You have the idea. You can see the first screen already. And the honest thing nobody says out loud is that building it is the fun part, so that is the part we start with.
Then four months later there is an app and about nine people have downloaded it. You find out what the market thought at the worst possible moment to find out.
Here is the mistake underneath that, and it is not laziness. We validate the idea, by asking people whether they like it. The thing worth validating is whether anybody is already paying to solve this, and what they are unhappy about while they pay.
So below are four ways to answer that, ordered by how long each one takes. Including the one we used before designing a single screen of an app that launched at 32% trial-to-paid, which is the share of free trials that turn into paying subscriptions.
What each one actually tells you
They are not alternatives. They answer different questions, and the first one answers the most important.
Competitor and revenue check
Is anybody making money here, and how much? Sensor Tower, Appfigures or Sensor Tower’s free tier will get you an estimate.
Answers: does demand exist at all.
Review mining
Read the one and two-star reviews on the four biggest apps in the category. Write down the complaints that repeat.
Answers: what the incumbents are failing at, in customer words.
Landing page test
A page describing the thing, an email field, and paid traffic pointed at it. Costs real money and needs real volume.
Answers: whether your specific promise is interesting.
Fake door
A real-looking purchase flow that ends in “coming soon”. Strong signal, and the one with an ethics bill attached.
Answers: whether people will act, not just nod.
If you only do one, do the first. It is days of work and it tells you the thing that decides everything else.
AI Decorator is the clearest example I have, because the validation happened before any design did.
Before a single screen was drawn, we reviewed more than ten top-performing AI home design apps and used Sensor Tower to estimate their revenue. It ran from $400K to over $2M a month.
Then we tore down the top four in detail: the AI tooling each one used, the business model, and what sat behind the paywall. Features where the AI results were weak got cut rather than shipped.
The app launched into a category with large, established competitors and reached 32% trial-to-paid from release. The launch itself is a separate story, and the short version is that none of it was decided on launch day.
Notice what that process did not include. Nobody was asked whether they would use it.
This is the part that runs against instinct. Finding four well-funded competitors feels like finding out you are too late.
It is the opposite. Those competitors are proof that people in this category open their wallets, which is the single hardest thing to establish and the thing you cannot create. An empty category is usually empty for a reason, and the reason is rarely that nobody thought of it.
What you are looking for is not a gap in the market. It is a specific complaint that repeats across the reviews of every app already in it.
Daphne Tideman, in How to Make an App People Will Pay For, recommends mining app store reviews, competitors’ included. Her reason is that users describe their situation and the outcome they wanted in their own words.
Those words are better than the ones you invented at your desk, and they are free. They are also the words that should end up in your screenshot captions and your keyword field, which means this research pays for itself twice.
Sort by most recent and most critical. Look for the complaint that appears under every app in the category, not the one that appears under the worst one.
“Would you use an app that…” People are kind, and the cost of saying yes to a friend is zero. Nothing about a hypothetical answer predicts a download, let alone a payment.
“How much would you pay for…” Worse, because it sounds like market research. Stated willingness to pay is reliably higher than actual willingness to pay, and you are asking somebody to price a thing that does not exist yet.
Asking people who like you. Your friends, your followers, the people in your Discord. They are answering a question about you, not about the product.
If you are going to talk to people, ask about the last time they had the problem. What did they do? What did they use? Did they pay for anything? That is a question about a past event, and past events are the only thing anybody can report accurately.
A landing page or a fake door gives you a number, and the number has the same problem as any other test: it needs to be big enough to mean something.
Forty visitors and three sign-ups is not a 7.5% conversion rate, meaning the share of visitors who act. It is three people.
The sample size table applies here just as it does to a paywall test. At the volumes an unlaunched idea can buy, most demand tests are a feeling with a decimal point attached.
Which is fine, as long as you call it that. A judgement call recorded honestly is a better input than a number you have over-read.
Start with whether anybody is already making money in the category, because that is the hardest thing to establish and the thing you cannot create. Estimate competitor revenue, tear down the top three or four, then read the one and two-star reviews across all of them for the complaint that repeats. That takes days and it answers the question that decides everything else.
No, usually the opposite. Established competitors are proof that people in that category pay for things. AI Decorator launched into a category with large incumbents, after a check that found competitor revenue running from $400K to over $2M a month, and reached 32% trial-to-paid from release. An empty category is more often empty for a reason.
No. People are kind, and the cost of saying yes to a friend is zero. Stated willingness to pay is also reliably higher than the real thing, and you are asking somebody to price something that does not exist. If you talk to people, ask what they did the last time they had the problem, because a past event is the only thing anybody can report accurately.
Only at a volume most unlaunched ideas cannot buy. Forty visitors and three sign-ups is not a 7.5% conversion rate, it is three people. The same sample size arithmetic that governs a paywall test governs this one, so either get the traffic or record the decision as a judgement call rather than a result.
Conversion design for subscription apps.
Screenshots, onboarding and paywalls, connected around one goal: getting more users to pay.