claude research
Use these definitions when a readout starts to swell. The words are working terms for qualitative marketing research in B2B. They are not a taxonomy of an industry, and they are not permission to turn notes into a representative statistic or a promised return.
The qualitative sequence this guidebook teaches: interviews with people who can shape a B2B purchase, observation of markets and touchpoints you are allowed to see, a search for repeated friction, and a positioning synthesis. It is a method name, not a sample design.
A person who can advance, delay, or stop a purchase. Not a synonym for “someone who uses the product.” See decision-maker research for the split among buyer, champion, user, reviewer, and blocker.
The set of roles around one industrial or B2B decision. Classic descriptions, including Webster and Wind, treat it as a social unit. Claude research keeps the roles separate instead of averaging them.
A planned conversation about one past purchase, renewal, or abandonment. Questions follow the timeline and the artifacts. The interviewer does not pitch. Full method: customer interviews.
A prompt that supplies the flattering answer. “How innovative did we seem?” is leading. “What did you compare us with in the staff meeting?” is not. Leading questions ruin language capture.
Writing the buyer’s phrases as said, including awkward ones. Paraphrase belongs in a later editorial column, never in the quote column. Positioning should be able to point back to a phrase.
A record of behavior and artifacts in places you may legally and ethically be: consented calls, ride-alongs, public pages, materials your company already holds. Not covert surveillance. Method: market observation.
A contact you can tie to an artifact: a forward, a questionnaire, a pilot log, a redline. A stage name on a poster is not a touchpoint until an artifact exists. See touchpoint mapping.
A stall that repeats: resale, proof, review, or path. It is not a complaint and not a percentage. Detail: purchase friction.
What buyers already do instead of buying or instead of using you as designed. Renewing the incumbent, running a pilot on a side budget, or emailing a diagram by hand are workarounds. They are competitive context.
A single story. Useful for mechanism, unfit to represent others. Label the role. Keep it out of the headline.
The same mechanism across roles or deals, stated with a boundary. “Three stalled deals restarted on a missing diagram” can be a pattern inside that set. It is not proof about a segment. Process: pattern finding.
A sentence written by the researcher to show the shape of several notes, clearly labeled composite. It is not a fake quotation from a named person and not a statistic.
A memo that cuts: primary reader, real alternative, friction you will name, language to retire, and claims you refuse. It is not a slogan contest and not a forecast. See positioning synthesis.
A conversation focused on why a specific deal was won or lost, often near the decision. Claude interviews can include wins and losses, but they also stay with stalls that never became a clean loss code in the CRM.
A number that claims to describe a defined population within a stated error. This guidebook does not produce those numbers. Counting “four of the six files we opened” is a description of a folder, and it should be worded that way.
Choose the instrument for the question. Stacking all of them still does not create a representative market study.
Shows one decision’s sequence and language. Does not show that a segment shares the attitude.
Shows artifacts and absences you were allowed to see. Does not prove motive by itself.
Shows answers to a tight question on a real list. A rating does not explain a political stall.
Shows an editorial cut with limits. Does not show ROI or market share.
If two columns disagree, keep both in the memo and say which one guided the cut. Agreement is not the goal of the glossary. Fit between claim and method is.
Start with the conversation, then keep every later word smaller than the notes.
Read the interview method