The discovery interview script we actually use

People ask for our interview questions more than they ask for anything else, so here they are. This is the script we use in the first week of a discovery sprint. It is short on purpose, because a discovery interview is forty minutes and most of that should be the other person talking. The questions are prompts to keep them going, and the phrasing matters more than the order.

The rule behind every question

Ask about things that already happened. Never ask someone to predict what they would do. Rob Fitzpatrick's The Mom Test makes this case better than I can: people are polite, and when you ask "would you use this," they say yes to be kind. When you ask "when did this last happen," they have to remember, and memory is where the truth lives. Every question below is built on that rule.

Opening: get them talking about the problem, not you

"Walk me through the last time you dealt with this." That is the whole opener. For Dwell it was "walk me through the last time you looked for a room." For ShiftFlow it was "walk me through building the last roster." If they give a summary, ask for the story: what day was it, what did you open first, who did you message. The detail is the data. Ashmita's "twenty rooms over six months, mostly dingy homes" came out of this question, and it changed the product.

Follow with "what did you do about it?" and then "what happened after that?" Keep pulling the thread until you reach the moment they gave up, paid someone, or built a workaround. The workaround is the most useful thing you'll hear all interview. It's the price they're already paying.

Middle: find the cost

"How much time did that take?" and "did it cost you anything?" are blunt on purpose. You are trying to find out whether the problem is expensive enough to be worth solving. A problem that costs someone twenty minutes a month is not a business. A problem that costs a store manager most of a day every two weeks, as rostering did before ShiftFlow, is.

Then: "have you tried anything else?" If the answer is no, be careful. A problem nobody has tried to solve is usually a problem nobody cares about enough. If the answer is a spreadsheet, a WhatsApp group, or a paid tool they abandoned, you have found both the competition and the bar you have to clear.

Later: test the model, not the mock-up

We do not show designs in discovery interviews. We describe the model and let them poke holes. For Dwell, the question to a realtor was "if a verified renter came to you with a trust score, what would you do with it?" His answer, that marketing on Facebook Marketplace is free so there is no reason for a small landlord to pay per unlock, took out the pricing model in one sentence. A screen mock-up would have gotten a compliment instead.

Two questions that help here: "who else would need to say yes to this?" and "what would make this a no for you?" The first finds the hidden buyer, which for Dwell turned out to be property management companies rather than individual owners. The second gives people permission to be negative, which they otherwise won't be.

Closing: get the next interview

"Who else deals with this?" Every interview should produce the next one. If they can't name anyone, that is also a finding. And "can I come back to you when we have something to show?" gets you a tester for the prototype stage without a recruiting round.

The two questions we stopped asking

"Would you pay for this?" and "what features would you want?" Both produce confident, useless answers. The first gets a polite yes that has never once predicted an actual payment. The second gets a wish list from someone who has spent five minutes thinking about a problem you have spent five weeks on. If you want to know about price, ask what they pay for the workaround today. If you want to know about features, ask what they did last time and build for that.

What the notes look like afterwards

Each interview becomes one page: who they are, the story they told, the cost they named, what they've tried, who else has to say yes, and one sentence on what this changes in the plan. That last line is the point. If six interviews don't change anything, either the plan was already right, which is rare, or the interviews weren't asking hard enough questions. In Dwell's case, two interviews changed two things. That is a good ratio, and it is why the build is still waiting on the other seven.

Sources and further reading

• The Mom Test, Rob Fitzpatrick. Most of the phrasing rules here come from it.
• Continuous Discovery Habits, Teresa Torres. On interviewing for stories rather than opinions.
• Dwell case study. Interview notes from Ashmita and the realtor.