A fake door test puts an entrance to a product that does not exist yet in front of real prospects, such as a landing page with a Buy button, a pricing page or a menu item, and measures how many try to use it. Set the pass mark in advance, send enough targeted traffic, and tell people the truth when they click.
It is the cheapest test that measures behaviour rather than opinion. An interview tells you what someone says they would do. A fake door tells you what a stranger did when it cost them a click, an email, or a moment of reaching for their card.
The name comes from Polyvore cofounder Jess Lee, by way of Alberto Savoia's Pretotype It, which describes the technique as pretending a feature exists and seeing if anybody clicks on it. The same idea is also called a smoke test or a painted door.
Two famous examples, and what they actually tested
Buffer's founder Joel Gascoigne described the first test in 2011: a two-page site that explained the product and collected emails. When people signed up, he added a pricing page between the two, so visitors had to pick a plan before leaving an email. A small number clicked on paid plans, and that was the signal he took to start building. The second version is the important one. The first measured interest; the pricing step measured something closer to intent to pay.
Dropbox's version was a video. Eric Ries recounts that Drew Houston posted a roughly three-minute screencast of the product working, and the beta waiting list went from 5,000 to 75,000 people overnight. The product was hard to demo as a prototype, so the video stood in for it. It tested whether people had the file-sync problem badly enough to queue for a fix.
Pick the kind of door
There are two broad versions. Which one you use depends on whether you have a product yet.
| Variant | What it looks like | Best for | Watch out for |
|---|---|---|---|
| Landing page | A page describing the product, a price, and a Buy, Pre-order or Choose plan button | A new product with no existing users | You have to buy or earn traffic, and that traffic has to look like your real customer |
| In-product | A menu item, button or upgrade option for a feature you have not built | A new feature or paid tier inside a product people already use | Annoying existing users; keep it short-lived and explain immediately |
| Ad as the door | A search or social ad for the product; the click is the signal | Comparing several positionings or audiences cheaply | Ad clicks are cheap and curious; still send them to a page with a price |
Step 1: decide what counts as a pass, before you launch
This is the step almost everyone skips, and it is the one that makes the test worth running. Savoia's book puts the metric as a simple ratio: actions taken divided by opportunities for action. You choose both the action and the number in advance.
Write three things down and share them with someone who will hold you to them: the action that counts (clicked Buy on the pricing page, not visited the page), the audience (visitors from a specific ad or community, not your friends), and the threshold, with what you will do on each side of it. For example: "If at least 4% of visitors from the ad click a paid plan, we build the first version. Under 2%, we change the offer or drop the idea. In between, we rerun with a sharper page."
There is no universal right threshold. Base it on the economics. If you need a hundred paying customers and can realistically reach five thousand people in your niche, you need roughly a 2% conversion all the way to payment, so a click rate on a fake door well below that is a warning. Do the arithmetic for your own market; the market sizing guide shows how to get the reachable number.
Step 2: put a price behind the door
A free waitlist is the weakest door you can build. It costs the visitor nothing, so it measures curiosity rather than demand. Make the door ask for something, in rising order of strength: choosing a paid plan from a pricing table, entering an email after seeing the price, starting a checkout, or paying a refundable deposit.
Show the real price you intend to charge. A test run at a price you would never charge tells you nothing about the business you are going to run.
Step 3: send enough of the right traffic
Here is the honest arithmetic. A conversion rate measured on a small sample is mostly noise. The table shows the 95% margin of error around a measured rate, using the standard approximation for a proportion.
| Visitors | If you measure 5%, the true rate is roughly | If you measure 2%, the true rate is roughly |
|---|---|---|
| 100 | 0.7% to 9.3% | 0% to 4.7% |
| 400 | 2.9% to 7.1% | 0.6% to 3.4% |
| 1,000 | 3.6% to 6.4% | 1.1% to 2.9% |
Margins computed as 1.96 × √(p(1−p)/n). At low rates and small samples the approximation is rough, which is the point: below a few hundred visitors you cannot tell a 2% door from a 5% one.
So plan on at least a few hundred visitors per version you want to compare, and more if your threshold sits close to the rates you expect. If you are buying ads, multiply that by your expected cost per click and decide whether the answer is worth that budget before you start.
Who the visitors are matters more than how many. Two hundred people from a community of your exact customer are worth more than two thousand from a broad ad. Never count traffic from your own network; people who know you click to be kind.
Step 4: be honest with the people who knock
A fake door is only ethical if the deception lasts seconds. When someone clicks, tell them straight away that the product is not available yet, thank them, and offer something real: early access, a short call, or a note when it launches. That screen is also your best research opportunity. Ask one question, such as what they hoped it would do for them, and many will answer.
Do not take money for something you are not going to deliver. If you collect deposits or pre-orders, say clearly that it is a pre-order, give a delivery estimate, and refund promptly and without argument if you do not build it. Check your payment provider's rules on pre-orders before you charge anyone. In-product doors have an extra cost: an existing customer who clicks a button that goes nowhere is mildly annoyed, so run them briefly and to a small slice of users.
Step 5: read the result without flattering yourself
Compare the number to the threshold you wrote down, not to the one you wish you had written. Then look at what sits underneath it.
- Clear pass. Talk to the people who clicked, ideally that week. Those conversations are the start of your customer interviews, and the people are your first design partners.
- Clear fail. Before you drop the idea, check that the test was fair: was the traffic your real customer, was the page clear about what it does, was the price sane? If all three hold, the market has answered. If one does not, fix it and run once more, not five times.
- Borderline. Change one thing, such as the headline, the audience or the price, and rerun. Changing everything at once tells you nothing about which change mattered.
- High clicks, no follow-through. Lots of clicks on Buy but nobody replies to the follow-up or books a call usually means the page promised more than the product will deliver. That is useful to learn now.
Common mistakes
Most failed fake door tests fail in one of six ways, and none of them is the idea.
- Moving the goalposts. Deciding the threshold after seeing the number. Write it down first.
- Measuring the wrong action. Page views and scroll depth are not demand. Count the click on the door.
- Testing on friends. Your network converts out of goodwill. Exclude it.
- No price. A free signup proves people like free things.
- Too little traffic. Fifty visitors and two clicks is an anecdote, not a rate.
- Stopping at the click. The test's best output is the list of people who tried to get in. Talk to them.
Where a fake door fits
A fake door answers one question: will strangers try to get this at this price? It does not tell you whether the problem is real enough to keep them, how big the market is, or who else is already serving it. Run it after you have evidence the problem exists and before you write the real product. The validation tools guide covers what to use for each of the other questions.