Running a Fake-Door Test Before You Build

What a fake-door test is

A fake-door test, sometimes called a painted-door test, puts the entrance to a product or feature in front of real people before the thing exists. It might be a landing page, a button in the app or a new pricing option. You measure who walks through it, tell them honestly that it isn't ready yet, and ask whether they'd like to hear when it is.

It's one of the cheapest ways to find out whether anyone wants something. It's also easy to do badly.

Where most tests go wrong

The usual failure is deciding what the numbers mean after you've seen them. If the click-through comes in low, someone says the copy was off. If it comes in high, everyone agrees to build. Either way, the test hasn't decided anything.

So I pre-register the test. Before it runs, we write down what we're testing, who sees it, how many people need to see it, and the thresholds that mean go, rework or stop. The team agrees to those numbers in advance, so when the result arrives the decision has already been made.

I've run a pre-registered fake-door test across two AI products at once. Agreeing the thresholds up front turned "should we build this?" into a question the data could actually answer.

What you get

A test designed around the decision you need to make, the page or in-product entry point built and live, thresholds agreed in writing, and a read-out that says what to do next. Each test usually runs one to three weeks, and designing it well takes longer than running it. If you're about to commit a quarter to something nobody has asked for yet, let's talk.