How to Validate Your Product (and Pivot Quickly if Nobody Wants It)

Second of six on the steps of go to market.

A Bay Area investor called Kevin Fong described how he sorted the business plans that arrived on his desk, in a line Omer Khan quoted at SaaS Club on 16 May 2016 (opens in a new tab):

We divide business plans into three categories: candy, vitamins, and painkillers. We throw away the candy. We look at vitamins. We really like painkillers. We especially like addictive painkillers!

Founders picked it up as a self-assessment, which is a much harder job; ranking your own product against the customer carrying on exactly as they are.

Which pile you're in depends on who's holding the product

The SaaS Club piece carries a caveat that disappears whenever the framework gets quoted: you don't always have to offer a painkiller, addictiveness matters more than which category you're in, and the same product is candy, a vitamin or a painkiller depending on the user. The same invoice-chasing tool is a painkiller to the freelancer who lost most of March to it, a vitamin to the agency owner who already pays a bookkeeper, and candy to everyone else.

There's also a warning in there for anyone selling a subscription, which is that a painkiller solving the pain permanently may have no repeat revenue at all.

So the best quesiton has a person in it: who was in enough pain last month to go and do something about it, and what did they do? The piles move when you change who you're selling to.

Five to ten conversations

Rob Fitzpatrick's The Mom Test is the one method here that always is worth time: five to ten interviews before any significant building, every question about past behavior and past spend.

The questions you need are the boring ones. When did you last hit this problem, what did you do about it, what did that cost you, and what are you paying for now. "Would you use this?" asks somebody to forecast their own behavior and then hand the forecast to a someone hoping for a yes - it feels nice, but it’s not the accuracy needed at this stage.

Building got cheap, so the risky assumption moved

Eric Ries's Build-Measure-Learn loop assumed the build was the expensive part, and when The Lean Startup came out in 2011 it was. So now, feasibility is rarely what kills a solo product any more - it’s demand. The cheapest experiment available is a conversation, and "minimum" in minimum viable product means the minimum needed to test your riskiest assumption.

Pincus splits an idea in three so you can tell which part isn’t right

Mark Pincus, who built Zynga, set his version out in Entrepreneur on 21 July 2026 (opens in a new tab). Everything gets split into Proven, Better and New.

Zynga Poker had a proven table layout, made better by removing the download because they built it in Flash, and new in exactly one respect: real Facebook profile pictures at the table. "When we saw that 25% of people were joining their friends at tables, we knew we were onto something big."

Isolation is what makes a pivot small

StackBlitz built a first prototype of what became Bolt in February 2024 and abandoned it, because the models weren't good enough yet. They restarted in June 2024 after Claude 3.5 Sonnet shipped, went heads-down over the summer and launched in October 2024 on a tweet: zero to $40M ARR in five months with under twenty people, per Eric Simons speaking to Lenny Rachitsky on 13 March 2025 (opens in a new tab).

The case for skipping all of it

Sometimes the test costs more than the build. If you can ship the real thing in four days, then four days of interviews about a hypothetical version of it feels onerous and might not even be worth it.

Building it cheaply answers whether you can build it. Whether anybody finds it, works out what it replaces and pays for it is a separate test.

What to write down before the first conversation

Which pile the problem sits in for one named person, and what that person last spent on it. The proven thing they use today, the one improvement you’re claiming over it, and the single new element you’re betting on. A fail condition with a number and a date in it, written before the interviews start, because written afterwards it will fit the results.

The two frames answer different questions. Candy, vitamin and painkiller sorts whether the problem is worth somebody's money; Proven, Better and New sorts whether your answer to it is worth building. Both get settled by what a person already did and what they already paid.