How do I conduct useful customer-discovery interviews?

Run customer-discovery interviews around a decision you have not made yet. Recruit people who experience, buy or influence the work, ask them to reconstruct a recent real occasion, and separate what they describe from your interpretation. Finish by deciding which assumption to test next—not by counting compliments.

The conversation can be warm and still be demanding. You want the person to feel comfortable telling you that the problem is rare, the current workaround is adequate or somebody else controls the budget. Those answers may save you from building a very polished solution to a problem nobody needs you to solve.

Choose the decision before writing the questions

“Learn about customers” is too broad to finish. Try a decision small enough that a finding could change it: whether to investigate a particular customer group, whether the problem occurs at the time you can serve it, or whether you have identified the actual buyer.

Imagine a fictional founder exploring a coordinated linen-pickup service for independent inns. The first idea is that properties with twenty to sixty rooms struggle with unpredictable Friday pickups and would pay for a more reliable arrangement. That range is a chosen research boundary, not a reported market segment or evidence of demand.

There are several assumptions inside the idea. Pickup unpredictability must occur. It must cause a consequential problem. Existing alternatives must leave something unresolved. The person bothered by it must have a route to a purchasing decision. Finally, the proposed business would have to supply a worthwhile improvement at a workable cost.

A first interview need not settle all of those. I would start with the first two and listen for who belongs in the next conversation. Otherwise, you can spend forty-five minutes discussing a hypothetical service while never establishing what happened last Friday.

The NSF's I-Corps description includes talking to potential customers, partners and other stakeholders in customer discovery. That wider cast matters: the user, payer and person who can prevent a purchase may be different people.

Write assumptions that can be contradicted

Turn the founder's broad idea into a short research note before recruiting:

• Event: a pickup arrived outside the agreed window during a recent busy period.
• Consequence: somebody had to do additional work, change service or absorb a cost because of that event.
• Current response: the property uses a workaround that still leaves a problem worth addressing.
• Decision route: someone at the property can explain how a different arrangement would actually be considered.

Beside each assumption, write what would weaken it. Perhaps pickup timing is reliable and the disruption comes from inaccurate stock counts. Perhaps the property has inexpensive spare linen and sees little reason to change. Perhaps it values reliability but cannot consider a new supplier during the proposed launch period.

These are possibilities to examine. They are not obstacles the interviewer must argue away. If every answer can be rewritten as support for the idea, the interview has become a sales exercise with unusually patient prospects.

Recruit for the experience you need to understand

Ask about the relevant role and recent experience when arranging a conversation. An owner who outsourced every linen decision may be less informative about the last pickup than the operations manager. A front-desk employee may describe the guest-facing consequence but have no purchasing authority.

For the fictional first round, the founder plans eight forty-five-minute sessions across six properties: four operations managers, two front-desk staff and two purchasing decision-makers. Two properties supply a second role. That is eight people, not eight independent businesses.

The mix is a deliberate starting point, not a statistically representative sample. Keep a note of how participants were found and the groups missing. If everyone comes from one friendly introduction, they may share a supplier, neighborhood or way of operating that the rest of your target market does not.

Use appropriate contact channels, explain the purpose and let people decline. Request only information necessary for the question. For this example, notes use property codes; there is no need to collect guest information, supplier account credentials or confidential contract copies.

The supplied research budget is $900. Twelve staff hours cover preparation, recruitment and the six hours of interviews; another six cover analysis and two cover administration. All twenty hours are paid at $35, totaling $700. Eight $20 participant honoraria add $160, and the existing secure-note process has $40 of incremental tool cost. These are invented complete costs, not recommended rates.

The founder has the $900 separately available and twenty paid staff hours scheduled without displacing accepted work. The honorarium is for participating, with no reward for a favorable opinion. Budgeting the work does not guarantee eight people will attend or that their answers will be useful.

Open with their last occasion, not your product

A useful opening explains what you are investigating without announcing the conclusion you hope to reach. For the inn case, the interviewer could say, “I'm trying to understand how linen pickup fits into a busy week. Could we walk through the most recent pickup you were involved in?”

Then stay with the sequence. When was it expected? What actually happened? Who noticed a difference? What did they do next? Ask what the participant personally observed and what they heard from someone else.

Use the person's answer to choose the next question. If the pickup was on time, there is no reason to force a late-pickup story. Ask what made the arrangement work and whether another occasion differed. If the person cannot remember, keep that uncertainty rather than supplying a plausible event for them.

These original prompts can help when the conversation becomes general:

1. “What happened the last time that affected your day?”
2. “What did you do instead, and who else became involved?”
3. “Which part was difficult enough that you changed something?”
4. “What did the current supplier or your own team do to resolve it?”
5. “How was the last decision about this service made?”
6. “What would have to happen before a different arrangement could be considered?”

Do not turn the list into an interrogation or insist on a dollar figure the participant cannot know. A description of forty minutes rearranging work may be better evidence than an invented estimate of “thousands lost.” Record the unit and whose time it was before trying to value it.

Pew Research Center's question-writing guidance emphasizes clear wording and the effects of question design. A qualitative interview needs its own conversational flexibility, but “Was our faster, cheaper and easier idea helpful?” still bundles several judgments and supplies the desired answer.

Keep the pitch out of the evidence

If you explain the proposed service early, participants may spend the rest of the meeting helping you improve it. That can be generous and useful without establishing that the original problem matters.

Finish the experience questions first. If you later show a concept, mark the transition in the notes. A response to an advertised feature is different evidence from the account given before the feature was mentioned.

Similarly, “I'd probably use that” is not a purchasing commitment. You can ask what would be required to evaluate it, but do not turn politeness into a forecast order. The first-paying-customer guide handles the separate step of making a defined offer and learning from a real purchase.

A participant may ask for an immediate solution. Be clear about what exists and what does not. This research round has not created a pickup operation, priced its routes or established that it can accept an order.

Write notes that preserve the disagreement

Use separate fields for the event described, the participant's interpretation, your inference and the next question. That small separation prevents a sentence such as “Friday pickup is broken” from replacing several different accounts.

The following notes are invented illustrations for the first four sessions in the fictional research plan. They are not quotations or reports from actual interviews.

Four different accounts, four different implications

Scroll across to compare the described event with the inference still to test. Use the arrow keys when the table area is focused.

Only three properties appear in those four accounts. P1's two roles add depth to one organization's story; they are not two separate customer opportunities. The remaining planned sessions have not happened in this illustration, so they supply no findings.

The evidence already challenges the original Friday-specific framing. One event occurred Thursday, another concern arose from stock control, and one property had a satisfactory buffer. Keep those differences. Compressing them into “four people discussed linen problems” would make the research sound positive while discarding what it taught.

Choose the next question the notes actually justify

Before the next session, update the research note. You might investigate properties without a spare-linen buffer, ask how pickup and counting errors are distinguished, or learn which buying windows fit an eventual service. Explain why that next conversation could change the decision.

Do not quietly rewrite the original assumption and then claim it was confirmed. Keep the earlier version and the evidence that changed it. A useful research history lets a colleague see why the idea narrowed rather than inherit a confident slogan with its awkward findings removed.

Interviews also have a stopping point. If the consequential unknown is now whether a route can arrive on time at the required cost, another conversation may not settle it. Design a suitably bounded operational test. The business-trial guide examines whether the test faces the conditions the final operation would actually need to handle.

If the unknown is whether someone will pay, an appropriate defined offer can be more informative than another hypothetical-price question. Keep the delivery obligation attached to that next step; discovery does not excuse selling a service you cannot provide.

Hand over the learning, not an endorsement count

A short research handoff should show the decision, original assumptions, participant roles and recruitment limits, relevant event notes, contradictions and the next action. Preserve uncertainty in ordinary language: “We have not spoken with the budget owner” is more useful than a high-confidence score nobody can explain.

The served-customer follow-up guide addresses a different setting: examining an existing service group and whose replies are missing. This discovery round explores a proposed problem and purchasing route before an offer exists. It cannot establish market prevalence, conversion rates or a lifetime of future customers.

In Infinite Upside, the expansion discussion pulls apart the steps between customers existing in an area and those customers choosing, paying for and sustaining a particular operation. That is the spirit worth carrying out of these interviews: give each assumption a chance to be wrong while changing course is still possible. Read the opening for the owner's wider story of turning an appealing idea into commitments to other people.