Short answer: Interview people who made a real decision about your product recently: recent buyers, lost deals, churned users and the people who vetted or paid for the purchase. Ask only about what happened (when they started looking, what they compared, what nearly stopped them, the words they used) and never about what they would do. A 2022 systematic review found that 9 to 17 interviews usually reach saturation in a homogeneous group, so plan about 12 per buyer segment. Code every transcript for triggers, alternatives, criteria, objections and exact phrases, and count which come up unprompted. Write the positioning statement, the message hierarchy and your homepage’s first screen from those counts and quotes, in that order.
This guide is for founders of funded B2B tech companies who want to run these interviews themselves or judge someone who will.
How is a positioning interview different from customer discovery?
A discovery interview checks whether a problem is worth solving. A positioning interview reconstructs a decision someone already made: what pushed them to look, what they compared, why they chose and which words they used. Those answers map onto April Dunford’s positioning components. She starts with competitive alternatives, “what would customers do if our solution didn’t exist,” an order she says came from Clayton Christensen’s Jobs to be Done theory, in which customers hire a product to make progress in a specific situation.
| Interview type | Question it answers | Who you talk to |
|---|---|---|
| Customer discovery | Is this problem worth paying to solve? | People who have the problem, often before there is a product |
| Positioning interview | Why do the right customers choose us, and over what? | Recent buyers, lost deals, churned users, the buying group |
| Win-loss interview | Why did this deal close or not? | Buyers from recent closed deals |
| Usability test | Can people use this screen? | Target users, with a prototype |
The “five users” rule comes from usability testing, where Jakob Nielsen showed that five users find most interface problems. Positioning needs more conversations per segment.
Who should you interview?
Bob Moesta and Chris Spiek, who developed the Jobs to be Done switch interview, recommend interviewing people who switched in the last 30 to 90 days, before the details fade. In B2B you may need to reach back six to twelve months and accept fuzzier memories.
| Group | Why they matter | What you mostly ask about |
|---|---|---|
| Recent buyers | They remember the trigger and the shortlist | The path from first thought to signature |
| Lost deals, including “no decision” | They show what your pitch failed to prove. Dunford writes that enterprise software companies typically lose 25% of deals to “no decision” | What the winner, or the status quo, gave them |
| Churned users | They show where the promise outran the product | When they first thought about leaving |
| Power users | They show where the value is real | What changed, and how they know |
| Technical evaluators | Security, IT and the engineer who ran the pilot decide what counts as safe | What they checked and what nearly failed |
| Budget holders | They sign, often for different reasons than the user | The business case they had to defend |
Without revenue, interview prospects about today’s workaround, plus domain experts who see many buyers. Leave out friends, investors and your own team. Y Combinator’s Gustaf Alströmer notes that people you know are easy to book but may be less honest, so they don’t offend you.
How many customer interviews are enough?
- Hennink and Kaiser (2022). A systematic review of 23 studies found that those using empirical data reached saturation within 9 to 17 interviews, or 4 to 8 focus groups, mostly in fairly homogeneous groups with narrow objectives.
- Hennink, Kaiser and Marconi (2017). In a study of 25 interviews, every main issue had appeared by interview 9, but the subtler ones took 16 to 24 to understand. The authors call it the difference between having “heard it all” and being able to “understand it all.”
- Griffin and Hauser (1993). In this classic marketing study, 20 interviews surfaced more than 90% of the customer needs found in 30.
All three assume one homogeneous group, so count per segment. If the VP who signs and the engineer who uses the product describe it differently, you have two segments.
| Your situation | Plan |
|---|---|
| One clear buyer segment, product in market | About 12 interviews |
| Two segments, such as the buyer and the daily user | About 12 per segment |
| Pre-revenue, or a new category with no obvious alternative | 15 to 20 per segment |
| Lost deals and churned users | 5 to 8 more, on top of the above |
Stop a segment early when three conversations in a row add nothing to your codebook. Keep going while you can name the themes but cannot yet explain them.
How do you recruit B2B interviewees?
Start with your own data: closed-won deals from the last 90 days and closed-lost deals from the last six months in the CRM, then churned accounts and your most active users in product analytics. Ask sales for the two deals they understand least.
Customers should hear from the founder. Lost prospects and churned users speak more freely to someone who is not their account owner. Whoever asks, promise it is not a sales call, and keep the promise.
To a recent customer:
Subject: 30 minutes on how you chose [product]
Hi [name], you signed with us in [month], and I’d like to understand how you decided: what started the search, what else you looked at and what nearly stopped you. It’s research, not a sales call, and nothing goes to your account team without your OK. Could we do 30 minutes next week? [link]
To a lost prospect:
Hi [name], you chose [other option, or to wait] in [month], which was the right call for you then. I’d like to learn where we fell short. Could I have 30 minutes? No pitch, and I’m happy to send a $100 gift card or donate to a charity you choose.
To a churned user:
Hi [name], your team stopped using [product] in [month]. Could I have 20 minutes to hear what happened, in your words? I’m not calling to win you back. I want to fix what didn’t work.
What should you offer as an incentive?
Pay people who owe you nothing, such as lost prospects. User Interviews, a research recruiting platform, recommends a baseline of $100 an hour for remote sessions with professionals, based on nearly 20,000 projects in its 2022 report. Many customers will talk for free, and some employers ban gifts, so offer a gift card, a donation, early access or a summary of what you learned.
How do you handle consent and recording?
- Ask when you book and again once the recording starts, so the consent is on tape. Say who will see the recording, how you will quote them (by role, unless they approve their name) and how long you keep the files.
- In the EU and the UK, recordings and transcripts are personal data. Article 13 of the GDPR requires you to tell people, when you collect the data, who you are, why and on what legal basis, who receives it (name your transcription tool), how long you keep it and what their rights are.
- In the US, federal law and many states allow recording with one party’s consent, but some states, including California, Florida and Pennsylvania, require everyone’s consent. When participants are in different states, ask everyone.
The interview guide
Ground rules
Rob Fitzpatrick’s The Mom Test has three rules: “Talk about their life instead of your idea. Ask about specifics in the past instead of generics or opinions about the future. Talk less and listen more.” Alströmer adds that you should not introduce your idea until the end of the call, or at all, because it biases the answers. Teresa Torres of Product Talk anchors her interviews on “Tell me about the last time you…”, because people overestimate what they “usually” do. Moesta and Spiek frame the call like a documentary in which the boring details matter.
25 questions, grouped by what they give you
Context
- What does your team own, and how is its success measured?
- Tell me about the last time you [did the job our product helps with].
The switch moment
- When did you first start thinking you needed something different?
- What was going on in the business at that time?
- What happened that turned “we should look at this” into “we need to fix this now”?
- What set the deadline for the decision?
Alternatives
- How were you handling this before?
- What else did you look at, including building it yourselves or doing nothing?
- How did you find the options you looked at?
- Why did you drop the other options?
What nearly stopped them, and who decided
- What almost made you not go ahead?
- What did you worry could go wrong after you signed?
- Who else had to agree, and what did each of them need to see?
- What did security, IT or procurement ask for?
- What was hard about giving up the old way of doing it?
Value and evidence
- What has changed since you started using it?
- How do you know? What do you look at or measure?
- If you had to stop using us next month, what would you go back to?
Their words
- How would you describe what we do to a colleague in another team?
- What did you call this problem internally before you bought anything?
- What did you type into a search box when you started looking?
- Who else should I talk to about this?
For lost deals and churned users
- Walk me through the last two weeks before you decided. What tipped it?
- What did the option you chose show you that we didn’t?
- When did you first think about stopping, and what was happening then?
Questions 3 to 6 follow the Jobs to be Done timeline from first thought to decision. The first thought usually comes earlier than people report. Home buyers Moesta studied said their search started in April, but 90% had their real first thought between Thanksgiving and New Year’s.
Questions 11, 12 and 15 cover the two forces that hold buyers back, the anxiety of the new and the habit of the present, and they produce your objection answers. Question 18 is the only one about the future. It names the fallback Dunford calls the competitive alternative, so check it against question 7.
Which questions should you avoid?
| Avoid | Why it fails | Ask instead |
|---|---|---|
| ”Would you pay for this?” | People are polite about the future | ”What did you pay for the last thing you tried?" |
| "What do you think of our new dashboard?” | It invites compliments, which The Mom Test counts as bad data | ”Tell me about the last time you prepared that report." |
| "What features would make it better?” | Alströmer’s point is that their job is to describe the problem and yours is to design the solution | ”What’s the hardest part of doing this today? Why?" |
| "Don’t you find spreadsheets slow?” | It hands them the answer | ”How do you do this today?" |
| "How do you usually choose vendors?” | General questions get the answer of the person they aspire to be, Torres notes | ”Tell me about the last vendor you chose." |
| "How did you find us, and what did you compare us with?” | Two questions at once, and they answer the easier one | Ask them one at a time |
Follow-up techniques
- Ask what a word means. Alströmer’s standard follow-ups are “What do you mean by that?” and “Can you tell me more about that?”
- Repeat their words back. In a User Interviews podcast, Torres shows how summarizing what she heard keeps the participant talking.
- Go back in time. “What was going on before that?” until you reach the first thought.
- Ask to see it. If they mention a spreadsheet or a report, ask them to share their screen.
- Get the order slightly wrong on purpose. Moesta and Spiek call this “connecting the wrong dots,” and people correct you with the real sequence.
- Turn compliments back into stories. “Thanks. When did you last use it for that?”
- Interview in pairs. One person asks, the other takes notes and listens for gaps in the timeline.
A 45-minute running order
| Minutes | Block | Questions |
|---|---|---|
| 0-3 | Consent, recording, documentary framing | |
| 3-8 | Context | 1-2 |
| 8-25 | Timeline from first thought to signature | 3-15 |
| 25-35 | Value and their words | 16-21 |
| 35-40 | Permission to quote, referrals | 22 |
| 40-45 | Buffer |
How do you turn transcripts into positioning?
1. Write a one-page summary per interview
Within a day, fill in the same page for every call: segment, role, trigger with a date, alternatives, deciding criteria, the moment they nearly stopped, the outcome with its evidence, and five verbatim phrases with timestamps.
2. Code every transcript
Use the same codes for every interview, each with a one-line definition so two people apply it the same way, and add codes when something does not fit.
| Code | What counts | Where it ends up |
|---|---|---|
| Trigger | The event that started the search, with a date | Why-now slide, first-screen subhead |
| Alternative | What they used or considered, including spreadsheets, an internal build or doing nothing | Your competitive frame |
| Criterion | What they compared on, and why they dropped options | Differentiators and message pillars |
| Objection | Anxiety about the new option, or habit of the old one | FAQ, sales deck, onboarding |
| Outcome | What changed, with the evidence they gave | Lead value and proof |
| Buying role | Who had to agree and what each person needed | Messages for each audience |
| Phrase | Their words, verbatim | One-liner, headline, ads |
3. Count, compare and find the best-fit segment
Put interviews in rows and codes in columns, and mark whether each mention was unprompted. Treat a theme as a pattern when a third or more of a segment raises it without prompting. One person repeating a point five times is still one mention.
Then compare segments. If budget holders talk about risk and users talk about time saved, write a message for each. Dunford defines best-fit customers as the ones who care most about your differentiated value, so look for the interviewees who describe it most strongly and with numbers. The position usually starts with them, even if that group is smaller than your current market.
Do not code alone. In Griffin and Hauser’s study, a single analyst found 54% of the customer needs on average, while seven analysts together found an estimated 99%. Have two people code the same three transcripts separately, compare, tighten the definitions and then split the rest.
4. Write the positioning decision
| Positioning component (Dunford) | Comes from |
|---|---|
| Competitive alternatives | Alternative codes, ranked by frequency |
| Differentiated capabilities | Criteria where you won, and why people dropped other options |
| Value | Outcome codes that came with evidence |
| Best-fit customers | The segment and triggers of the people who described that value most strongly |
| Market category | The words buyers used for the problem, and the options they put next to you |
Write the positioning statement only after these five decisions are made. Our San Francisco guide to brand strategy gives a fill-in template and explains why Dunford warns against starting with one.
5. Build the message hierarchy and choose proof
The one-liner comes from phrase codes, the lead value from the most frequent outcome with evidence, the pillars from criteria, and the objection answers from anxieties and habits. The SF guide shows the full hierarchy, and one story for the pitch deck, website and sales deck shows how one message map feeds all three.
For proof, use the order from our review of 30 funded B2B startup websites: a named customer with a number, then a video testimonial, a named quote, customer logos, partner or investor logos, and finally awards and press. When someone gives you a number in an interview, ask before you hang up whether you may quote it with their name or their role.
A worked example structure
This is the sheet we suggest for each segment. The brackets are placeholders, not client data.
SEGMENT: [role] at [company type, size] INTERVIEWS: [n]
recent buyers [n] | lost deals [n] | churned [n]
TRIGGER [code] unprompted in [n] of [n]
"[verbatim quote]" (interview [#], [timestamp])
ALTERNATIVE [code] [n] of [n]
WON ON [criterion] [n] of [n]
NEARLY STOPPED [anxiety or habit] [n] of [n]
OUTCOME [what changed] + [evidence] [n] of [n]
PHRASES "[...]" "[...]" "[...]"
DECISION
Best-fit: [segment + trigger]
Frame: compared with [alternative], in the category [buyer's words]
Lead value: [outcome, in their words]
Proof: [named customer + number, or the best available]
Objections: [anxiety] -> [answer] [habit] -> [answer]
OPEN QUESTION [what the next few conversations must settle]
From interviews to assets
| Asset | What it takes from the interviews | How to check it |
|---|---|---|
| Website first screen | A headline from the outcome phrase best-fit customers used, a subhead naming the trigger or the alternative, one proof line | A 5-second test with someone in the target role |
| Pitch deck problem slide | The most frequent trigger and its cost, in customers’ words, with the count (“[n] of [n] [role]s we interviewed”) | Would your customers describe it the same way if an investor called them? |
| Sales deck | One slide per buying role, and an answer to each anxiety and habit that stalled a lost deal | Every objection from lost deals has an answer |
| Onboarding copy | The outcome buyers pictured when they signed, as the first milestone, in their words | The first session delivers it |
In our 30-site review, only 13 of 30 homepages were clear within 5 seconds, usually because a slogan sat where the buyer’s outcome should have been. Our London guide covers the problem and why-now slides. For onboarding, the Jobs to be Done timeline describes a “consumption in the mind” at the moment of purchase, when the buyer has already imagined using the product, and the gap between that picture and the first real week decides satisfaction.
Common mistakes and how to fix them
| Mistake | Fix |
|---|---|
| Interviewing only happy customers | Add lost deals, no-decision deals and churned users |
| Pitching halfway through the call | Keep the product out until the last five minutes, or leave it out |
| Asking about the future | Rewrite every “would you” as “tell me about the last time you” |
| Mixing segments in one count | Plan and count per segment |
| Treating feature requests as positioning | Ask what the request is for. Alströmer recalls early Gmail users asking for the inbox and an open email on one screen, because emails loaded slowly |
| Asking an AI model to play your buyer | A model can summarize published material. Only your buyer knows what they used last spring and who nearly blocked the deal |
| Ending with a findings deck | End with written decisions on segment, alternative, lead value and proof, each with an owner |
Tools
Any of these will do for 10 to 30 interviews:
- Scheduling: Calendly, Cal.com or your calendar’s booking page.
- Calls and recording: Zoom, Google Meet or Microsoft Teams.
- Transcription and note-taking: Otter, Fathom, Grain or the transcripts built into your video tool. Tell participants when a note-taking bot joins.
- Analysis: a spreadsheet is enough at this size. Dovetail and Condens are built for tagging research, and Miro or FigJam work for affinity mapping.
- Recruiting people who are not customers: User Interviews and Respondent run participant panels, including professionals.
How dp.vision runs customer interviews
Every dp.vision project starts with interviews with our client’s real customers. The research decides the message, the message decides what we design and build, and the person who ran your interviews reviews the build. If you only want the research, Validate starts at $2,500 and covers customer interviews, competitor and market research, surveys and a written recommendation naming the main problem, the scope and the next step.
For PaceSovereign, conversations with prospective clients, their advisors and longevity physicians set the position before the brand, the landing page and the MVP existed. The interviews defined two buyer profiles, and the landing page speaks to both. Interviews and positioning took 50 hours, a market valuation of $4,250-5,000 at our rate of $85-100 an hour (full budget). If you are comparing studios, our comparison of eight branding agencies that use AI shows how each uses AI in brand work and what it charges.
If you want interviews with your customers before your next round, launch or sales push, tell us what you are working on.
Frequently asked questions
How many customer interviews do you need for B2B positioning?
Plan per segment. A 2022 systematic review by Hennink and Kaiser found that 9 to 17 interviews reached saturation in fairly homogeneous groups, and a 2017 study by the same authors needed 16 to 24 to understand the subtler issues. Plan about 12 per segment, or 15 to 20 if you are pre-revenue.
Who should a B2B startup interview for positioning?
People who made a real decision recently: recent buyers, lost deals (including no decision), churned users, power users, and the technical evaluators and budget holders behind the purchase. Leave out friends, investors and your own team.
What questions should you ask in a customer interview, and which should you avoid?
Ask about what already happened: when they started looking, what else they considered, what nearly stopped them and who had to agree. Avoid hypotheticals, pitches, feature requests, leading questions and yes-or-no questions, which produce polite guesses.
How long should a customer interview be?
Book 45 minutes and plan about 30 minutes of questions. To reconstruct a full purchase in detail, Bob Moesta and Chris Spiek recommend 45 to 90 minutes.
Should you pay B2B customers for interviews?
Pay people who owe you nothing, such as lost prospects. User Interviews recommends a baseline of $100 an hour for remote sessions with professionals. Many customers talk for free, and some employers ban gifts, so offer a donation or a summary of what you learned instead.
Do you need consent to record a customer interview?
Ask every time, before you record. In the EU and the UK, the GDPR requires you to tell people who you are, why you collect their data, how long you keep it and what their rights are. Some US states, including California, require consent from everyone on the call.
How do you turn customer interviews into a positioning statement?
Code each transcript for triggers, alternatives, criteria, objections, outcomes and exact phrases. Count which codes come up unprompted in each segment and choose the segment that values you most. Write the statement last, as a summary of those decisions.
Sources
- Hennink and Kaiser (2022), Sample sizes for saturation in qualitative research: a systematic review of empirical tests, Social Science & Medicine
- Hennink, Kaiser and Marconi (2017), Code Saturation Versus Meaning Saturation: How Many Interviews Are Enough?, Qualitative Health Research
- Griffin and Hauser (1993), The Voice of the Customer, Marketing Science, MIT Sloan reprint
- Jakob Nielsen, Why You Only Need to Test with 5 Users, Nielsen Norman Group
- April Dunford, A Quickstart Guide to Positioning
- Clayton Christensen, Taddy Hall, Karen Dillon and David Duncan, Know Your Customers’ “Jobs to Be Done”, Harvard Business Review
- Bob Moesta and Chris Spiek, How to Run a JTBD Switch Interview, The JTBD Timeline and The Four Forces of Progress
- Rob Fitzpatrick, The Mom Test
- Y Combinator, Gustaf Alströmer, How to talk to users
- Teresa Torres, Customer Interviews: How to Recruit, What to Ask, and How to Synthesize What You Learn, Product Talk
- User Interviews, How to Interview Customers Continuously with Teresa Torres
- User Interviews, The UX Research Incentives Report
- Article 13 GDPR: Information to be provided where personal data are collected from the data subject
- Digital Media Law Project, Recording Phone Calls and Conversations
- dp.vision, We Reviewed 30 Funded B2B Startup Websites
