The Last Time
The Living Brainstorm · talking to people without collecting flattery
Interviews
People lie to you. Not maliciously, they're being kind, they're guessing about their own future behaviour, and they can tell what you want to hear. So the discipline isn't detecting lies. It's asking questions where the kind answer and the true answer are the same answer.
Core discipline adapted from Rob Fitzpatrick's The Mom Test. The same anchoring move appears independently in Teresa Torres's story-based interviewing in Continuous Discovery Habits; two sources arriving separately at one move is why it sits at the centre of this page. The Itch is adapted from the internal-trigger hunt in Nir Eyal's Hooked, compulsion framing dropped. All credited as inspiration. Nothing reproduced.
01
The move
Refuse the generality. Get one specific incident. Ask what actually happened. Everything else on this page exists to support that one conversion, and if you only remember one thing, remember to ask when the last time was.
✕ Generality
"How do you usually handle invoicing?"
gets you a self-portrait
✓ Incident
"Walk me through the last invoice you sent."
gets you evidence
The move
One dated incident
People describe themselves as who they want to be. Specifics break the portrait.
"Usually" and "always" are invitations to ask when the last time was.
Yours: two independent sources land on this same move. That's why it's the name of the page.
02
The kindness test
Run every question through one filter before you ask it: would this still be useful if you asked it of someone who loves you and badly wants you to succeed? If their affection could change the answer, the question is broken.
✕ Affection changes the answer
"Do you think this is a good idea?"
"Would you use something like this?"
"Would you pay for it?"
✓ Answer holds either way
"When did you last pay to solve this?"
"What did you try before?"
"Talk me through the last time it went wrong."
The filter
Their past, not your idea
Ask about what happened. Never about what they think of you.
Every "would you" is a request for a guess about the future, given by someone motivated to be nice.
03
Praise, Vapor, Wishlist
Three things people hand you that feel like data and aren't. Learn to hear them mid-sentence and convert them on the spot. Write them down anyway; just never let them into the evidence column.
Praise
"That's a great idea." "I love this." "You're really onto something."
Convert Stop pitching. Praise usually arrives right after you've pitched, which is the actual bug.
Vapor
"I usually…" "We always…" "I'd definitely pay for that." "Generally we just…"
Convert Pin it to one dated event. "When did that last happen? Talk me through it."
Wishlist
"You should add tagging." "It'd be great if it synced with our CRM."
Convert Chase the motivation, not the feature. What would they do with it, and when did they last need it?
Bad data
Hear it, convert it
Three inputs that feel like signal and carry none.
Praise is a symptom. The pitch that caused it is the disease.
Yours: record all three, in a separate column. The ratio of evidence to vapor is its own diagnosis.
04
The Give
Enthusiasm is free, so it's worth exactly what it costs. The only signals that count are the ones that cost something. If a conversation ends without a give, record it as a zero rather than a maybe; a pipeline of maybes is how people talk themselves into building things.
Nothing"I love it." "Great idea."
Nothing"Definitely let me know when it launches."
Nearly nothing"Email me and I'll take a look."
SomethingAn hour of their time, scheduled, that they turn up to.
RealAn introduction to their boss or a peer, made by name. They've spent reputation.
RealAccess to their data, their team, or their process.
Settles itMoney. A deposit, a pre-order, a pilot fee, a signed letter of intent.
The Give
Time, reputation, money
Three currencies. Money is the only one that ends the argument.
No give means no signal. That's normal and fine; record the zero honestly.
Yours: this is the bridge to The Number. "I'd buy that" and a deposit aren't weak and strong versions of one signal. One of them isn't a signal.
05
The Itch
Under most feature requests there's an emotion, and the emotion is the durable thing; features change, the itch doesn't. Ask why until the answer stops being about the product and starts being about how they feel. Then stop.
They say"I want a weekly export."
Why?"So I can send it to my manager."
Why?"She asks about the numbers on Mondays."
Why?"Because last quarter I missed something and she found it before I did."
The itchHe doesn't want an export. He doesn't want to be caught out again.
The Itch
Build against the feeling
Three or four rounds of why. The answer gets less technical each time.
Stop when it gets uncomfortable. Past that it's interrogation, and it poisons the rest of the call.
Yours: the brake on this lives in The Doubt List under Harm. Finding what someone can't put down is not automatically permission to build on it.
06
Running one
Five phases, and the last one matters as much as the call itself. Sorting good data from bad happens after the conversation as much as during it.
BeforeWrite your three riskiest unknowns. Three, not ten. Without them you'll pitch instead.
OpeningDon't present. If they ask what you're building, deflect once and give a one-line version late. Everything after a pitch is contaminated with politeness.
ShapeAsk about their world. Drop every generality to the last instance. Ask what they've already tried and what they've already paid for.
CloseAsk who else to talk to, by name. Ask for a give. Then ask what you should have asked.
AfterWithin the hour, write what they said in their words. Mark each line: evidence, vapor, or your own inference.
The session
The notes are the deliverable
Five phases. The one after the call is the one people skip.
Quote, don't summarise. The specific words are the data; your paraphrase is already an interpretation.
What they've already spent money on is the best available predictor of what they'll spend money on.
07
Ways this goes wrong
Most of these feel productive while they're happening. That's the problem with them.
Pitching first
Guarantees praise and destroys everything after it. The single most common failure.
Asking about the future
People are terrible at predicting their own behaviour, and confident while doing it.
Leading
"Wouldn't it be useful if…" is not a question; it's a request for agreement.
Talking to people who like you
Friends are the easiest sample to get and the worst one to have.
Counting instead of learning
Twenty interviews that all confirmed you is one interview run twenty times.
Summarising
Your paraphrase drops the specifics, and the specifics were the evidence.
Anti-patterns
Productive-feeling failure
All six leave you with more notes and no more knowledge.
The test: whether you heard something this week that you didn't want to hear.