A hundred dollars is enough that I can’t hand-wave the result. Thirty days is enough that the market has seen me more than once. Five angles is enough that I’m not judging the idea by the single execution I happened to try first.

The question the test answers isn’t is this good? It’s does this want to exist? Those are different questions, and most founders conflate them.

What the five angles look like in practice §

If the idea is a piece of software, the five angles aren’t five landing pages. They’re five framings of what the product is — five different stories about why someone should care. Same core; different wrapper.

Sometimes the wrapper matters more than the core. Sometimes the core is just wrong and no wrapper rescues it. Five tries is the smallest number that tells me which one I’m looking at.

The part that surprises people §

About thirty percent of what I test gets discarded after the $100 is spent. Not “pivoted.” Discarded. The idea goes into the vault and I move on.

I used to feel bad about that number. Now I think it’s the ratio that keeps me honest. If I were discarding five percent, it would mean my filter upstream was doing too much work — I was pre-rejecting things I hadn’t actually tested. And if I were discarding seventy percent, the filter would be doing too little work, and I’d be wasting money on things my gut already knew were dead.

Thirty percent is the band where I’m surprised often enough to trust the process, and wrong often enough to know the process is real.

Why the test is cheap on purpose §

You could run a better validation test for $5,000. You’d learn more. But the point of $100 isn’t to get the best possible signal — it’s to get enough signal to stop arguing with yourself.

Most arguments about whether an idea is good end not because someone was persuaded, but because someone got tired. The $100 test ends the argument for you. The market either leaned in or it didn’t, and you stop relitigating.

That is the actual product of this test. Not the ideas. The decisions.

A failure rate measures how often the market tells you no. A discard rate measures how often you listen. They sound similar. They are extremely different.

I discard about thirty percent of what I test. Some of those discards were genuinely bad ideas — poorly scoped, thinly reasoned, a pitch I’d already half-abandoned before the test ran. Those are easy. You run the numbers, the numbers are flat, you shrug and move on.

The hard ones are the ideas I still liked after the data came in.

The seduction of almost-working §

There is a particular failure mode where an idea gets a little bit of traction and you decide to keep going. Not because the traction was decisive — it wasn’t — but because you’d rather do something than admit the test didn’t land.

I’ve done this. Everyone I know has done this. The pattern is easy to spot from the outside and nearly invisible from the inside.

The test isn’t “did anyone want this?” The test is “did enough people want this, through enough angles, that the market was choosing it instead of tolerating it?”

Most almost-working ideas are being tolerated. That’s not a product. That’s a courtesy.

Why the number stays around thirty percent §

I’ve tried to drive the rate down. It doesn’t work. The more patient you are upstream — the more ruthless your pre-filter — the less information you get from the test itself. You start believing you already know the answer before the data arrives, which is exactly when you stop being calibrated.

Thirty percent is the tax I pay for keeping my filter honest. If I lowered it to ten, I’d be congratulating myself on good judgment while quietly filtering out the data that would have updated me. I’d rather waste a few hundred dollars a year learning things I didn’t expect.

The thing that actually makes this work §

You need somewhere for the discarded ideas to go. Not a graveyard — a vault. The idea doesn’t get deleted, it gets archived with its reasoning: what I tested, what I learned, what would have to change for me to pick it up again.

Most of those never come back. A few do, once the surrounding context has moved — a technology matures, a market opens, a partner appears. When that happens I don’t start over. I pull the file.

The discard rate is not the end of those ideas. It’s the beginning of their patience.

The system is permanent. Whatever specific idea I am currently testing is not. It is a contestant. It has to earn its place every month, and it has to accept being retired when the data says so. My job is not to win with any one idea. My job is to keep the system honest enough that the winners, when they appear, are actually winners.

Guilty until purchased §

Here is the rule I try to live by: assume an idea is guilty until it has been purchased.

Not complimented. Not shared. Not bookmarked. Purchased.

A compliment tells you somebody is polite. A purchase tells you somebody was willing to trade a real thing — time, money, attention — for the thing you made. Everything short of that is noise, or it’s too early.

This is unpopular advice because it’s slow. You don’t get to feel good about an idea until the market has voted with a wallet. You have to sit in the uncertainty. You have to test the idea enough times that if the data is flat, you can’t tell yourself “they just didn’t see it.”

What this costs §

It costs ego. The system doesn’t care that I’ve been working on an idea for three months. It doesn’t care that the idea is clever. It doesn’t care that my friends said it was promising. The system only cares about the data, and the data is sometimes unkind.

It also costs momentum. Most founders want to feel like they are winning. The system doesn’t let you feel that often. You feel it when you get a real purchase, and you don’t feel it otherwise, and you have to be okay with that.

What it’s worth §

What it’s worth is every idea I didn’t keep pushing past its expiration date. Every month I saved by retiring something instead of dragging it. Every attention dollar I spent on a live contestant instead of a zombie.

The system is not a shortcut. It is the opposite of a shortcut. It is the discipline of running each idea through the same loop — $100, 30 days, five angles — and letting the data do what data does, which is disappoint you and occasionally surprise you.

I would rather be disappointed on a schedule than surprised too late.