I Like Business Ideas. I Trust Experiments More.
Ideas are the cheapest thing in any room I walk into.
I had coffee in Lisbon a few weeks ago with two people I like arguing with, and between us we produced three business ideas before the second cup arrived. All three sounded good. I would have defended two of them in front of a serious audience. None of them cost us anything to have.
That last part is what people skip. Generating a plausible idea is free, fast, and pleasant. It feels like progress because it is progress — of a kind that produces nothing on its own.
Let’s imagine a specific Tuesday.
- A small software team sells scheduling tools to independent fitness studios. On three separate calls, customers mention that failed card payments are a headache.
- The team hears the pattern. Someone says the sentence that ends every discovery meeting: “There’s something here.”
- It goes on the roadmap. It gets a name. It gets a slide.
- Six weeks of engineering later, the feature ships behind a toggle.
- Almost nobody turns it on.
Nothing in that story is stupid. Every step follows sensibly from the one before it. The mistake happened at step two, and it was invisible: a belief got promoted to a plan without ever passing through a test.
The test, in that case, would have taken a day. Email the twenty studios most likely to care, describe the thing in three sentences, ask who wants it switched on for them next week. Count how many say yes in writing. That single day, run before the six weeks, changes everything downstream — either it kills a costly feature or it buys the team six weeks of building with actual wind behind them.
A business idea is a recipe. You can read a recipe as many times as you like and it will never tell you that the sauce is over-salted. Only the pan does that. The whole skill sits in getting to the pan sooner and paying less for the burn.
So I’ve stopped treating “I believe in this idea” and “this idea has survived a real test” as neighbors on the same scale. They’re different categories. The first is a feeling I generate for free. The second is information the world gave me, usually reluctantly, and usually not the answer I wanted.
Here’s the uncomfortable part. Most operators — me included, on bad weeks — skip the second step precisely because the first one feels like enough. Conviction is warm. Testing is cold, slightly embarrassing, and carries the risk that you’ll have to abandon something you’ve already described out loud to people whose opinion you care about.
That’s the real cost of an untested idea. Not the money. The identity you build around it while it’s still a hypothesis wearing a suit.
What makes a test worth running
The one thing I check first: does a real “no” exist? If there’s no outcome that would make me drop the idea, I’m not running an experiment, I’m running a ceremony. Write the kill line down before you start — the number, the date, the specific thing that has to happen. Written afterwards, it always moves to wherever the result landed.
The second thing: does the answer cost the other side anything? “Sounds interesting” is free to say, so it carries no information whatsoever. A calendar slot costs time. A signature costs review. A deposit costs money. Access to their data costs trust. Anything a person gives up to say yes is worth more than a hundred enthusiastic replies, and it’s usually available much earlier than people assume.
The third: am I testing the assumption that would hurt most if it’s wrong, or the one that’s easiest to check? Those are almost never the same. It’s tempting to validate the part you already have a method for. The riskiest assumption is usually the one you’ve been quietly avoiding — that anyone will pay, that the integration is possible, that the person who loves this has a budget line for it.
And a note on the pace: this isn’t a case for caution. It’s the opposite. Cheap honest experiments are what let you take more swings, because each one costs a day instead of a quarter. The people I’ve seen move fastest weren’t the boldest, they were the ones who’d made being wrong survivable.
Where I learned this in my own work
I run my sales pipeline with an agent system I built and operate myself. Not a product, not something a vendor sold me — a set of tools I’ve been shaping for a while now, and it does the parts of the job that used to eat my mornings so I can spend the day on the parts that need a person.
Every single piece of it started as a guess. Some of those guesses were confident. Most were wrong.
What made it work wasn’t better guessing. It was the discipline of changing one thing at a time, running it on a slice of real work rather than the whole pipeline, and deciding in advance what result would make me revert it. The system that exists today is simply what’s left after everything that failed its own test got removed. That’s the honest description of it.
The habit transfers cleanly to anything else I touch. Before agreeing that a change to a process is a good idea, I ask what would have to be true for it to be a bad one, and how quickly we could find out. Half the time the answer is “by Thursday, for the cost of one afternoon.” That answer is almost always available, and almost never reached for.
The tests that aren’t tests
A few things wear the costume convincingly. Asking friends. Surveys where nobody spends anything. An internal deck that argues the case. A pilot with a customer so eager to help that they’d have said yes to the opposite idea. A metric picked after the fact because it happened to move.
The most common one, and the hardest to catch in yourself: designing an experiment you can’t fail. If every possible outcome leaves the idea alive, you’ve built a machine for producing reassurance.
I’ve watched a lot of founders present in Portugal over the past couple of years — the pitch nights here have a good, blunt energy to them. The pitches that stay in my head afterwards aren’t the ones with the biggest market slide. They’re the ones where someone says: we ran this small thing, here’s the ugly number it gave us, here’s what we changed because of it. Ten seconds of that outweighs ten minutes of vision, because it’s the only part of the presentation that came from outside the founder’s own head.
The good news is that the scarce skill here is learnable, and it’s not intellectual. Designing a fast, cheap, honest experiment is mostly a matter of being willing to look slightly silly for a day in order to avoid looking very silly for a quarter. That’s the whole trade.
I still like ideas. I collect them, I argue about them over coffee, and I’ll happily spend an evening building a case for one I have no intention of pursuing. The enthusiasm isn’t the problem — it’s the fuel. It just can’t be the evidence too.
If you’re sitting on something you believe in, I’d rather hear about the test than the thesis. Tell me the smallest version you could run this week, and I’ll try to help you make it smaller.