--- title: "How to Run Customer Interviews Properly" description: "How to run customer interviews that surface real behavior instead of opinions: recruiting participants, writing questions, and turning notes into decisions." canonical_url: "https://builderscamp.com/guides/other/how-to-run-customer-interviews" date_published: "2026-09-16" date_modified: "2026-09-16" author: "Andre Albuquerque" publisher: "Builders Camp" guide_class: "other" --- # How to Run Customer Interviews **TL;DR:** Running a customer interview well means treating it as a continuous habit, at least one a week, not an occasional research sprint, and asking about specific past behavior rather than opinions or hypotheticals. A mix of power users, prospective users and churned users, synthesized into themes every couple of weeks, is what turns interview notes into an actual decision instead of a pile of unused transcripts. ## What makes a customer interview useful instead of just a conversation? The difference comes down to what you ask about. Questions about opinions and hypotheticals ("would you use a feature like this?") produce answers people cannot reliably predict about their own future behavior. Questions about specific past behavior ("tell me about the last time you tried to solve this problem") produce answers grounded in something that actually happened, which is far harder to get wrong or to answer with a polite guess aimed at pleasing the interviewer. A useful interview also has a clear hypothesis behind it before the conversation starts: a belief you are currently treating as true, stated specifically enough that a piece of evidence in the interview could actually change your mind about it. An interview run with no hypothesis in mind tends to produce a pleasant conversation and very little that changes a product decision afterward. ## How often should interviews actually happen? Continuous discovery, the practice popularized by product researcher Teresa Torres, treats customer interviews as a standing habit rather than a project: at least one interview per week, every week, rather than a burst of ten interviews before a big roadmap decision and then silence for the next two quarters. The weekly cadence matters because it keeps evidence current; a product and its users both change, and research done once a year goes stale well before the next research sprint happens. ## How do I find people to interview without a dedicated research team? The hardest practical part of continuous interviewing is recruiting, and the most sustainable fix is recruiting from inside your own product rather than relying entirely on cold outreach. A single, simple prompt to active users, asking whether they have 20 minutes to talk about their experience in exchange for a small incentive, is a common, low-friction way teams keep a steady interview pipeline running without a dedicated research operations function. The mix of who you talk to matters as much as how many. A useful interview pool includes: - **Power users**, who use the product heavily and can speak to advanced or edge-case behavior. - **Prospective users**, who represent your target market but have not adopted the product yet. - **Churned users**, whose reasons for leaving carry information neither of the other two groups can provide. ## How do I write questions that actually surface truth instead of politeness? Start from the hypothesis, not from a list of things you are curious about. Write down who does what, and why, according to your current belief, then ask what specific piece of evidence would change that belief if it turned out to be false. From there, build questions that probe behavior directly: "walk me through the last time you did X" surfaces far more reliable information than "do you usually do X," because the first forces a specific, checkable memory and the second invites a general self-image answer that may not match reality. Cut leading questions during review, before the interview happens. A question like "don't you find this frustrating?" tells the participant what answer you are hoping for, which contaminates the result before the conversation even starts. A [reusable discovery interview script](https://builderscamp.com/guides/templates/discovery-interview-script) is a faster starting point than writing every guide from scratch, as long as it still gets edited for the specific hypothesis at hand rather than reused verbatim. ## How do I turn a pile of interview notes into an actual decision? Individual interviews rarely change a roadmap on their own; synthesis across several interviews does. Periodically, roughly every couple of weeks, a team should look across recent interview notes together and cluster what was said into a small number of named themes, each one tied back to a specific quote or observation rather than a vague summary. Mapping those themes onto a visual structure connecting them to the outcome you are trying to move, an opportunity solution tree is the most commonly referenced format for this, keeps the discovery work connected to a decision instead of accumulating as unread notes in a shared folder. The step people skip most often is the final one: stating, explicitly, what decision the evidence now supports, weakens, or kills. An interview process that produces themes but never forces a decision statement has done research without doing discovery. ## Who should be involved, and should AI be part of the process? The strongest continuous discovery habit involves the full product trio, the PM, the designer and the engineer, in the interviews directly rather than having one person interview and summarize secondhand for the other two. Specific details that change a design or an engineering approach are exactly the details that get lost in a secondhand summary. AI tools can speed up the mechanical parts of this process: drafting an interview guide, generating behavioral follow-up probes, and clustering raw transcripts into candidate themes faster than a person doing it by hand. What AI should not do is replace the actual conversation with a real user, since the value of an interview comes specifically from evidence you could not have anticipated, which a model trained on existing text has no independent access to. ## Who should use this guide, and who should look elsewhere? This guide fits a product manager building or restarting a discovery habit, a researcher or designer trying to make interviews a continuous practice instead of a one-off project, and a founder validating a product decision before committing engineering time to it. It matches the workflow Builders Camp's AI Prompting for Customer Discovery bootcamp teaches directly: turning an assumption into a testable hypothesis, generating a sharper interview guide, and synthesizing evidence into a decision, using AI to speed up the thinking without replacing the actual conversation. It is not the right fit if you are looking for a one-time research sprint format instead of a continuous habit; the whole premise here is that interviews work best as a weekly practice, not an occasional event. It is also not a guide to quantitative research methods like surveys or A/B testing; those answer different questions than a qualitative interview can, and neither should be used as a substitute for the other. Themes that come out of a strong interview practice often point directly to a [North Star Metric worth tracking](https://builderscamp.com/guides/other/north-star-metric-examples), and once you have more opportunities than time to pursue, the [RICE prioritization framework](https://builderscamp.com/guides/other/rice-prioritization-framework) is a reasonable next step for deciding which one to act on first. ## Frequently asked questions ### How many customer interviews should a product team run per week? Product researcher Teresa Torres recommends at least one customer interview per week as a minimum sustainable habit, run continuously rather than in occasional bursts before a big decision. The goal is a steady stream of evidence, not a single research sprint that goes stale within a quarter. ### How do I find people willing to do an interview without a big research budget? Recruit inside your own product. The simplest and most sustainable method is prompting active users directly, for example asking whether they have 20 minutes to talk about their experience in exchange for a small incentive, rather than relying entirely on cold outreach or a recruiting agency. ### Should I only interview my most engaged users? No. A useful interview mix includes power users who use the product heavily, potential users who represent your target market but have not started using it yet, and churned users whose reasons for leaving carry information the other two groups cannot give you. ### What is the biggest mistake people make when writing interview questions? Asking about opinions and preferences ('would you use a feature like X?') instead of asking about specific past behavior ('tell me about the last time you tried to do X'). People are unreliable predictors of their own future behavior, but they can usually describe what they actually did last time accurately. ### How do I turn a stack of interview notes into an actual product decision? Periodically, roughly every couple of weeks, look across the interview notes together as a team and map the recurring opportunities onto a visual structure, an opportunity solution tree is one common format, that connects what you heard to a specific outcome you are trying to move. Interviews that never get synthesized this way just accumulate as unused notes. ### Who should be in the room for a customer interview: just the PM? Ideally not. Involving the product trio, the PM, the designer and the engineer together in discovery, rather than having the PM interview alone and report back secondhand, is a core habit behind sustained, continuous discovery. Secondhand summaries lose the specific detail that changes a design or engineering decision. ### Can AI help with customer interviews, or does it replace the human part? AI can speed up the parts of the process that are mechanical: drafting an interview guide, generating follow-up probes, and clustering raw notes into themes. It should not replace actually talking to a real user, since the value of an interview comes from evidence you could not have predicted in advance, which an AI model has no access to on its own. ## Sources - [Lenny's Newsletter: Teresa Torres on how to interview customers](https://www.lennysnewsletter.com/p/teresa-torres-on-how-to-interview) - [User Interviews: How to Interview Customers Continuously with Teresa Torres](https://www.userinterviews.com/blog/how-to-interview-customers-continuously-with-teresa-torres-of-product-talk) ## How this guide was made Researched from Builders Camp's bootcamp, track and masterclass material and the sources listed on this page, drafted with AI, and fact-checked against every source cited.