19 May 2026 · 4 min read

Build the prototype, not the business case

What killed most good ideas was never that they were bad. It was that finding out cost too much. That has changed, and it means far more ideas now get the chance to become something.

There is a familiar sequence for deciding whether to build something. Someone has an idea. They write it up. There is a document with a market size, a rough cost, a set of assumptions and a projection. People read it, add comments, and eventually the thing gets approved or it does not.

The sequence has one structural flaw, which is that the document is written by a person who already knows what answer they want. Every number in it is a number they chose. The market size is the largest defensible one. The cost is the most optimistic build anyone could argue for. The assumptions are the assumptions that happen to make the idea work.

This is not dishonesty. It is what happens when you ask somebody to argue for something. But it means the document is not evidence about the idea. It is evidence about how much its author wants to build it.

What changed

The reason this sequence existed at all is that building was expensive. If getting something in front of a real user cost six months and a team, you had better be fairly confident first, and a long argument on paper was cheaper than being wrong in code.

That ratio has moved a long way. A working prototype of a reasonably contained idea is now days rather than quarters. Not a polished product, and not something you would put a support contract behind, but something a real person can open and use and have an opinion about.

Once building is cheaper than arguing, the argument stops being worth having.

What a prototype tells you that a document cannot

A business case can tell you whether an idea sounds good. A prototype tells you whether it feels good, which is different and more predictive.

The interesting failures are almost never the ones the document anticipated. They are things like: the flow makes sense until step four, and then everyone stops. Or: people understand it immediately but do not care. Or, most usefully, people use it for something other than the thing it was built for, which is where the actual product usually turns out to be hiding.

None of those show up in a projection. All of them show up within about twenty minutes of watching someone use the thing.

The filter used to be money

This is the part that matters.

When building was expensive, something had to decide which ideas got built, and that something was cost. Not merit. Not whether anybody wanted the thing. Not whether it would work.

Which means the real filter was: who can afford to find out? Ideas belonging to people with a budget got built. Ideas belonging to people without one did not, however good they were. A great many of them were better than the things that did get funded, and nobody ever found out, because the cost of finding out was the barrier rather than the idea itself.

That is an awful filter. It selects for access, not for quality. And for about thirty years it is the filter almost every software idea had to pass through.

More ideas get to live now

The interesting consequence of cheap building is not that you can kill things faster. It is that far more ideas get to exist at all.

An idea that could never have justified six figures and two quarters can now be built and put in front of real people. Some of those turn out to be businesses. Some turn out to be features of something else. Some turn out to matter to a small number of people very much, which used to be a reason not to build a thing and increasingly is not.

The point of getting the cost down is to widen what is worth attempting. Every idea that gets a genuine attempt is one that was previously decided on by a spreadsheet, and spreadsheets have a poor record on this.

Giving something a real go

An idea only gets a fair test if it gets a fair build. A grudging version put together to prove a point does not tell you anything, and neither does a demo abandoned the first week it is quiet.

The rules we try to hold to:

• Build the riskiest part first. Not the easiest, and not the bit that demos well. The part that everything else depends on.
• Put it in front of someone who will be rude about it. Enthusiasm from people who like you is not data.
• Give it long enough to be understood. Most things are quiet before they are anything. Quiet in week one is not a verdict.
• Write down what would change your mind. Then check it against what actually happened, rather than against your memory of what you expected.

None of this is novel. What has genuinely changed is the price, and the price has changed enough that the old question, "can we justify finding out", has mostly stopped being a question worth asking.

The best reason to make building cheap is not that you can afford to be wrong. It is that far more things get the chance to be right.