Customer Discovery for Startups: Interview Questions and a Test Plan

Write down what you need to learn: who coordinates a particular workflow, how a purchase gets approved, or why a recent workaround failed. Avoid research objectives that already assume the answer, such as proving that customers want your dashboard. Your study should be capable of changing your customer definition or product direction.

NSF’s I-Corps guidance for Phase I awardees explicitly connects customer-discovery interviews with market evidence and engagement across the surrounding ecosystem. That illustrates why the conversation extends beyond the person who might click a button. Read the NSF guidance. For your own study, identify users, budget owners, implementers and anyone who can block adoption.

Begin with people who recently performed the task or adopted an alternative. Include different ways of handling it, rather than recruiting only people already enthusiastic about your idea. Ash Maurya’s discussion of interviewing mistakes warns that premature assumptions about an ideal early adopter can narrow discovery too far. Read his prospecting guidance.

For a hypothetical supplier-document tool, recruit someone using spreadsheets, someone using an established platform and someone who recently abandoned a new process. Record how each participant was recruited. Warm introductions may be convenient, but they can overrepresent one network. Explain the purpose and expected length of the conversation, ask permission before recording, and avoid requesting confidential employer documents.

Open with a recent situation and follow its sequence. Ask the participant to describe the trigger, the first action, the people contacted and the eventual result. When an answer becomes general, return to the specific event. If someone says the process is always slow, ask what happened during the last occasion and which step created the delay.

Keep any demonstration for a clearly marked second part of the session. Otherwise you risk hearing a reaction to your pitch instead of an account of the existing workflow. You can acknowledge an idea without defending it. Leave room for a participant to say that the problem is unimportant or that an alternative already works well.

Imagine a procurement coordinator describing a supplier onboarding delay. The coordinator reports sending reminders, but the vendor’s document was actually waiting for internal approval. A founder who hears only the word reminders might build automated email. A better record distinguishes the visible symptom from the approval dependency and identifies the person who understands that dependency.

Do not turn an interpretation into an invented quotation. Capture a paraphrase as a paraphrase and a direct quotation only when accurately recorded with permission. Then add a separate inference and a next question. This separation makes it easier for another teammate to challenge your reasoning without arguing about what the participant said.

After each conversation, record the participant’s context, recent event, workaround, consequences and unresolved questions. Group similar observations without erasing contradictory cases. If one segment describes a costly delay and another does not, preserve the difference. A simple matrix is often more useful than a single score summarizing every interview.

At the end of the study, state what you learned, what remains uncertain and what action follows. You might interview a missing stakeholder, narrow the problem, test a workflow or stop pursuing the idea. Use the idea-validation checklist for that broader decision. If support would help you plan the next step, Idea2Unicorn describes guidance for early ideas and prototypes.

Explore Idea2Unicorn if you want structured support for your early customer-learning plan.