Fake door test: definition, ethics, and the honest version

The term and the ones next to it

These four get used interchangeably and should not be. The differences are about what the participant is told.

Fake door test: An experiment that presents an entry point to a product or feature that does not exist, and measures how many people try to go through it. The classic form is a button inside an existing product; the landing-page form is the version used before any product exists.
Reveal: What the person sees immediately after they act. This is the only part of a fake door test with an ethical dimension, and the entire difference between a survey and a small con lives here.
Demand test: The broader family: any experiment that measures willingness to act rather than stated preference. A fake door is one instrument; a pre-order, a priced waitlist, and a concierge run by hand are others.
Priced waitlist: A demand test where the price is stated before the signup rather than after. It converts worse than an unpriced one by design, because the people it loses are the ones whose answer would have been noise.
Vanity signup: A signup made with no cost, no commitment and no stated price. Correlates with launch revenue weakly enough that a large list of them can be actively misleading, because volume reads as progress.

Where the line actually is

The objection to fake doors is usually phrased as though the deception is in the button. It is not. Showing someone an option that does not exist yet is, on its own, no more dishonest than a menu with a dish that has sold out. What determines whether an experiment was research or a small confidence trick is what happens in the two seconds after the click.

There are three common reveals. The first says nothing — the button spins, then errors, then the visitor leaves confused. That one is straightforwardly bad and also useless, because it contaminates the measurement with people who thought they had hit a bug. The second says "coming soon, leave your email" without ever admitting the feature was never there. That is the version the criticism is aimed at, and it deserves it. The third says plainly that the thing does not exist, explains what the click was measuring, and gives the person something for their trouble.

Only the third produces data you can defend in a room. It also converts worse, which is why it is rare, and which is the entire reason it is worth doing: the people who continue after being told the truth are a different population from the people who continue before being told.

Four rules that keep a fake door honest

Applied to every page on this domain. If one of them is broken here, the criticism lands and the fix is ours to make.

State that it does not exist, above the fold, before the signupNot in the footer and not after the click. Every landing page here carries the word unbuilt in the first screen, and the price is labelled as expected rather than current.
Never take money for a thing that is not thereA pre-order is a different instrument with a different contract. If you take the money, you owe the product or the refund, and most people running a fake door are not prepared for either.
Give something back on the reveal pageA template, a calculation, a checklist. The person spent attention on you; the reveal is where you settle it. This is a real deliverable and not a consolation prize.
Tell them the outcome, including when the answer is noThe failure email is the one that costs something to send and the one that makes the whole exercise legitimate. Committing to it in writing, before launch, is the cheap version of an ethics review.
Do not run it on people who cannot walk awayA fake door inside a product a customer already pays for spends trust they gave you for something else. On a public page a stranger arrived at from search, they can simply close the tab, and mostly do.

The fourth rule is the one that separates an experiment from a harvest, and the only one with a cost attached.

Three reveals, three different experiments

Same button, same traffic. The reveal changes what you measured and whether you can use it.

Dimension	Silent error	"Coming soon"	Stated as unbuilt
What the visitor believes afterwards	The site is broken	The feature is nearly ready	The feature does not exist
Click-through, relative	Highest	High	Lowest — deliberately
What the number means	Curiosity plus confusion	Curiosity	Curiosity that survived a price
Defensible in public	No	Arguable, and argued about	Yes
Cost if the idea dies	Wasted traffic	An email list you owe an explanation	One email, already promised

The relative click-through ordering is the consistent finding across published write-ups of these tests; the magnitudes vary far too much between contexts to quote a number.

The questions people actually search for

Is a fake door test unethical?

Not inherently. It becomes unethical when the reveal conceals rather than explains, when money changes hands, or when it is run on existing customers who cannot decline without cost. A public page that says the product is unbuilt, takes no payment, and commits to reporting the outcome is a survey with a better response mechanism.

How much traffic does a fake door test need?

Enough to distinguish "nobody" from "someone", which is a much lower bar than statistical significance and the only bar most of these tests can actually clear. A few hundred visitors and a handful of high-intent signups is a signal. Two thousand visitors and one signup is also a signal, and a clearer one.

What is the difference between a fake door and a pretotype?

Mostly vocabulary and lineage. Pretotyping, as Alberto Savoia framed it, is the wider practice of faking the product to test the demand; the fake door is the specific instrument where the fake is an entry point inside a real surface.

Can I run one without a landing page?

Yes, and it is often better. A post in the forum where the buyer already complains about the problem, with the price stated, gets a faster and more honest answer than a page nobody has found yet. The landing page matters when you want the test to keep running after you stop posting.