how many customer interviews before building a product? more than zero

this morning a founder walked me through a remittance and payouts product. the slides were good. two break-even models and a sensitivity analysis.

model one... a few hundred remitters each sending about a thousand dollars a month, at a fee just over half a percent, covers the floor. model two... a small group of startups paying a modest monthly subscription. and at scale the rail rate drops, so the spread gets better. he'd thought about it.

then i asked who he'd talked to.

"all the customer research is is basically reams and reams of scraped information right and we haven't started chatting with a person yet."

not one remitter. not one startup customer.

i told him the truth. it resonates on paper. the math holds. and none of that tells you if it's good. the only thing that tells you is putting it in front of humans.

so why haven't you talked to two hundred people and tried to sell the mvp already?

people search how many customer interviews before building a product, and they want a number. there are guides that will give you one. but this founder's number was zero, and the problem wasn't that zero is below some threshold. the problem is that every hour he spent on models was an hour he could have spent on a call learning that one of his assumptions was wrong.

i know this one from the inside.

back at unocoin, around 2012 and 2013, we had a bitcoin remittance thesis. it was beautiful. money moving across borders cheaper and faster than the old rails. on paper it was obvious.

in the real world it proved very hard to crack. regulation was a major blocker, and no slide had that on it.

that's the lesson i carry. a thesis can be compelling and still not work, and the spreadsheet is the last place you find out why.

and that was before ai.

now here's what's changed. ai makes research free. it can scrape every competitor, model every scenario, and write the sensitivity table in an afternoon. ai writes the code. ai builds the landing page.

so research isn't a moat anymore. everyone has the same reams of scraped information.

the one input that's still scarce is human time. a real person telling you in their own words why they would or wouldn't switch. that's the only data your competitor can't download.

which flips the old advice. it used to be that talking to customers was the cheap step before the expensive build. now the build is cheap and the conversations are the expensive part. so they're the part that matters.

my advice to him, and to you if this sounds familiar. stop polishing the model. pick the twenty people most likely to use this. call them. don't pitch. ask how they send money today, what it costs them, what annoys them, what they've tried. then show them the thing and ask them to pay.

if they won't, the model was fiction. better to find out in week one than after the raise.

and notice what happens to your models after twenty of those calls. they get better. the fee you assumed, the volume you guessed, the customer you thought was obvious... real people will move every one of those numbers. the spreadsheet isn't useless. it's just the last step, not the first. you write it after the conversations, so it describes a market instead of a hope.

and if you're shy about reaching out to strangers, flip it around. don't ask for their time. give them something first. a question about their world, a spotlight on their story. that's the whole idea behind the spotlight: stop asking, start giving.

so how many real people have told you, out loud, that they'd pay for what you're building?