a line. Do not pretend to be a buyer, a student with no client, or a journalist if you are none of those. A Claude memo opens with the limit: these interviews describe the decisions you were allowed to hear. They do not represent a segment, a country, or a quarter’s pipeline. If you need the places the decision moved when nobody was narrating it, continue into market observation.