Before an MVP there is a cheaper instrument, and most teams either skip it or run it in a way that guarantees a flattering answer. It is a page that describes a product as though it already exists, with a real call to action, shown to people who have never heard of you.

Done properly it costs a few days and a small ad budget, and it can save a quarter. Done carelessly it produces a number that means nothing and gets quoted in a pitch deck for the next eighteen months.

What it is actually measuring

The page is not measuring whether people like your idea. It is measuring whether a stranger, having read a description of the value, will perform a small deliberate act at a rate higher than strangers normally do.

That act is the whole test. A click, an email address, a card entry on a pre-order, a booked call. The stronger the act, the stronger the evidence — and the fewer people will do it.

This is sometimes called a smoke test or a fake door test. The mechanism is the same: present the offer, count the behaviour, compare against a benchmark you set in advance.

Set the threshold before you look

This is the step that gets skipped, and skipping it invalidates everything downstream. Human beings are extremely good at finding a story that makes whatever number they got sound encouraging.

Decide beforehand: what rate, from what traffic source, at what sample size, means we proceed?

Useful reference points, drawn from practitioner guidance on fake-door testing:

  • Internal buttons — a stub inside an existing product — typically see 2–5% click-through for genuine interest, with above 5% reading as strong demand.
  • External ads to cold audiences should beat the channel's own average, roughly 1% for paid social and 3–5% for search, where intent is higher.
  • Sample size of at least 200 views gives a directional signal; 500 or more per variant before comparing variants against each other.

Those are starting points, not laws. The rate that matters is relative to the benchmark for your channel and your market, and a B2B page selling a £2,000-a-month product should expect far lower raw conversion than a consumer waiting list — while each conversion is worth vastly more.

Two published examples give a sense of scale: Polyvore saw roughly 2% of users click a "buy this outfit" button before the functionality existed; Notino recorded 3.5% on a language-switcher stub. Neither number is impressive out of context. Both were decisive in context.

What goes on the page

Keep it to one screen of substance. You are testing a proposition, not a website.

A specific headline naming the outcome and the audience. Not "the future of X." Something a member of the target group would read and think that is about me. If your headline could describe three competitors, it is not testing your idea — it is testing the category.

The problem, in their words. This is where the discovery conversations pay off. Language lifted from interviews outperforms language written in a workshop, consistently and by a lot.

Three concrete capabilities. What it does, plainly. No adjectives doing structural work.

A price, or a clear signal of price band. This is the part teams are most tempted to leave out, and leaving it out is what makes the result unreliable. Interest at an unknown price is not interest. If you are not ready to commit to a number, say "from £X per month" or "pricing for teams of 10–50."

One call to action. One. A page with "join the waitlist" and "book a demo" and "download the guide" measures indecision.

Enough credibility to be plausible. A real company name, a real face, a way to contact a human. A page with no evidence that anyone is behind it converts badly for reasons unrelated to your idea.

What to leave out: a long feature list, a roadmap, a team section, a blog, an FAQ that argues with the reader. Every additional element gives people somewhere to go that is not the call to action.

Where the traffic comes from

This choice affects the result more than anything on the page.

Paid search is the highest-quality traffic for this test, because people searching for a solution have already decided they have a problem. If you cannot find search volume for the problem you are solving, that is a finding in itself — it may mean the problem is unarticulated, which is not fatal but changes your go-to-market completely.

Paid social reaches people who were not looking. Cheaper, lower intent, and much more sensitive to creative. Useful for consumer products and for discovering which framing resonates.

Communities — a relevant subreddit, a Slack group, a niche forum — produce small samples of extremely well-qualified traffic. Handle with care and disclose what you are doing; the reputational cost of getting this wrong in a community you need is high.

Your own network is the one source to exclude. They are not strangers, they are not representative, and they will click to be supportive. Keep the sample clean.

Budget is usually smaller than people expect. A few hundred pounds against a well-chosen search term gets you to a directional answer.

The honesty problem

You are showing people something that does not exist. The guidance from practitioners is consistent and worth following: keep it honest.

In practice that means the page after the click tells the truth — that the product is in development, that you will be in touch, and gives them a way to leave. It means not taking money for something with no delivery date unless you are genuinely prepared to deliver or refund. And it means not running the same fake door repeatedly at the same audience, which erodes trust for a result you already have.

There is also a self-interested argument here. The people who convert on this page are your first fifty users and your best source of interviews. Burning them for a data point is a poor trade.

Where the test lies to you

A good rate with bad economics. The most useful lesson in the Dropbox history is not the video. It is the paid-acquisition experiment that failed: roughly $233 to acquire a customer for a $99 product. Conversion was not the problem. The arithmetic was. Always compute what a converted user cost you, and compare it to what they could plausibly be worth.

A good rate from a tiny sample. Three conversions out of forty visits is 7.5% and means nothing. The confidence interval on small samples is wide enough to drive a business into a wall.

A good rate on the wrong promise. If your headline promises something you cannot build, you have validated a product that is not yours. Read your page back and ask whether the thing you intend to ship would satisfy the person who clicked.

A bad rate from a bad page. This is the inverse error and it kills good ideas. If the page is confusing, slow, or visually untrustworthy, you have measured your execution rather than your idea. Before concluding the idea failed, check the obvious: does it load quickly, does it read clearly on a phone, does the call to action work.

That last point is not a small caveat. Page experience genuinely affects whether people stay long enough to decide — Google's Core Web Vitals thresholds put a good largest contentful paint at 2.5 seconds or less and cumulative layout shift at 0.1 or below, measured at the 75th percentile of real visits. A page that misses those badly is measuring its own load time.

Reading the result

Clearly above benchmark: proceed, and go interview everyone who converted. Their reasons are your positioning.

Clearly below: something is wrong with the proposition, the audience, or the framing. Change one at a time. Most teams change all three at once and learn nothing.

Right at the benchmark: this is the genuinely difficult result, and the honest answer is that you do not have a signal yet. Test a different framing or a different audience segment before deciding. Do not proceed on a marginal number because you want to start building — that impulse is exactly what the test was designed to check.

A result that says stop is not a wasted week. It is the cheapest no you will ever get.

Then what

If the signal holds, you have a segment, a message that worked, and a small list of people who have already raised their hand. That is a much better starting position for scoping a build than a specification written from assumptions — and it makes cutting scope without cutting the value a far less arbitrary exercise.

It is also worth keeping the page. It becomes the first version of the real site, the message has already been tested against strangers, and you can keep measuring it as the product arrives. The work we do in web design and development for early-stage companies often starts exactly here — a single page that has to earn attention from people with no reason to give it, before there is a product to point at.

And if you have not yet done the conversations that make the headline worth writing, that comes first: validating the idea is what gives this page something true to say.

Common questions

What conversion rate means my idea is validated?

There is no universal number. Set the threshold against your channel's benchmark before you run the test: roughly 1% for paid social and 3–5% for paid search as baselines to beat, and 2–5% click-through for a stub inside an existing product. What matters is beating the benchmark for your channel and audience, not hitting an absolute figure.

How much traffic do I need for a reliable result?

At least 200 visits for a directional signal, and 500 or more per variant before comparing two versions against each other. Below that, the confidence interval is wide enough that an encouraging number and a discouraging one are statistically indistinguishable.

Is a fake door test dishonest?

Not if it is run transparently. The page after the click should say the product is in development, give a realistic sense of timing, and let people leave. Taking money for something with no delivery plan, or running the same stub repeatedly at the same audience, crosses the line and costs you the trust of your earliest users.

Should I show a price on the test page?

Yes. Interest at an unknown price is not interest, and omitting price is the most common reason these tests produce optimistic, unusable numbers. If you have not settled on a figure, show a band or a starting point rather than nothing.

Sources and further reading