Why do our messaging interviews give us answers we already had?
By Greg Rosner
Founder of PitchKitchen · Author of StoryCraft for Disruptors
· 7 min read
TL;DR
A truth extraction interview is a one-on-one conversation inside a messaging engagement whose only job is to get a specific, opinionated sentence out of a founder's head and onto the record. Most discovery interviews fail because they ask for summaries, and a founder answers a summary question with the version they've already rehearsed for investors. Extraction asks for episodes instead: one deal, one customer, one thing somebody actually said. The truth lives in the particulars. Expect ninety minutes, one interviewer, recording on, and roughly a dozen questions you'll find harder than you expect. What comes out is raw material for a Magnetic Messaging Framework, not a finished message.
Your discovery interviews give you back the answers you already had because the questions asked you to summarize. A founder who gets asked what makes you different produces the sentence they've said two hundred times to investors, recruits and their own board, and that sentence is polished, generic, and useless as raw material. Truth extraction fixes this by refusing to ask for positions at all. It asks for episodes: one deal, one customer, one thing somebody actually said out loud.
This is the part of a messaging engagement nobody photographs. It's not a workshop with sticky notes and it isn't a strategy presentation. It's a recorded conversation where one person asks a founder uncomfortable questions for ninety minutes and writes down the moments the polish slips.
What actually happens in a truth extraction interview?
One interviewer, one subject, ninety minutes, recording on. No slides, no framework taught at the top, no group. The interviewer talks maybe fifteen percent of the time and spends most of it asking some version of what do you mean by that.
The one-on-one part carries more weight than it looks. Put the leadership team in a room together and you'll get the account everyone can live with, which is the account already on the homepage. That group conversation is worth having, and it's a genuinely different exercise with a different job: a brand positioning workshop is where a leadership team agrees. Extraction happens first, separately, because you can't reconcile accounts you haven't collected. It's also distinct from the free ninety-minute strategy session that happens before anyone signs anything, where the job is diagnosis rather than raw material.
A full engagement runs a handful of these: the founder, one or two revenue leaders, two or three customers. The customer interviews are the ones founders underestimate. Customers routinely describe a company better than its own website does, and they do it in words the company would never have chosen.
Why do we give the rehearsed answer without noticing?
Every founder carries a small library of answers that have survived contact with an audience. The fundraise version. The recruiting version. The version for the board meeting where somebody questioned the roadmap. These answers exist because they worked, and they're the first thing that arrives when a question sounds familiar.
Call it the Rehearsed Answer. It isn't dishonesty and it isn't laziness. It's the sentence that got you through the last hundred versions of this conversation, retrieved automatically, and its defining feature is that it's already been sanded down for a room that wasn't buying your product. Investors want a market. Recruits want a mission. Neither of them is your buyer, and the answer that satisfied them has had every specific, arguable, slightly dangerous edge filed off it.
This is why summary questions fail. What's your differentiation, who's your ICP, what's your value proposition: each one is an invitation to retrieve. The interviewer gets fluent, confident, category-standard language back and mistakes the fluency for clarity. Then that language ends up on a homepage, which is a large part of why so many B2B websites sound like each other.
What do the questions look like when they're working?
The move is to stop asking about the company and start asking about events. A position is a claim you can rehearse. An episode is a thing that happened, and you can't retrieve a pre-polished version of it because nobody ever asked you for it before.
| The summary question that fails | The episode question that works |
|---|---|
| What makes you different from your competitors? | Tell me about the last deal you lost. Who won it, and what did the buyer say when they told you? |
| Who's your ideal customer? | Which customer, by name, would be furious if you shut down tomorrow? Why that one and not the bigger account? |
| What problem do you solve? | What was your customer doing on a Tuesday afternoon before they had you? Walk me through the whole hour. |
| What's your point of view on the market? | What does everyone in your category believe that you think is wrong? Who would argue with you about it? |
Notice what the right-hand column has in common. Every one of them is unanswerable in the abstract. You have to go get a specific memory, and specific memories carry the detail that makes a message land: a real objection in the buyer's own phrasing, a named alternative, a number somebody actually tracked.
The last row is the one that does the most work and gets skipped the most. A company that can't name what it's arguing against hasn't got a position, it's got a description. That's the difference between problem-centric messaging and describing pain everyone already knows, and it's the same gap that makes naming the problem you solve so much harder than founders expect.
What does the data say about companies that never did this work?
We can put a number on it. In August 2026 we scored 302 healthtech homepages against the 19-signal Brand Signal Score rubric for the 2026 Healthtech Messaging Index. Two of those nineteen signals are, in effect, the output of extraction: whether the page names a status quo it's rebelling against, and whether it owns language a competitor couldn't copy.
The rebellion signal is one of the weakest in the whole rubric. 122 of 302 companies score a flat zero on it, which means the homepage names nothing it's against. On owned language, 45 score zero and 63 earn full marks.
- 1The 25 companies scoring zero on both average 13.24 out of 38, against a field average of 19.18. Not one of them reaches the top two bands. Sixteen land in Weak, eight in Invisible, one in Critical.
- 2The 15 scoring full marks on both average 27.60 out of 38, and every one of them sits in Approaching or Magnetic.
- 3142 of 302 score zero on at least one of the two, and they average 15.65 against 19.72 for everybody else.
- 4Here's the part that matters for the interview: 9 of those 25 double-zero companies earn full marks on Solution Clarity. They explain their product perfectly well. They still average 17.0, while the 176 companies with full Solution Clarity average 22.09.
That last group is the whole argument in one number. Those nine companies can describe. Describing is what a founder does when you ask them to summarize, and they're good at it, because they've had practice. What they can't do is say what they're against or say it in words only they own. That's the material an interview has to go and get, and it doesn't come out of a question that could be answered from memory.
One caution on how to read those numbers: this is an association across a single vertical at a single moment, not proof that running better interviews raises a homepage score. It's a fair description of what's missing across a large sample, which is what we use it for.
How do we prepare for one of these interviews?
Don't prepare answers. Preparing answers is the single most common way founders make the session less useful, because a prepared answer is a Rehearsed Answer with fresh polish on it.
- The last three deals you lost, and what the buyer said when they told you.
- A deal you won that you didn't expect to win.
- The exact words a customer used when they described you to somebody else. Forwarded emails and call recordings beat your memory of them.
- The objection your reps hear most, in the phrasing the buyer uses rather than the phrasing your team uses back.
- Anything you believe about your market that you'd hesitate to put on the website.
That last one is where the good material usually is. The sentence a founder is slightly nervous to publish is generally the sentence a buyer would remember, and most buyers remember nothing at all about what makes a vendor different.
What comes out the other side?
Transcripts and a synthesis. Not a message, not a tagline, and definitely not a deck. The interviews are the collection phase, and the reconciliation happens afterward, when the founder's account, the sales leader's account and the customers' accounts get laid against each other and the disagreements get resolved on purpose rather than by whoever writes the copy.
The output of that reconciliation is a Magnetic Messaging Framework: the documented version of the company's narrative identity, written so a rep, a homepage and an AI tool can all work from the same source. That framework is what makes the extraction durable, because a transcript in a folder helps nobody and a founder's memory doesn't scale past the founder.
“The raw material for a clear message is already inside the company. Nobody wrote it down because nobody ever asked a question that couldn't be answered from memory.”
If you've sat through discovery before and got back a document that said what you already said, the interview wasn't built to extract anything. It collected summaries, and summaries are where the truth goes to die. Worth knowing what a real engagement produces and how you'd know it worked: here's what a messaging engagement actually delivers, and what to expect in the first 30 days.
Questions People Ask
FAQ
What is truth extraction?
Truth extraction is the interview phase of a messaging engagement where a strategist pulls the specific, opinionated story of a company out of the people who lived it, before anyone writes a word of copy. It exists because founders know exactly what they built and why it matters, and still can't say it without sounding like every other company in their category. The strategist isn't there to teach a framework. They're there to get on the record what's already true and never been written down.
How long does a truth extraction interview take, and who's in the room?
Ninety minutes is the working length, one interviewer and one subject, recorded with permission. One-on-one matters more than it sounds: put four executives in the room and you get the version everyone can live with, which is the version already on the website. A full engagement runs several of these separately, typically the founder, one or two revenue leaders, and two or three customers, then reconciles where their accounts disagree. The disagreements are usually the most valuable material in the transcript.
How do we prepare for a messaging interview?
Don't prepare answers. Bring artifacts instead: the last three deals you lost and why, a deal you won that you didn't expect to, the exact words a customer used when they described you to someone else, and the objection your reps hear most. Preparing polished answers is actively counterproductive, because a rehearsed answer is the thing the interview is trying to get underneath. If you arrive with notes, bring evidence rather than positioning.
Why can't we just run these interviews ourselves?
You can, and some teams get real material out of it. Two things make it hard from the inside. An internal interviewer shares the company's vocabulary, so the jargon that would stop an outsider passes without a follow-up question. And a founder answers a colleague in the register the company already uses, because that register is what the room expects. If you run it internally, have the least senior person on the team ask the questions and instruct them to keep asking what do you mean by that until the answer stops containing category language.
What do we get at the end of the interviews?
Transcripts and a synthesis, not a finished message. The interviews produce raw material: the villain you're actually arguing against, the language your customers use that you don't, the sentence you said once and never wrote down. That gets reconciled into a Magnetic Messaging Framework, which is the document your website, your sales deck, your reps and your AI tools all work from. Expect the framework to contain things you said in passing and forgot you'd said.