How to run a customer interview that actually changes what you build and sell

Interviews drift when the team cannot agree what they are trying to learn. Before you touch a calendar, write a single-sentence research question that names the decision the research is going to inform. Good examples look like "what forces push mid-market revops leaders to replace a legacy CRM in the first 18 months of a new role" or "why do self-serve trial users who activate still fail to invite a second seat." A sharp question lets you cut the sample, pick the script, and later judge whether the research actually paid for itself. A fuzzy question such as "learn about our customers" produces a transcript library that nobody reads and no decision follows.

• Write the research question in one sentence and share it with product, marketing, and sales before recruiting.
• Name the decision the research will inform, such as a positioning change, a packaging change, or a roadmap bet.
• Agree in advance what a strong finding looks like so the debrief does not relitigate the scope.

The sample decides what you can learn. Aim for eight to twelve participants spread across the segments that matter for the research question, with a deliberate mix of active customers, recent churns, lost prospects, and users of the workaround your product replaces. Weight the sample toward the ICP tier you actually want to grow, not the one that happens to answer email fastest. Published research from Bob Moesta on jobs-to-be-done and from Teresa Torres on continuous discovery both make the same point from different angles: the switch stories you need live in the lost and churned cohort, not in the champion cohort, and a sample that leaves them out will mislead the team.

• Draw at least one third of the sample from lost deals, churned accounts, or former users of a workaround.
• Balance by ICP tier, company size, and role so themes can be read by cut rather than aggregated.
• Recruit through a neutral channel such as a researcher or product marketer, not the account team.
• Offer a modest incentive such as a $100 charitable donation or gift card to lift response rates in the hardest-to-reach cohort.

Standardize the script so you can compare across interviews, then keep every question open so the participant tells the story in their own words. Group the questions into four blocks: context, the switch moment, the forces at play, and the outcome they were trying to produce. Keep the whole script to a single page. Avoid leading language such as "what did you love about us," which trains the participant to flatter rather than inform, and avoid feature-level questions in the first two thirds of the call because they short-circuit the story. Save any rating scales for a short closing section so you do not interrupt the narrative with scales mid-flow.

• Context: how the project started, who was involved, and what triggered the search.
• Switch moment: the first time they thought the old way was not going to work.
• Forces: what pushed them away from the old, pulled them toward the new, held them back, and worried them about change.
• Outcome: what better looked like in their own words and how they would know they had it.

The first 10 minutes of a customer interview are not throwaway minutes. They set whether the participant tells you a real story or a sanitized one. Open with their role, how their team is organized, and what the project you are researching looked like from their seat. Ask who else was involved and how decisions got made. The purpose is twofold: you need the context to interpret anything that comes later, and the participant needs the warm-up to stop performing for you. Reforge research-methods materials and Teresa Torres both emphasize this early-call investment because the quality of every answer later in the hour depends on the trust and specificity built in the first stretch.

• Start with the person, not the product: role, team shape, what a typical week looks like.
• Anchor on a specific project or decision so the participant tells a story, not an opinion.
• Ask who else was involved and how they influenced the decision so you can map the buying group later.

The heart of a jobs-to-be-done interview is the moment the participant decided the current approach was not going to work. Walk them backward and forward from that moment in small steps. What were they doing just before? What did they try first? Who did they talk to? Bob Moesta's four forces model is the sharpest tool here: push from the current situation, pull of the new solution, anxiety about change, and habit holding them back. All four show up in every real switch. If you only hear pull, you are getting a sales pitch played back at you. If you only hear push, you are missing why they chose your product rather than the next option on the list.

• Push: what about the old way had become intolerable right before the switch.
• Pull: what the new way promised that the old way could not.
• Anxiety: what worried them about picking the new tool or process.
• Habit: what made the old way comfortable even once they knew it was failing.

Features change, jobs do not. Once you have the switch story, slow down and probe the job the participant was really hiring a solution to do, the outcome they wanted to produce, and the anxiety that nearly stopped them. Harvard Business Review's long-running work on jobs-to-be-done frames this as separating the functional job (what they need to get done) from the emotional and social job (how they want to feel and be seen while getting it done). Ask what better looks like in their own words. Ask what they would show their boss to prove the project worked. Ask what the worst outcome would have looked like, and what they did to protect against it. Those three questions almost always surface the outcome and the anxiety the product has to speak to.

• Functional job: what work are they trying to get done, in their own language.
• Outcome: how they will know, concretely, that the job got done well.
• Emotional and social job: how they want to feel about the choice and how they want others to see it.
• Anxiety: the worst-case picture they were defending against when they picked a vendor.

A single interview is an anecdote. Eight is a dataset. After each call, hold a 20 to 30 minute debrief with the interviewer, note-taker, and at least one teammate from the function the research is going to inform. Tag the transcript against a small, stable set of theme codes such as push forces, pull forces, anxieties, habits, functional job, outcome, buying group, and workaround. At the end of the research window, roll the themes up across the sample and separate findings by cohort so wins, churns, and lost prospects each tell their own story. The deliverable is a short written synthesis with the top three to five themes per cohort, representative anonymized quotes, and a prioritized list of actions by owner.

• Debrief within 24 hours while the specifics are still fresh and the quotes are easy to pull.
• Tag each transcript with 4 to 8 theme codes from a shared taxonomy, not a free-form list.
• Separate findings by cohort because wins, churns, and no-decisions almost always point in different directions.
• Pull two or three verbatim anonymized quotes per theme so the final report has human texture, not just percentages.

Research that never changes a decision is wasted budget. Close the loop by assigning each theme to a single accountable owner with a date for the first visible action. Product owns functional-job and outcome themes. Marketing owns pull-force and emotional-job themes because they drive positioning and messaging. Sales owns anxiety and habit themes because they drive objection handling and discovery questions. Customer success owns workaround and buying-group themes because they shape onboarding. Publish a short executive summary to the whole go-to-market team, not just leadership, and open the next quarterly readout with a status check on last round's commitments. Over two or three rounds, the research program becomes the most trusted source of voice-of-customer evidence in the business.

• Assign each theme to a single accountable owner in product, marketing, sales, or customer success.
• Attach a specific action and a target date for the first visible change in the product, the site, or the pitch.
• Publish a one-page summary with ranked themes, verbatim quotes, and the committed actions.
• Review progress on last round's commitments at the top of the next readout so the program visibly earns its keep.