Fake Door Test: How to Validate Demand Without Writing a Single Line of Code

A founder spends a weekend building a real, working web app for an idea she has not yet validated. The product is polished. The landing page is polished. She spends $400 on Instagram ads. The ads drive 1,200 visitors. Seven people sign up. She spends the next six months trying to figure out what went wrong.

Nothing went wrong with the product. The product was fine. The product was answering a question almost nobody in the channel she chose was asking. The same six months, spent on a single landing page with no working product behind it, would have produced the same answer — at one-fiftieth of the cost. She would have found out that the click-through was 0.6%, that the language on the page produced confusion, and that the channel was wrong for the segment. Then she would have changed one of those three things before writing a line of code.

That is what a fake door test is for. It is the cheapest credible demand test in the validation toolkit. A founder can run one in a week, with no product, no engineering, and a few hundred dollars in ad spend. The result is not a verdict and it is not a green light. The result is a smaller list of things the founder could still be wrong about — which is the entire purpose of the validation loop.

This article is the strategic pillar for the fake door test. The operational playbook (page copy, CTA wording, traffic sources, conversion tracking, the eight-step launch checklist) lives in the Fake Door Test Guide. This page owns the why, the when, the decision rule, and the failure modes.

Quick answer

A fake door test is a pre-build demand test in which a founder exposes an interface element — a button, a pricing card, a waitlist signup, a feature callout — for a product or feature that does not yet exist, then measures how many visitors perform the targeted action. If the action rate clears a pre-committed threshold on qualified traffic, the founder has evidence that the described product is worth building. If it does not, the founder has learned that at least one of the four inputs — the audience, the channel, the value proposition, the price — is wrong, without having spent engineering time finding out.

The term "painted door" comes from Alistair Croll and Benjamin Yoskovitz's Lean Analytics (2013). The practice is older: a saloon that painted a door on a blank wall and watched to see if anyone tried to walk through it. The same idea, applied to software products, became a named validation technique once Lean Startup vocabulary entered practitioner use in the early 2010s. Eric Ries described the same method as the "smoke test" in The Lean Startup (2011); "fake door" and "smoke test" are now used interchangeably in practice.

The six steps, in order, are: (1) name the assumption being tested, (2) define the single action that counts, (3) build the smallest surface that can produce it, (4) pre-commit to the decision threshold, (5) drive qualified traffic to the surface, (6) read the result against the threshold and decide. The two most-cited case studies are Buffer's October 2010 two-page landing site (Joel Gascoigne collected roughly 120 email signups and 15% click-through on the pricing page before writing product code) and Dropbox's Spring 2008 demo video (Drew Houston grew the waitlist from ~5,000 to ~75,000 in a single weekend before public launch).

The method is cheap. It is fast. It is also the easiest validation test to misread, because the four inputs — audience, channel, proposition, price — are confounded in a single number. The rest of this article is the discipline of separating them.

Key takeaways

• A fake door test is not a website. It is an experiment. A real landing page with a working product is a launch. A page that exists only to measure whether the described product is worth building is a fake door test. The two look identical from the visitor's side; the founder's intent is what makes it a test.
• The original term is "painted door," from Alistair Croll & Benjamin Yoskovitz's Lean Analytics (2013). The metaphor is older than the book — saloons in the 19th-century American West reportedly painted doors on blank walls and watched for collisions. Croll and Yoskovitz named the technique; the technique was already in use.
• The smoke test is Eric Ries's term for the same method in The Lean Startup (2011). "Smoke test" and "fake door test" are now used interchangeably in practice. A founder who has read one book will recognise one name; a founder who has read both should know they describe the same test.
• The four inputs are confounded in a single number. Audience (who saw it), channel (where they came from), value proposition (what the page said), and price (if shown) all affect the action rate. A 1% conversion rate could mean "the audience doesn't want this" or "the page is wrong" or "the channel is wrong." The test cannot separate these on its own.
• The Unbounce 2024 Conversion Benchmark Report is the most-cited public benchmark. Across 41,000+ landing pages, 464M visits, and 57M conversions, the median landing-page conversion rate across all industries was 6.6%. B2B SaaS sits lower — the median is roughly 2.1% to 3.0%, with the top quartile at 6.8% and the top 10% at 7.12%. A smoke-test landing page should be benchmarked against the same vertical's top quartile, not against the all-industries median.
• The minimum useful sample is large enough that the result is statistically distinguishable from noise. A 1% action rate on 20 visitors is not evidence of demand. A 1% action rate on 1,000 visitors is evidence of weak demand. The number matters because small samples routinely produce extreme results that don't replicate.
• A fake door test does not prove willingness to pay. A waitlist signup measures interest under zero friction. A deposit or pre-order measures interest under cost. The two produce different evidence, and a founder who confuses them will sign up a thousand people and convert none.
• The two documented case studies are real but not reproducible by instruction. Buffer's October 2010 landing page succeeded partly because Joel Gascoigne already had a Twitter audience and a network of early-adopter peers. Dropbox's Spring 2008 demo video succeeded partly because it was a video — a format that had not yet been saturated on Digg. Reproducing the play without the conditions produces a different result.

What a fake door test actually is

A fake door test exposes an interface element that promises something the product does not yet deliver, then measures the rate at which visitors attempt to use the element. The element is usually a button, a form, a pricing card, a feature callout, or an entire landing page. The promise is usually a waitlist signup, a deposit, a "we'll email you when it ships," or a multi-step pre-order flow.

Three things make the test a test rather than a launch.

First, the visitor is told (or allowed to believe) that the product is real. A landing page that says "Sign up to be notified when we launch" is doing this. A landing page that says "Beta testing this concept — sign up to participate in a 30-minute user research call" is doing this differently — the visitor knows the product is at an early stage and is being asked to participate in shaping it. Both can be fake door tests. The first is closer to the painted-door method; the second is closer to a concierge pilot with extra steps.

Second, the founder has pre-committed to the decision rule before traffic arrives. "If 5% or more of 500 qualified visitors sign up, I'll build the product. If under 2%, I'll change the value proposition. Between 2% and 5%, I'll change the channel and re-test." A founder who decides what the number means after seeing it has not run a test; the founder has run a survey of their own preferences.

Third, the test is bounded. A founder running a fake door test is not collecting feedback, interviewing visitors, or iterating on the page in real time. The page is frozen; the traffic source is named; the duration is fixed. Bounded tests produce results that can be compared. Unbounded tests produce anecdotes that look like results.

A fake door test is not the same thing as a coming-soon page. A coming-soon page is a placeholder. A fake door test is a measurement instrument whose result will decide what the founder builds next. The distinction is the founder's intent, not the page's design.

The painted-door origin

Alistair Croll and Benjamin Yoskovitz named the technique in Lean Analytics (2013), in the chapter on disruptive innovation techniques. The term is a metaphor. In the nineteenth-century American West, a saloon that wanted to know whether customers would use a back exit reportedly painted a door on the wall and watched to see if anyone tried to walk through it. If customers kept bumping into the wall, that was evidence a real door was needed. If nobody tried, the wall was fine.

The same logic, applied to software, is: show the affordance, see whether anyone attempts to use it, and only build what people try to walk through. The Croll & Yoskovitz framing places the technique in a family of low-cost pre-build tests — landing pages, concierge MVPs, Wizard-of-Oz setups, paper prototypes — all of which exist to test the same hypothesis (would the customer act on this if it existed) at a cost far below the cost of building it.

How it relates to the smoke test

Eric Ries described a method he called the smoke test in The Lean Startup (2011). The smoke test, in Ries's framing, is the smallest possible version of a product used to test whether customers are interested enough to pre-order. A pre-order page with a "buy now" button on a product that doesn't exist is a smoke test. A landing page with a "sign up" form on a product that doesn't exist is also a smoke test. The fake door test is the same method, viewed from the visitor's side: instead of offering the product, the test exposes only the affordance that would lead to the product.

In current practitioner use, the two terms describe the same family of tests. A founder reading both books will recognise the family even if the names differ. The important thing is not which name the founder uses; it is whether the founder treats the result as a measurement or as a confirmation.

When a fake door test is the right test

A fake door test is the right test when the founder needs evidence of one specific thing — whether the named customer will perform the named action for the described product — and is willing to accept that other assumptions remain untested. That last clause is the one most often skipped.

The test is well-suited to:

• Testing the value proposition. "If we describe the product this way, will qualified visitors attempt to use it?" A founder with two different framings of the same product can run two fake door tests, one per framing, on the same channel, and compare the action rates. The result is a clear winner among framings, not a verdict on the product.
• Testing the audience-channel fit. "If we run ads in this channel, will the visitors be the right kind of people?" A founder running cold traffic from a new channel can use the fake door test as a one-week channel audit. The test is not whether the product is good; it is whether the channel reaches qualified people at an acceptable CAC.
• Testing the price point (crudely). "If we show this price, will visitors attempt to buy?" A founder comparing two price points can run two parallel pages, one with each price, and measure the click-through. The test is not a substitute for willingness-to-pay research; it is a directional sanity check that the price is not visibly wrong.
• Testing the segment. "If we describe this product for this segment, will they attempt to use it?" A founder who has narrowed the segment but is not sure the narrowing is right can run parallel fake door tests, one per segment description, on the same channel.

The test is poorly suited to:

• Testing willingness to pay. A waitlist signup is not a purchase. A deposit or a pre-order is closer, but still not a purchase. The fake door test measures interest under zero cost; willingness-to-pay tests measure commitment under cost. The two are not interchangeable.
• Testing retention. A fake door test produces one action, not a return visit. The signal is about the first move, not the second. Founders who need retention evidence should run a paid pilot or a concierge MVP, not a smoke-test landing page.
• Testing a solution's effectiveness. A landing page describes a product. It does not deliver the product's outcome. The evidence is about the visitor's intent to try, not about the product's ability to deliver.
• Testing in the absence of qualified traffic. A landing page reached only through the founder's personal network measures the network's willingness to act, not the segment's. The test must reach qualified visitors through a named channel that does not depend on the founder's existing audience.

How to run a fake door test (the six steps)

The six steps below are ordered so that each one produces information the next one needs. Skipping a step is possible; each skip produces a test whose result cannot be cleanly interpreted.

Step 1 — Name the assumption being tested

A single sentence in the form "We believe [named customer] will attempt to [specific action] when exposed to [specific description]." If the founder cannot write the assumption in this form, the founder is not yet ready to test. The action must be observable. The customer must be specific. The description must be the one the page will use, not the one the founder believes in private.

The discipline of naming the assumption is the same discipline Steve Blank introduced in the Customer Discovery framework — a startup is a search for a repeatable business model, and the search begins with hypotheses. A founder who runs a fake door test without naming the hypothesis is running an opinion poll.

Step 2 — Define the single action that counts

One action. One. A founder who designs a page with three CTAs ("sign up for updates," "join the beta," "follow us on Twitter") has designed three tests, not one. The test result is the action rate on the chosen CTA, measured against the qualified-visitor count.

The action can be a waitlist signup, a deposit, a pricing-card click, an "add to cart," a calendar booking, or a multi-step form completion. The harder the action (multi-step form vs. single email field), the lower the action rate and the cleaner the evidence. The easier the action (single email field), the higher the action rate and the more confounded the signal.

Step 3 — Build the smallest surface that can produce it

A landing page. One page. One description of the product. One CTA. No navigation menu, no blog link, no "about us" page, no other CTAs. The page's job is to produce the one action. Anything else on the page is a competing argument for the visitor's attention.

The page should describe the product in the customer's vocabulary. The vocabulary comes from the problem interviews (Step 1's underlying customer discovery). A page that uses the founder's vocabulary measures the founder's ability to describe the product, not the customer's willingness to try it.

The mechanical details of the page — the headline formula, the social-proof placement, the CTA wording, the form length, the tracking setup — are in the Fake Door Test Guide. The strategic decision on this page is that the surface is one page, not three.

Step 4 — Pre-commit to the decision threshold

Before traffic arrives, the founder writes down four numbers: the action rate that justifies building, the action rate that justifies re-testing the proposition, the action rate that justifies changing the channel, and the action rate that justifies shelving the idea. The thresholds are not universal; they depend on the cost of building, the cost of changing channels, and the founder's other constraints.

A defensible default for B2C SaaS-style landing pages, derived from the Unbounce 2024 Conversion Benchmark Report: the median landing-page conversion rate across all industries was 6.6%, B2B SaaS sits at roughly 2.1–3.0%, the top quartile of B2B SaaS lands around 6.8%, and the top 10% at 7.12%. A solo founder running cold traffic to a smoke-test page should expect something in the middle of that range; a top-of-funnel B2C page with a strong hook and warm channel traffic can reasonably aim higher.

The threshold the founder writes down should be benchmarked against the right vertical, not the all-industries median. A B2B SaaS founder who benchmarks against the 6.6% cross-industry median will set a threshold that is too high and interpret every result as a failure.

Step 5 — Drive qualified traffic to the surface

Qualified traffic is the variable the founder can least afford to skimp on. A landing page reached through the founder's personal network measures the founder's network, not the segment. A landing page reached through organic SEO measures the SEO position, not the proposition. A landing page reached through targeted paid ads in a named channel measures the channel-proposition fit, which is closer to what the founder needs to know.

For a B2C consumer product, qualified traffic often means a small ad spend in one or two named channels (Reddit, Facebook, Instagram, TikTok), with audience targeting that approximates the named segment. For a B2B product, qualified traffic often means cold email to a named list, LinkedIn outreach to a named role, or sponsored content in a publication the segment reads. The mechanic matters less than the discipline: the founder must be able to name the channel and the segment before the test runs.

The minimum sample size is not a single number; it depends on the expected action rate and the size of the effect the founder is trying to detect. A rough rule of thumb, conservative on purpose: at least 200 qualified visitors per variant, ideally 500 or more. A 1% action rate on 20 visitors is not evidence of anything. The same rate on 1,000 visitors is evidence of weak demand.

Step 6 — Read the result against the threshold and decide

The result is one number: the action rate on qualified traffic. The founder compares it to the pre-committed threshold and takes the action the threshold specified.

Three failure modes show up at this step and produce false negatives or false positives:

• Reading a low action rate as proof the idea is bad. The action rate may be low because the page is wrong, the channel is wrong, the segment is wrong, or the price is wrong. The test cannot separate these. A low action rate is evidence that something is wrong, not evidence that the idea is wrong.
• Reading a high action rate as proof the idea will work. A 10% action rate on 50 visitors reached through the founder's Twitter network is not a 10% action rate at scale. The test result is conditional on the traffic the founder drove. A different channel may produce a different result.
• Iterating on the page during the test. Changing the headline, the CTA, or the price after the first 30 visitors arrive turns the test into a sequence of one-sample experiments. The result is no longer interpretable. The page is frozen for the duration of the test.

The decision rule the founder wrote in Step 4 is the only thing that should determine the next move. If the founder finds themselves arguing with the rule, the rule was too soft at the writing step — the founder should have written a stricter one before the data arrived.

How to read the results (the decision rule)

The output of a fake door test is one number: the action rate on qualified traffic. The number does not say "the idea is good" or "the idea is bad." It says one of three things, depending on where it falls against the pre-committed threshold.

Above the top threshold. The test passes. The named customer, reached through the named channel, will perform the named action when exposed to the described product. The next move is to build, with the assumption that the same action will convert to a paid action when the product exists. The next move is not to declare victory. The test is one piece of evidence in a sequence; the rest of the sequence (problem validation, willingness-to-pay validation, retention evidence) is still needed.

Between the two middle thresholds. The test is inconclusive. The most likely cause is a confounding input — the channel, the proposition, the audience, or the price. The next move is to change one input at a time and re-test. Changing more than one input per re-test makes the result uninterpretable. A founder who changes the channel, the headline, and the price between test 1 and test 2 has not learned which one was wrong.

Below the bottom threshold. The test fails. Something in the four-input system is sufficiently wrong that qualified visitors are not performing the action at a rate the model can support. The next move is to return to customer discovery — additional problem interviews, a re-examination of the segment definition, a search for a different channel — not a code sprint.

The decision rule is binary on purpose. A founder who treats the middle band as "interesting, let me run it a bit longer" turns a measurement into a justification. The middle band is for re-testing, not for hoping.

Two documented case studies

Two case studies show up in nearly every article on the topic. They are documented in the founders' own writing; they are not invented. They are also not reproducible by instruction; the conditions that produced them were specific to the moment. The two cases below are presented for what they are: real evidence that the method works, with the conditions named.

Buffer — October 2010

Joel Gascoigne, working on what became Buffer, had been building a social-media scheduling tool. He stopped, deleted the work in progress, and built a two-page website instead: a landing page that described the product and offered a signup form, and a pricing page that listed paid plans ($0, $5/month, $20/month) with a "click to sign up" CTA on each. There was no backend. There was no scheduling algorithm. There was no product.

He tweeted the link to his network. Roughly 120 people signed up for the updates list. Of the visitors who reached the pricing page, more than 15% clicked through to the paid plans, including the paid tiers — despite the fact that the plans did not yet exist. The signal was strong enough that he spent the next seven weeks building the actual product. The first paying customer arrived four days after launch.

The conditions that produced this result are not generic. Gascoigne already had a Twitter audience of early-adopter peers; he was reaching qualified traffic through a channel he had spent years building. The pricing page was a real-time test of willingness to act at a specific price, not just interest in the concept. The result should be read as evidence that the method can produce strong signal — and as a reminder that the audience-and-channel inputs were not zero.

Dropbox — Spring 2008

Drew Houston had been working on Dropbox for several months and had collected a waitlist of roughly 5,000 people. In March/April 2008, he posted a roughly three-minute demo video to Digg (with a Reddit cross-post). The video was a faked demonstration of the product — the file-syncing technology was not yet built — but it showed what the experience would look like, included Easter eggs aimed at tech early adopters (references to Office Space, Tron, XKCD, "Chocolate Rain"), and was, in his own framing, a fake demo of a real product.

The video hit Digg's front page. The waitlist grew from roughly 5,000 to roughly 75,000 in a single weekend. The team's internal target had been to push the list to roughly 15,000; the actual fifteen-fold jump was a surprise. The public launch followed in September 2008.

Again, the conditions matter. Houston was reaching an audience that was already primed to find the video interesting (early-adopter tech readers on Digg). The video format was undersaturated on Digg in early 2008, which contributed to its organic reach. A founder replicating the play in 2026 against a different channel-and-format combination will not produce the same result by copying the tactic. The play works because the inputs (audience, channel, format) match the test, not because the test itself is novel.

The two cases together make the case for the method: a pre-build test, designed to measure intent at near-zero cost, produced the evidence that justified a real build. The cases do not make the case that the method is a marketing tactic. They make the case that the method, run under the right conditions, produces signal.

Common mistakes

The seven failure modes below account for most of the fake door tests that produce false positives or false negatives. Each is a specific way a founder can mistake the result for something it is not.

1. Reaching the wrong audience. A landing page shared on the founder's personal Twitter account measures the founder's network's willingness to act, not the named segment's. The fix is to drive traffic through a named channel that does not depend on the founder's existing audience — paid ads, organic SEO, cold email, sponsored content. The test is not informative if the audience is the founder's audience.
2. Driving unqualified traffic. A landing page reached by visitors who do not match the named segment measures the channel, not the proposition. The fix is to name the audience before the test runs and to verify the traffic source's audience roughly matches. A 10% action rate on 1,000 visitors reached through "anyone who clicked a generic ad" is weaker evidence than a 5% action rate on 200 visitors reached through a tightly-targeted channel.
3. Using a multi-CTA page. A page that asks visitors to "sign up for updates," "follow us on Twitter," and "join the beta" produces a result that cannot be cleanly interpreted. The fix is one page, one CTA, one action. The result is the action rate on that single action.
4. Changing the page during the test. A founder who edits the headline after the first 50 visitors and edits the CTA after the next 50 has run two tests, not one. The fix is to freeze the page for the duration. Iteration belongs between tests, not during them.
5. Reading a high action rate as proof of willingness to pay. A 10% waitlist signup rate is not a 10% purchase rate. The fix is to treat the waitlist as a measurement of intent under zero friction, and to follow up with a separate test (a deposit page, a Van Westendorp survey, a paid pilot) to test commitment under cost. The full willingness-to-pay discipline is in How to Test Whether People Will Pay Before You Build.
6. Setting the threshold after seeing the data. A founder who decides "5% is good" after seeing a 4.8% result has not run a test. The fix is to write the threshold before traffic arrives, using the Unbounce vertical benchmark as a starting reference (B2B SaaS top quartile is roughly 6.8%; top 10% is roughly 7.12%) and adjusting for the founder's specific channel and segment.
7. Confusing the fake door test with the MVP. A fake door test is a measurement of intent at the proposition level. The MVP is the smallest version of the product that delivers the experience the customer is being asked to value. The two are different rungs in the validation sequence. The full discipline is in MVP Validation.

What the fake door test cannot prove

The fake door test is the cheapest credible demand test. It is also one of the weakest in absolute terms. A founder who runs a fake door test should know exactly what the test cannot do.

• It cannot prove the customer will pay. A waitlist signup measures intent under zero friction. A deposit or pre-order measures intent under cost. A purchase is the only signal that proves willingness to pay. The fake door test result is the first evidence, not the only evidence.
• It cannot prove retention. A signup says the customer was interested once. It says nothing about whether they will return, whether the product solves the problem at the second encounter, or whether they will refer. The full retention discipline is in Product-Market Fit Validation.
• It cannot prove the solution works. The page describes the product. It does not deliver the product's outcome. A founder who treats a strong fake door result as proof of solution effectiveness has skipped the rungs that test the actual experience.
• It cannot separate the four inputs. Audience, channel, proposition, and price all affect the action rate. A low result is evidence that something is wrong, not evidence that the product is wrong. A founder who has not designed the test to isolate one input at a time has run a confounded experiment.
• It cannot predict what happens at scale. A 7% action rate on 500 qualified visitors reached through one channel is not a 7% action rate at scale across all channels. The result is conditional on the inputs the founder chose. Changing the inputs may change the result.
• It cannot substitute for ongoing customer research. The fake door test is one experiment in a sequence. Teresa Torres's Continuous Discovery Habits (2021) argues that customer research is a weekly practice, not a one-time gate. The fake door test is the gate; it is not the practice.

The right mental model is that the fake door test narrows the range of things the founder could be wrong about. It does not eliminate that range. A clean test result says "this proposition, on this channel, with this audience, at this price, produces intent at a rate the model can support." It does not say "this idea will succeed."

How the fake door test fits into the broader validation sequence

The fake door test is one rung in the pre-code validation sequence. The sequence, in the order a solo founder typically runs it, is:

1. Problem interviews. Five to ten conversations with members of the named segment, asking about their current workaround. The cheapest evidence; produces evidence the problem is real and felt. The interview discipline is in The Mom Test Explained for Solo Founders and the stop rule is in How to Know When Customer Interviews Are Enough.
2. Fake door test (this article). A landing page, paid or organic traffic, a single CTA, a pre-committed threshold. Produces evidence that the named customer will act on the described proposition at the channel the model can support.
3. Concierge or Wizard-of-Oz pilot. Three to ten customers receive the planned outcome manually, by the founder. Produces evidence that the customer will pay the planned price for the planned outcome. The full discipline is in Concierge MVP Explained.
4. Pre-order, deposit, or LOI. A direct request for money or a written commitment. Produces evidence the customer will commit cost. The full discipline is in How to Test Whether People Will Pay Before You Build.
5. Channel wedge test. Evidence the founder can acquire the named customer through the named channel at a CAC the model supports. The full ranking is in Distribution Channels Ranked for Solo Founders.

Each rung produces evidence the others do not. A founder who has run only a fake door test has problem and demand evidence; the founder does not yet have outcome evidence (the concierge pilot), commitment evidence (the pre-order), or reach evidence (the wedge test). The fake door test is the cheapest rung; it is not the last rung. The full pre-code discipline, including all five checks and the seven-question build gate, is in How to Validate Demand Before Coding.

FAQ

Is a fake door test the same as a landing page MVP?

The terms overlap in practice and mean different things in theory. A landing page MVP is a landing page used as the smallest possible product. A fake door test is a measurement instrument whose purpose is to decide whether to build the product. A landing page can be both at once — a measurement instrument that also functions as a starting point for the product. The two are the same surface; the founder's intent is what differs.

How much traffic does a fake door test need?

A defensible minimum is 200 qualified visitors per variant, ideally 500 or more. The exact number depends on the expected action rate and the size of the effect the founder is trying to detect. A 1% action rate on 20 visitors is not evidence of anything; the same rate on 1,000 visitors is evidence of weak demand. The founder should plan the test budget around the sample size, not the other way around.

What is a good conversion rate for a fake door test?

It depends on the vertical, the channel, and the traffic temperature. The Unbounce 2024 Conversion Benchmark Report — the most-cited public benchmark, analysing 41,000+ landing pages — found a 6.6% median across all industries. B2B SaaS sits lower (median roughly 2.1–3.0%); B2B SaaS top quartile is roughly 6.8%; B2B SaaS top 10% is roughly 7.12%. A solo founder running cold traffic to a smoke-test page should benchmark against the vertical's top quartile, not the all-industries median, and pre-commit to a threshold in that range before the test starts.

Does a fake door test prove willingness to pay?

No. A waitlist signup measures interest under zero friction. A deposit or pre-order measures interest under cost. The fake door test produces intent evidence, not commitment evidence. A founder who has run only a fake door test has not yet tested whether the customer will pay. The willingness-to-pay discipline is in How to Test Whether People Will Pay Before You Build.

Can I run a fake door test without paid traffic?

Yes, but the result will be conditional on the traffic source. Organic SEO traffic measures SEO position, not proposition. Twitter traffic measures the founder's network, not the segment. Cold email traffic measures the segment, but with high friction that depresses the action rate. The cleanest test uses one named channel, ideally paid, with audience targeting that approximates the segment. The founder should name the channel and its limitations before the test runs.

What is the difference between a fake door test and a smoke test?

They describe the same method. "Smoke test" is Eric Ries's term from The Lean Startup (2011); "fake door test" is the term popularised by practitioner use following Croll & Yoskovitz's Lean Analytics (2013). The mechanism is identical: show an affordance for a product that does not exist, measure the action rate on qualified traffic, decide whether to build. A founder who has read only one of the two books should recognise the method even if the name differs.

Should I run a fake door test before or after problem interviews?

After. A fake door test is a measurement instrument, not a discovery instrument. The proposition on the page is built from the vocabulary the customer uses in problem interviews. A founder who has not yet run interviews is testing the founder's vocabulary, not the segment's. The five-check sequence — interviews, fake door, concierge, pre-order, wedge — places the fake door test as the second rung, after interviews. The full sequence is in How to Validate Demand Before Coding.

How is this different from a coming-soon page?

A coming-soon page is a placeholder. A fake door test is an experiment with a pre-committed decision rule. The visual is the same; the intent is different. A coming-soon page does not change the founder's behaviour regardless of its traffic. A fake door test changes the founder's behaviour depending on its result. If the page is not connected to a decision, it is not a test.

Summary

A fake door test is the cheapest credible demand test in the validation toolkit: a single landing page, a single call to action, qualified traffic, a pre-committed decision threshold. The method was named "painted door" by Alistair Croll and Benjamin Yoskovitz in Lean Analytics (2013) and was described as the "smoke test" by Eric Ries in The Lean Startup (2011); the two terms are now used interchangeably. The four inputs (audience, channel, proposition, price) are confounded in a single number, which is why the test is easy to misread; the right mental model is that the test narrows the range of things the founder could still be wrong about, not that it eliminates that range.

The six steps are: name the assumption being tested, define the single action that counts, build the smallest surface that can produce it, pre-commit to the decision threshold, drive qualified traffic to the surface, and read the result against the threshold and decide. The benchmark for "good" depends on the vertical; the Unbounce 2024 report places B2B SaaS top quartile at roughly 6.8%. The two documented case studies — Buffer's October 2010 two-page site and Dropbox's Spring 2008 demo video — show what the method can produce under the right conditions; neither is reproducible by instruction alone.

The fake door test is the second rung in the pre-code validation sequence, after problem interviews and before the concierge pilot, the pre-order, and the channel wedge test. It produces demand evidence at the proposition level; it does not produce outcome, commitment, or reach evidence. A founder who treats the result as a green light has skipped the rungs the result does not cover.

Related reading

• How to Validate a Startup Idea Before Building — the four-assumption framework this page sits inside.
• How to Validate Demand Before Coding — the five pre-code checks; the fake door test is Check 2.
• Fake Door Test Guide — the operational playbook: page copy, CTA wording, traffic sources, tracking setup.
• Landing Page Validation — the broader landing-page method that contains the fake door test.
• MVP Validation — what changes after the pre-code checks are complete.
• The Mom Test Explained for Solo Founders — the behavior-over-opinion discipline behind a clean test.
• How to Know When Customer Interviews Are Enough — the stop rule for the first rung of the sequence.
• Concierge MVP Explained — the third rung, after the fake door test.
• How to Test Whether People Will Pay Before You Build — the discipline the fake door test does not cover.
• Distribution Channels Ranked for Solo Founders — the channel evidence the fake door test depends on.
• Startup Validation Glossary — definitions for smoke test, fake door test, MVP, ICP, willingness to pay.

Sources

• Alistair Croll & Benjamin Yoskovitz, Lean Analytics (2013) — the painted-door / fake-door technique as a low-cost pre-build validation tool. The metaphor derives from a saloon practice of painting doors on blank walls to test demand. Reference at leananalyticsbook.com.
• Eric Ries, The Lean Startup (2011) — the smoke test as the smallest version of a product used to measure customer interest; the Build–Measure–Learn loop. Reference at theleanstartup.com.
• Steve Blank, Four Steps to the Epiphany (2005) — Customer Discovery methodology; hypothesis-driven testing of startup assumptions; "get out of the building." Reference at steveblank.com.
• Rob Fitzpatrick, The Mom Test (2013) — the polite-yes problem; behavior-over-opinion as the only honest signal in pre-build validation. Reference at momtestbook.com.
• Teresa Torres, Continuous Discovery Habits (2021) — customer research as a weekly practice, not a one-time gate; pairs with the fake door test as a complementary, ongoing input. Reference at continuousdiscoveryhabits.com.
• Joel Gascoigne, Buffer (October 2010) — the two-page landing site and pricing-page click-through that produced the first ~120 signups before the product existed. Founder's own account at buffer.com/blog and Joel Gascoigne's blog.
• Drew Houston, Dropbox (Spring 2008) — the demo video posted to Digg that grew the waitlist from roughly 5,000 to roughly 75,000 in a single weekend. Founder account summarised at startupfounderstories.com/stories/drew-houston-dropbox-demo-video and via Steve Blank's lecture materials on Customer Discovery.
• Unbounce, Conversion Benchmark Report (2024) — 41,000+ landing pages, 464M visits, 57M conversions analysed; 6.6% median across all industries; B2B SaaS top quartile ~6.8%; top 10% ~7.12%. Reference at unbounce.com/conversion-benchmark-report.

Next action

If the founder has not yet run a fake door test, the next action is to write down the single assumption the test will examine, in the form "We believe [named customer] will attempt to [specific action] when exposed to [specific description]." Then name the single action that counts. Then name the channel and the qualified-traffic plan. Then write the decision threshold before any traffic arrives.

If the founder has already built a product, the next action is to run the test anyway on the next product. The fake door test is not a substitute for building; it is the cheapest way to know whether the build is justified. A founder who runs it once will run it again.