Fake Door Testing SOP

A fake door testing SOP measures demand for one feature or offer by counting how many people try to open it before it exists, and a full test runs for one to two weeks. You add a button, a menu item, a pricing tier or an email link that looks real. When someone clicks, a short message says it is coming soon and asks if they want to hear when it launches. A click is a decision made in context, in the middle of a real task, which makes it a sharper signal than a survey answer about the same feature.

What is fake door testing and when should you use it?

A fake door (also called a painted door) is an entry point to something you have not built. Alberto Savoia lists the Fake Door among his "pretotype" types in The Right It, next to the Mechanical Turk and the Pinocchio, and pairs it with his XYZ Hypothesis: "At least X% of Y will Z." The format forces you to state, before the test, what share of which people must take which action.

Use this SOP:

• When you have existing traffic (users, subscribers, store visitors or an ad budget) and want to test one new thing.
• When the team argues about which feature to build next. Three fake doors settle it with behaviour.
• When you want to test a premium tier or an add-on before writing any code or contracts.

A fake door tests one offer at the moment of choice. To test a whole new product on its own page, use the landing page test SOP. To compare several features on paper with customers, use the feature prioritization SOP.

What do you need before you start?

• An audience that already sees your product: at least a few hundred active users, a newsletter, or a $100 to $500 ad budget.
• One clearly named offer in a few words: "Auto-file VAT return", "Team plan", "Gift subscription".
• Event tracking on the door, with PostHog, Google Tag Manager or your analytics tool.
• A way to show the door to part of the audience, such as a feature flag in PostHog or LaunchDarkly.
• A baseline: the click rate of an existing link or button in a similar spot.

How do you run it, step by step?

1. Write the XYZ hypothesis

Fill in Savoia's format with numbers: "At least 10% of active users who see the menu will click 'Auto-file VAT return', and 40% of those will ask to be notified." Writing it first prevents you from calling any result a success.

2. Choose where the door goes

Place the door where the need appears: in the menu next to related features, as a third plan on the pricing page, as a link in your newsletter or as a separate ad. Context is the whole point. A door hidden in a settings page tests nothing.

3. Design the door and what's behind it

The door must look like every other button in that spot. Behind it, show a short, honest message: "We're building this. Want to hear when it's ready?" with a one-click "Notify me" button. Add the planned price if you want to test that too.

4. Show it to a measured share of your audience

Use a feature flag to show the door to 10% to 20% of users, so you can compare them with users who don't see it. For ads or emails, send to one segment at a time. Note the start date and the number of people exposed.

5. Run for one to two weeks

Let the door collect at least a few hundred exposures. Keep everything else on the page unchanged so the click rate means something.

6. Measure the funnel

Count exposures, clicks, notify-me clicks and the click rate of your baseline element over the same period. The notify-me rate separates curiosity from need: anyone clicks a shiny new button once, fewer ask to hear more.

7. Follow up with the clickers

Email everyone who asked to be notified within a few days. Ask what they hoped the feature would do and what they do today instead. Book five short calls. These conversations tell you what to build when the numbers pass.

8. Close the door and decide

Remove the door at the end of the test, even if it passed. Compare results with your hypothesis, write the decision down and, if you build, tell the notify-me list first.

Template you can copy

XYZ HYPOTHESIS
At least [X]% of [Y: who, where, when] will [Z: action].
Second step: at least [X2]% of clickers will [click Notify me / reply / pick a price].

DOOR COPY
Button / menu label: [2-4 words, same style as neighbouring items]
Behind the door:
  "[Feature] is on its way.
   We're building it now and want to get it right.
   [Notify me when it's ready]   [No thanks]"
  Optional: "Planned price: $[x]/month as an add-on."

TEST LOG
Start date | End date | Exposed users | Door clicks | Click rate |
Notify-me clicks | Notify-me rate | Baseline element click rate |
Follow-up replies | Calls booked | Decision

Worked example: a VAT feature in a freelancer app

This example is illustrative. A bookkeeping app for freelancers has about 2,400 monthly active users. The team debates adding automatic VAT filing. They add "Auto-file VAT return" to the main menu for 20% of users for two weeks, with a note behind the door that it is planned as a $6 per month add-on.

Their hypothesis was 10% clicks and 40% notify-me. Both passed, and the door beat the baseline. In calls, most freelancers said they wanted reminders and pre-filled numbers more than full automatic filing, which shrank the first build to a smaller feature.

What should you measure, and when do you stop?

Stop after one to two weeks or when you reach your planned exposure count, whichever comes first. A fake door stays up only as long as it teaches you something: each extra day disappoints more users. If results land near the line, rerun with a clearer label before you decide, because a vague label depresses clicks on a good idea.

Which mistakes should you avoid?

• Leaving the door up for months. People who click again and again learn not to trust your buttons.
• Hiding the honesty message. Say right away that the feature is not ready. Never show an error screen instead.
• Collecting payment. A fake door asks for clicks and emails. Taking money for an unbuilt feature is a pre-order campaign with its own rules.
• Using a vague label. "New tools" gets clicks from the curious and tells you nothing.
• Testing too many doors at once. Three new menu items compete with each other and with the real product.

How can AI speed this up?

AI assistants write five label variants for the door in seconds; pick the most literal one. They can also draft the behind-the-door message and the follow-up email. After the test, paste the call notes into an assistant and ask it to list the jobs people expected the feature to do. The human check: you decide the exposure share, and you read the replies yourself before you build anything.

The free F/MS idea validator gives a quick first read on a new feature idea before you put a door in front of users, and the startup playbook module places tests like this in a full startup plan.

Where does this SOP fit in your validation plan?

Fake door testing sits in Solution Validation and is the fastest behaviour test once you have any traffic.

1. Before: rank feature ideas with the feature prioritization SOP, so you only build doors for the top candidates.
2. Set up: make sure events fire correctly with digital behavior tracking.
3. After: if the door passes and you build, measure real use with product usage analysis and keep testing ideas with the growth experiment SOP.

See the whole Solution Validation category.

FAQ

Is fake door testing ethical?

Yes, when you tell clickers right away that the feature is coming, limit the exposure to a share of users and a short period, and never take money. Thank people and give them the first access when it launches.

What is a good click-through rate for a fake door test?

There is no universal number. Compare the door with a similar existing element in the same spot and with the X% you wrote in your hypothesis before launch.

How long should a fake door test run?

One to two weeks, or until a few hundred people have seen the door. Longer tests annoy users without adding much data.

What is the difference between a fake door test and a landing page test?

A fake door tests one feature or offer inside an existing product, email or ad, at the moment of choice. A landing page test presents a whole new product on its own page to visitors who arrive from outside.