Last updated
A fake door test puts a control for a feature on the page before the feature exists. A button, a pricing tier, or a menu item is real.
A fake door test puts a control for a feature on the page before the feature exists. A button, a pricing tier, or a menu item is real. The thing it promises is not.
The pattern is also called a painted door. The visitor sees an affordance that a normal product would honor. On click, they get a clear stop: the feature is not ready, sometimes with a way to leave an email. You never pretend the click completed the job. Because nothing was built behind the door, a high click rate only answers "would anyone ask for this?" A low rate answers it too, which is the point of running the test before the build.
Building the feature first and then discovering nobody uses it is the expensive order. A door reverses it. The cost is trust. If the stop message is late, or the button looks like a purchase, you have trained people not to believe the next control. Use a fake door when the build is large and the demand is actually unknown. Skip it when you already know people want the feature and the open question is how it should work. That second question needs a real version.
A B2B app considers a CSV export. Instead of building it, you add "Export CSV" on the report page for half of users. The other half does not see the link. Of 2,000 visitors who see it, 160 click, which is 8%. The click opens a short note: export is not available yet, and they can join a waitlist. You compare 8% with the threshold you set, say 5% of report viewers, and you compare the waitlist submits with the clicks. 160 attempts and 40 waitlist joins is a reason to scope the export. It is not a measurement of how often people would finish an export that does not exist.
Write the audience, the control, the click you will count, and the build threshold before launch. Make the dead end immediate and plain. Do not take payment. Run it long enough to clear the sample a click-rate test needs, including people who already know the product. If you test two labels, randomize them and keep the same stop message. Afterward, either build, or remove the door. Do not leave a permanent control that goes nowhere.
A common mistake is counting page views of the button as demand. The metric is the attempt. Another is shipping the door to everyone with no control, so you cannot tell a curious click from the traffic you would have had anyway. A third is using the click rate as the business case for a feature whose value happens after the click, such as retention. The door never saw that behavior.
It is a promise you cannot keep yet. The ethical version tells the visitor immediately that the feature is not available and, if you collected a click, what happens next, such as a waitlist or nothing. Leaving people in a dead end, or charging them, is not a demand test. It is a broken page.
The one you wrote down before the door went live, based on the cost of building and the value of a user who tries it. A 4% click rate is not a universal pass mark. Compare the rate with the closest real control on the same page, and decide from the threshold in the plan, not from how exciting the first day felt.
Yes, if the question is which offer people try, not which finished product is better. Randomize the label or the placement, keep the dead end honest and identical, and read the click rate. Do not describe the winner as a proven feature. You have only proven the attempt.
Visual editor, AI-generated variants with MidaGX, redirect testing, and GA4 integration — free forever in the Mida Sandbox. No credit card required.
Still deciding?
A quick screen share on your actual site — no slides, no generic tour. Just your questions answered.
30 min · no commitment · no sales pressure