Taste Maker #2. Marcelo Calbucci: The Document That Talked Him Out of His Own Startup

Bias to Action Is a Trap. He Measures Impact Instead.

Listen to this episode on Spotify or Apple Podcasts

In January 2025, Marcelo Calbucci sat down to write a PRFAQ for a new investment vehicle he was thinking of starting. By the end of the document, he had decided not to start it. The author of The PRFAQ Framework — six-time founder, ex-Amazon, ex-Microsoft — walks me through what writing forces you to see that a pitch deck never does.

“Let me say something that might surprise you,” Marcelo Calbucci tells me.

We are an hour into a conversation about a book he spent a year writing, and he says it the way a man lays a card face-up on the table — without flourish, watching to see what I do with it. “Most PRFAQs lead to a no-go decision.”

I should explain what a PRFAQ is, because the name does it no favors. It is the document at the center of Amazon’s Working Backwards process: a press release for a product that does not exist yet, followed by a long list of frequently asked questions about it. You write the thing as if it has already shipped, and then you interrogate it. Marcelo spent two years inside Amazon watching this process run at full weight — months of writing and review for a single bet, financial modeling, user research, the whole apparatus — and then he left and wrote a book adapting it for the people who never get that apparatus: founders, designers, product managers at companies with no research team to lean on.

So when the author of the book tells me that most of the documents the book teaches you to write should end in no, I do the thing I suspect he wants me to do. I push back. That sounds like failure. You do weeks of work, and the deliverable is that you don’t build the thing?

He has heard this before. I can tell because he doesn’t reach for theory. He reaches for himself.

“This is very personal,” he says. Last January, he sat down to write a PRFAQ for something he genuinely wanted: a new investment vehicle for founders. Not a thought experiment. A real next move for a man who has started six companies and raised more than forty million dollars doing it. He knew the shape of the thing. He could have built a pitch deck for it in an afternoon and felt good walking out of the room.

Instead he opened a blank document and started writing the press release for it — the announcement he’d make if the thing already existed and had already worked. Then the questions underneath it, the ones his own framework forces: who is the customer, whether the problem he was so sure existed actually existed or just felt like it did, how the people he wanted to serve were already solving it without him. Each answer he was confident about sent him to go check, and the checking sent him back to the document less confident than he’d started, and somewhere in that loop the thing began doing what he wrote an entire book about. It started talking back.

This is the part outsiders never see about a man with his record — six companies founded, more than forty million dollars raised across them, eighteen years spent deciding what to build and convincing other people to fund it. You would assume that by now the deciding gets easier, that a founder six times over walks into his seventh idea with the confidence of someone who has been right before. What he was describing instead was a person voluntarily sitting down with the one tool guaranteed to make a good idea harder to keep believing in. He did not have to write the document. Nobody was making him. He wrote it because he has learned not to trust an idea he has not yet tried to talk himself out of.

I want to know what it said. I want to know whether a man who has done this six times can really sit in a room with his own idea and let a Word file argue him out of it, or whether that is the kind of thing you say in an interview and not the kind of thing that happens at a kitchen table in January.

He pauses on the word realized, like he is deciding how honest to be about what came next.

What he decided that month — and the reframe he handed me when I called it a waste — is the part of the conversation I have been turning over since we hung up.

What this profile covers

1. What that document did to a founder who has been right six times before — and the reframe he handed me when I called the whole exercise a waste. The shift he names is small, a single change in what you choose to measure, and he uses it to separate the product managers who last from the ones who burn out. Once he said it, I could not unsee how much of product work is built to reward the wrong number.

2. Why the process can’t simply be lifted out of Amazon and dropped into your company — the part of his book that got ahead of every objection I came in with. It starts with a sixty-minute meeting that opens with twenty minutes of total silence, and it ends somewhere uncomfortable for anyone who has ever said “let’s just do it the way Apple does it.”

3. And the line he draws through the middle of 2026’s most tempting shortcut: what you can hand to an AI to write, and what you must refuse to. The example that makes it concrete is the most boring feature request in software — a customer asking for tags — and what Marcelo does with it is the whole argument for why the thinking is still yours.