A fake door test, also called a painted door test, shows visitors a button, link or offer for something that does not exist yet and counts how many try to use it.
The click is the signal: if too few people "knock" on the door, you have learned there is weak demand before spending anything on building the real thing.
Fake door testing in CRO and A/B testing
Alberto Savoia describes the Fake Door as one of his "pretotyping" techniques in The Right It, and credits the name to Jess Lee, then CEO and cofounder of Polyvore.
The idea is to put up a front door (an ad, a page, a button) that pretends the product exists, then measure how many people try to go through it.
On a website, it usually looks like one of these:
- A new nav link or menu item, such as "Subscribe and save" or "Gift cards", that leads to a "coming soon" message.
- A pricing tier or add-on shown on the pricing page that has not been built.
- A product-page button, such as "Save to wishlist" or "Try it on", that opens a short note and an email sign-up instead of the feature.
Run it as a proper experiment: show the door to a random share of traffic and keep a control group that never sees it.
That lets you measure two things, how many people click (the demand signal, often reported as a click-through rate) and whether the dead end costs you orders.
Track that second number as a guardrail metric.
The click is not a purchase, and it is not a promise.
Write the decision rule in your hypothesis before launch, for example "build it if at least 3% of product-page visitors click", so a weak number can't be explained away afterward.
Also tell clickers plainly that the feature isn't ready, and offer to notify them.
Leaving them on a broken page wastes their time and wears down trust.
What our database shows about testing a feature before building it
We don't have a pure painted door experiment in the database yet.
What we do have shows why checking demand first can pay off, and why clicks alone can mislead.
A fully built feature that didn't pay. Clean Commit tested adding a wishlist to a Shopify store: heart icons on product cards and product pages, login required, plus a wishlist link in the header.
Over 48 days and 113,055 visitors, conversion rose 4.2% and add-to-cart rose 4.3%, but revenue per visitor stayed flat, AOV fell 3.9%, and the result wasn't significant.
A fake door, a heart icon that just counted clicks, could have shown the level of interest for far less work.
The same feature working elsewhere. Clarins UK added a wishlist option to listing and product pages and saw transactions rise 1.84%, with more basket page views.
Demand for a feature depends on the store and its customers, which is the case for testing your own audience instead of copying someone else's result.
Clicks are not the same as revenue. Blend Commerce added a Subscribe and Save link to the mobile navigation for Jackson's.
Visits to the subscription page went up 24.77%, which is the kind of number a fake door produces.
But the business outcome was much smaller: conversion rose 3% and total revenue 3.4%, and AOV and revenue per visitor fell slightly as more people subscribed.
Expect your click-through rate to overstate real uptake.
Blend Commerce
Clean CommitRelated terms
- Experimentation: the broader practice that fake door tests are part of.
- Control and variant: the two groups you compare when the door is shown to only some visitors.
- Canary release: the step after a fake door, where the real feature goes to a small share of users first.
- Feature flags: how most teams switch a fake door, and later the real feature, on for part of their traffic.
Browse more terms in the glossary, or search all test results for features others have already tested.