Customer Discovery Interviews: A 20-Minute Script and Scoring Sheet

TL;DR. To do customer discovery interviews well, ask about the last time a problem actually happened, not whether someone would like your idea. A 20-minute interview needs five parts: context, the last occurrence, what they did, the alternatives they tried, and a close. After each call, score the notes on strength of evidence. Past behaviour scores high. Opinions and predictions score low. Decisions then rest on what people did, not on what they said they might do.

What is a customer discovery interview for?

A discovery interview tests whether a problem is real, frequent and costly enough that someone already spends time or money on it. It is not a sales call or a pitch. It is not a feature wish-list session either.

The useful output is evidence: specific events, actions, workarounds and costs that the person experienced. Compliments and "I would definitely use that" feel encouraging, but they cost the speaker nothing and predict little. The script below keeps the conversation on events that have already happened.

If you have not yet found people to talk to, start with Recruit B2B Customer Interview Participants.

How should a 20-minute interview be structured?

Plan for five blocks. The times are a guide, not a stopwatch.

Block	Time	Goal	Example questions
1. Context	3 min	Understand their role and how their work fits together	"Walk me through what your week looks like for [area]." "Who else is involved?"
2. Last time it happened	5 min	Anchor on one real, recent event	"Tell me about the last time [problem] came up." "What triggered it?"
3. What they did	5 min	Capture actions, effort and cost	"What did you do next?" "How long did that take?" "Who had to get involved?"
4. Alternatives	4 min	Find out what they already use or pay for	"What have you tried to fix this?" "Why did you stop using it?" "What does it cost you today?"
5. Close	3 min	Confirm, ask for a commitment and referrals	"Did I miss anything important?" "Could you share the spreadsheet you mentioned?" "Who else deals with this?"

Rules that keep the script honest

• Do not describe your solution until the close, if at all. Once people know what you are building, they start reviewing it politely.
• Follow the story, not the list. If the last occurrence is rich, spend more time there and shorten context.
• Turn generalities into an instance. When someone says "we usually", ask "can you think of the last time?"
• Ask about money and time already spent. Spending reveals priority better than stated importance.
• Ask for something small at the end. A shared document, an introduction or a follow-up call is a modest test of real interest.

How do you score an interview on strength of evidence?

Score each interview as soon as it ends, while the notes are fresh. Use five criteria, each rated 0, 1 or 2.

Criterion	0 – weak	1 – moderate	2 – strong
Specific occurrence	Only general statements or hypotheticals	A vague recent example	A dated, detailed event they walked through
Action taken	Did nothing; just finds it annoying	Some informal effort	A clear process, workaround or escalation
Cost or effort	No cost mentioned	Effort described but not sized	Time, money or people clearly committed
Existing alternatives	None tried	Tried something casually	Built, bought or replaced a solution
Commitment at close	Polite interest only	Agreed to a follow-up	Shared material, made an introduction or offered access

The maximum is 10. Treat the total as a sorting aid, not a measurement: it tells you which interviews to weight more heavily when you look for patterns. Also note one line on what the person said they want, kept separate from the score, so opinions are recorded without inflating the evidence.

What does scoring look like in practice?

The three interviews below are illustrative notes for a hypothetical tool that helps finance teams chase overdue supplier documents. All details and scores are invented for the example.

Interview A, operations lead. "This is a huge pain, I would love a tool for it." Asked about the last time, she could not recall a specific case. Her team "just deals with it". No tools tried. Happy to "see a demo sometime".

Interview B, finance manager. Described month-end two weeks earlier: three supplier documents missing, two days of email chasing, a late close reported to the finance director. Built a shared tracking spreadsheet last quarter. Shared it at the close and introduced a colleague.

Interview C, controller. Mentioned a missing document last month, resolved with a few emails. Previously tried a reminder feature in an existing tool but stopped. Agreed to a follow-up call.

Criterion	A	B	C
Specific occurrence	0	2	1
Action taken	0	2	1
Cost or effort	0	2	1
Existing alternatives	0	2	1
Commitment at close	0	2	1
Total (illustrative)	0	10	5

Interview A sounded the most enthusiastic and scored lowest. Interview B was calm and factual, and it supplies the strongest evidence: a recent event, real cost, a homemade workaround and a concrete commitment. Interview C suggests the problem exists but may be minor for some teams. The next step is to find more people who look like B and ask what made their situation costly, rather than counting how many people said the idea was good.

Where should the notes and scores live?

Store each interview's notes, score and the separate "what they want" line in one place, tagged by segment and problem. That makes it possible to compare high-scoring interviews across weeks instead of relying on memory. Organize a Customer Research Repository covers a structure that works for small teams.

When several high-scoring interviews point to the same costly problem, use Assess Product-Market Fit With Few Customers to judge whether the pattern is strong enough to act on.

What are the common mistakes?

Pitching too early. Showing the idea in the first minutes turns the call into feedback on your idea, and people are polite about ideas.

Asking about the future. "Would you use…" and "How much would you pay…" invite guesses. Ask what they did and paid last time instead.

Scoring enthusiasm. Tone is not evidence. A flat description of a painful week outweighs an excited opinion.

Leading questions. "Isn't it frustrating when…" supplies the answer. Ask open questions and let them name the problem.

Not asking for anything at the close. Without a small request you lose the clearest test of whether the problem matters to them.

Averaging away the signal. A few strong interviews from one segment matter more than a blended score across everyone you spoke to.

When does this approach not apply?

The script assumes the problem already happens. It suits less well when you are exploring something people have never done, such as a new kind of workflow with no current equivalent. In that case, look for adjacent behaviour they do today. It is also not a usability test: if you need to watch someone use a prototype, run a separate session. Pricing questions deserve their own conversation too; see B2B Willingness-to-Pay Interview Questions.

What else should you know about discovery interviews?

Frequently asked questions

How do you do customer discovery interviews without leading the person?

+

Anchor every question on a past event, such as the last time the problem happened. Ask what they did, who was involved and what it cost. Avoid describing your idea until the close, and avoid questions that suggest an answer, like asking whether something is frustrating.

Should I show my product in a discovery interview?

+

Usually not. Showing it early turns the conversation into polite feedback on your idea. If you want reactions, wait until the close, after you have captured their real behaviour, and treat any reaction as opinion rather than evidence in your scoring sheet.

Why score interviews at all?

+

Scoring forces you to separate what people did from what they said. It stops enthusiastic but vague conversations from carrying the same weight as detailed accounts of real cost. The total is a sorting aid that tells you which interviews deserve more weight when you look for patterns.

What if the person cannot remember a recent occurrence?

+

Treat that as information. If someone cannot recall the last time a problem happened, it is probably infrequent or minor for them. Note it, score the interview accordingly, and spend the remaining time on context and alternatives rather than pressing them to invent an example.

Related reading

• Recruit B2B Customer Interview Participants — find the right people to interview.
• Organize a Customer Research Repository — store notes and scores so patterns are visible.
• Assess Product-Market Fit With Few Customers — judge what your strongest evidence means.
• Jobs to Be Done for B2B Product Teams — frame the problems you uncover.