Painted door tests

A painted door test puts a UI element - a button, a pricing tier, a feature link - in front of users for something that doesn’t exist yet. You measure how many people try to use it and decide whether the underlying feature is worth building. Sometimes called fake-door, smoke, or concierge tests depending on what happens after the click.

The use case is everywhere CRO bleeds into product. Should we add a subscribe-and-save option? Build the button, route it to a “joining soon, leave your email” page, see how many people click. Should we offer a £400 tier above the current top? Add it to the pricing page, route the click to a contact form. The cost is a few hours of build. The signal is whether anyone wants it.

Variations

• Smoke test. The lightest version. The button leads to a “coming soon - notify me” capture page. Measures click-through and email signups.
• Fake door. Same idea but the user lands on a “thanks for your interest, we’re not offering this yet” message. More honest about the state, slightly worse for trust.
• Concierge. The feature appears to work but is fulfilled manually behind the scenes. Useful when the feature has usability questions, not just demand questions.
• Wizard of Oz. The user thinks they’re using software; you’re answering tickets in the back. The full version of concierge.

What the result actually tells you

A painted door measures intent, not conversion. Clicks ≠ purchases. A 4% click-through on a “subscribe and save” button means 4% of visitors are interested enough to investigate. The conversion to actual subscriptions, once the feature is built, will be a fraction of that. Plan accordingly.

The other limit: painted doors measure top-of-funnel interest. They don’t measure whether anyone would have paid, used, or retained. A high click rate can fund the decision to build, but pricing and viability still need a real test once the feature exists. Treat painted doors as evidence for strong hypotheses, not as substitutes for them.

Where it goes wrong

The biggest risk is trust erosion. A user who clicks a button expecting a feature and gets a “joining soon” page is mildly annoyed. A user who hits that three times across the site stops believing anything works. Use painted doors sparingly per session, and follow up with something that makes the click worth their time - early access, a discount when it launches, a useful answer.

The other failure mode is treating the result as conversion data. Painted door tests should be reported separately from primary metric results because they’re measuring a different thing. Mixing them into the same dashboard invites the wrong comparison and lets one bad painted-door win drag a real ship decision the wrong way.