Start here
How do I test an app idea before building it?
Updated 19 August 2026 · Thresholds read live from the verdict engine
Short answer
Build the smallest thing that can produce a no. For most consumer apps that is one page carrying the promise and the price, a checkout step that never charges, and $300 of cold traffic pointed at it over five days. You are looking for the share of visitors who commit, and the bar is 1% at the bottom and 2% at the top.
Pick the smallest test that can produce a no
A test is worth running when it can fail. Most founders design tests that cannot: they show the idea to people who like them, ask a question nobody can answer honestly, and treat enthusiasm as a result. Start from the opposite end. Write down the number that would make you walk away, then design the cheapest way to observe it.
The ladder below runs from cheapest to most convincing. Climb it only as far as your decision needs. If a $200 fake door says no three times across three angles, you never need the expensive rungs.
| Rung | Time | Cash | The no it can produce |
|---|---|---|---|
| Desk research | 2 hours | $0 | Nobody is spending here and no competitor has run an ad for a year. |
| Ten problem interviews | 1 week | $0 to $200 in incentives | The pain is real but people have already solved it another way. |
| Waitlist page plus ads | 3 days | $150 | Cold visitors will not even give an email for this promise. |
| Fake door with your price | 1 week | $300 | They want it in theory and will not commit at the price you need. |
| Concierge run for ten people | 2 to 3 weeks | Your labour | They pay, then stop caring by week three. |
The seven-day plan
This is the version we would run for a consumer subscription app. Two evenings of building, five days of traffic, one hour to read it.
- Day 1. Write the promise. One outcome sentence, one line of proof, one price. Steal the phrasing from how people describe the problem in reviews and forum posts, since that is the language your ad has to match.
- Day 1. Pick the audience you can buy. Name a country, an age band and an interest you can actually target. If you cannot describe the audience in ad-manager terms, you cannot test the idea this way.
- Day 2. Build the page and the commitment step. One screen of copy, one image, the price, the button, then a checkout that stores a card without charging or takes a refundable pre-order.
- Day 2. Instrument it. Page view, click, checkout reached, checkout completed. Four events, one tool.
- Day 3. Launch three creatives at $40 a day. One campaign, one ad set, broad targeting, one tracked link per creative so you can tell the angles apart.
- Days 3 to 7. Leave it running. Editing an ad set resets its learning and throws away the first days of delivery. Check spend once a day, change nothing.
- Day 8. Read it top to bottom. Impressions, clicks, visitors, funnel starts, checkouts. The worst drop in that chain is the thing to fix, and often it is not the idea.
The four things you have to make
Everything else is optional. These four decide whether the test means anything.
1. A promise a stranger can repeat
The headline states what the person has after using the app, in their words. Feature lists fail here because a cold visitor has no reason to assemble them into a benefit. If ten people cannot repeat the promise back to you after five seconds on the page, the test will measure your copy, not your idea.
2. Something to look at
One image of the thing. A mockup in Figma, a screen recording of a clickable prototype, even a well-shot photo of the physical problem. Pages with no visual of the product convert badly and the drop tells you nothing useful.
3. A commitment step that costs something
An email costs a visitor nothing, so it measures curiosity. A card, a refundable pre-order or a deposit costs trust and attention, and that is the behaviour that predicts revenue. Keep the wording honest on that screen and you stay on the right side of both platform policy and your own conscience.
4. Four events you can trust
Page view, click, checkout reached, checkout completed. Without the middle two a failed test is unreadable, because a broken button and a rejected offer look identical from the outside.
Matching the test to the kind of app
The fake door funnel is the default for consumer apps because those buyers can be bought on Meta and decide in one session. Other shapes need other tests.
| What you want to build | The test that fits | What to measure |
|---|---|---|
| Consumer subscription app | Paid traffic into a web funnel carrying the price | Visitor to checkout rate |
| B2B tool | 40 cold emails or 20 LinkedIn messages offering a paid pilot | Booked calls, then signed pilots |
| Marketplace | Test the hard side alone. Recruit ten suppliers by hand before any page exists. | Suppliers who agree to list and answer a message |
| Local or service business | One city, one landing page, geo-targeted ads or flyers | Bookings or deposits per 100 visitors |
| Developer tool | A README, a waitlist and a post where those developers already argue | Stars, replies from strangers, install requests |
Worked example
A B2B test that cost nothing but a week
A founder wanted to build a compliance tracker for dental practices. Instead of a funnel, they sent 60 plain-text emails offering a paid pilot at $200 a month, described in one paragraph, with a calendar link.
- 60 emails sent, 11 replies
- 6 calls booked
- 2 practices agreed to pay for a pilot before anything existed
Two paying pilots out of 60 emails is a workable outbound rate for a product at that price. It also produced two customers who would tell them what to build, which is worth more at that stage than any conversion percentage.
When testing first wastes your week
- You can build it in a weekend. If the build is two days, the build is the cheaper test and it produces a product at the end.
- You are the customer and you are unusual. Tools for people like you often start as something you needed. Ship it, then check whether anyone else is like you.
- The value only appears after use. Habit trackers, note tools and social products can look weak on a paywall and thrive once someone lives in them for a fortnight. Test them, but weight retention over intent.
- You cannot buy the audience. Doctors, procurement teams and school heads do not click consumer ads. Go and talk to twenty of them.
The numbers
Whatever shape the test takes, the bar it gets held to is the same one, and it comes from the verdict engine.
- Funnel starts before any verdict
- 30
- Below this, don't build
- 1.0%
- At or above this, build
- 2.0%
- Visitors before a segment counts
- 50
Nothing is judged under 30 funnel starts, so a test that buys 40 visitors has not run yet. Plan for 300 to 500 visitors on a first read, which at a $1 cost per click is $300 to $500 of traffic.
How VerifyToLaunch does this
VerifyToLaunch collapses days 1 and 2 into about twenty minutes. You describe the idea in a paragraph, it drafts the positioning, builds the six-screen funnel with your price on the paywall, generates ad angles with creatives, and gives you a tracked link per angle. Tracking for all four events is wired in already, and the dashboard shows the funnel with the drop at each stage.
Days 3 to 7 are still yours. You launch the links in your own ad account and pay for the traffic yourself, because we have no ad-platform API. If your idea is a marketplace or a B2B pilot, the funnel is the wrong instrument and the manual tests above will serve you better.
Common questions
- How much should I spend on a first test?
- Between $250 and $500 of traffic in a country whose buying power matches your price. That buys roughly 300 to 500 visitors at typical Meta costs, which is enough to settle a result that is clearly good or clearly bad.
- Can I test without spending on ads?
- You can post in communities where the audience already gathers, but the traffic is warm and small, so the number it produces is optimistic and fragile. Use organic traffic to sharpen the promise, then buy cold traffic to measure it.
- What if the page converts but I cannot build the app?
- That is a good problem and a common one. Demand for something technically hard is still demand. Either narrow the promise to a version you can ship in six weeks, or sell the manual version by hand while the software catches up.
Read next
- How do I validate an app idea?The whole method: pick a commitment that costs the visitor something, buy cold traffic, count who commits.
- Should I build an MVP or validate first?What each path costs, what each measures, and when building really is the cheaper test.
- What is a fake door test and how do I run one?Offer the thing before it exists, count who reaches for it, and stay on the right side of the ethics line.
- What does validating an idea cost?The bill line by line, and what $150, $400 and $1,000 each buy in certainty.
Get the test built for you
Describe the idea and get the funnel, the ad angles, the creatives and the tracked links in about twenty minutes, with the verdict maths already wired in. The ad budget and the ad account stay yours.
One-time licence, 30-day money-back guarantee