Ways to run the test
How do I run a smoke test with a landing page?
Updated 19 August 2026 · Thresholds read live from the verdict engine
Short answer
A smoke test is one page that states the promise, shows the thing, names the price and asks for a commitment, plus enough tracking to see where people stop. Buy 300 to 500 cold visitors, change nothing while it runs, then read the funnel from top to bottom. The stage with the worst drop tells you what to fix, and it is usually the ad or the headline rather than the idea.
What the page has to do in eight seconds
A cold visitor arrives from an ad with no context and low patience. Before they scroll, the page has to answer three things: what this is, who it is for, and what it costs. Everything else on the page is support.
That is why the classic mistakes are so expensive. A clever headline that hides the category costs you the visitor at second two. A page with no image of the product leaves them assembling one from your adjectives. A page that hides the price until checkout converts curiosity instead of intent, and you find out too late.
The page, block by block
Above the fold
- Headline. The outcome, in the words your audience uses. Match the phrasing of the ad they clicked, or the mismatch will read as a mistake.
- One line of support. Who it is for and the mechanism, in under twenty words.
- One image. The product, a mockup on a device, or the outcome in the physical world. One, not a carousel.
- The button. The verb that matches what happens next. Not "Learn more", which measures nothing.
Below it, in this order
- The problem, in their sentences. Two or three lines lifted from reviews or forum posts. This is where a cold visitor decides you understand them.
- How it works. Three steps, one line each. Screenshots beat descriptions.
- Price and plan. Visible without a click. If you plan a trial, show the trial and what happens after it.
- Proof. Anything true. A founder note, a named beta user, a number from your own testing labelled as such. Invented testimonials are both dishonest and easy to spot.
- The objection you know is coming. Two lines answering it. Privacy, effort, switching cost, whichever applies.
- The button again. Same words as the first one.
Four events, or the test is unreadable
A failed smoke test with only a page-view counter cannot tell you whether the offer was rejected or the button was broken. Track these four, in any tool you like.
| Event | Fires when | What a drop here means |
|---|---|---|
| Page view | The page renders for a confirmed human visitor | A gap against ad clicks means slow loading or bot traffic |
| CTA click | The main button is pressed | Weak headline, wrong audience, or a price that stops them cold |
| Checkout reached | The commitment screen renders | Friction or fear in the step between, usually the form |
| Checkout completed | Card stored or pre-order taken | The price, the trust gap, or a broken payment step |
Plausible, PostHog, Vercel Analytics and a plain Google Analytics 4 property all do this in an afternoon. Add UTM parameters on every ad so each creative can be read separately, since one bad creative can drag a good page under the kill line.
Sending traffic without poisoning the sample
- Budget for 300 to 500 visitors. At a $1 to $2 cost per click in the US or UK that is $400 to $800 of spend, and a fraction of that in cheaper markets, which read differently.
- One country per test. Buying power varies enormously and a blended rate across ten countries answers no question you actually have.
- Three creatives, one ad set. Splitting a small budget across five ad sets means none of them ever leaves the learning phase.
- No incentives. No giveaway, no discount for signing up, no prize draw. They lift the top of the funnel and destroy the meaning of the bottom.
- Keep your own people out of it. Exclude your IP, and if you post the link to your audience, tag it so you can read cold and warm traffic apart.
Reading the drop-offs
Work down the funnel and stop at the first stage that is clearly below the expected range. Fixing anything below that stage is wasted effort until the stage above it is healthy.
| Stage | Typical range on cold paid traffic | If you are under it |
|---|---|---|
| Ad click to page view | 70% to 90% | The page is slow or the link is broken. Test on a real phone. |
| Page view to CTA click | 10% to 30% | Headline and audience mismatch. Rewrite the promise before touching the design. |
| CTA click to checkout reached | 40% to 70% | The step in between asks too much. Cut fields, cut screens. |
| Checkout reached to completed | 15% to 40% | Price, trust, or a payment step that scares people. Say plainly that nothing is charged. |
Worked example
The drop that was not the idea
A test on a $49 a year budgeting app returned 0.6% visitor to checkout, which sits under the kill line. The funnel underneath told a different story: 22% of visitors clicked the button, and 71% of those reached the checkout, but only 4% of the people who reached it finished.
The checkout was asking for a full billing address before it explained that nothing would be charged. They moved the reassurance above the form and reran the same creatives. Completion at the last step went to 26%, and the funnel landed at 2.7% overall. The idea was never the problem.
The numbers
The whole test lands on one ratio, and these are the bands the verdict engine holds it to.
- 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
There is a second number worth watching if your page collects an email before the commitment step. Of the visitors who reach the email step, under 8% handing one over means the step itself is broken, and 25% or more means it is doing its job.
Read them in order. A page converting at 1% with a healthy email step has an offer problem. A page converting at 1% with a leaking email step has a page problem, and you have not measured the offer yet.
How VerifyToLaunch does this
VerifyToLaunch generates the page as a six-screen funnel instead of one long scroll, because splitting the pitch across screens is what web2app funnels do and it gives every stage its own measurement. The four events are wired in from the start, bot traffic is filtered before anything is counted, and the dashboard shows each stage against the range it should be hitting.
If you prefer a single page you have written yourself, everything above works and we would rather you ran that test than no test. What you lose by hand is the per-stage benchmark and the confidence interval, both of which you can rebuild with a spreadsheet if you enjoy that sort of thing.
Common questions
- What is a smoke test in product terms?
- It is a small, cheap version of the offer put in front of the audience you would sell to, built to produce a measurable action rather than an opinion. The name is borrowed from hardware testing, where you power the board and see whether smoke comes out.
- Should the landing page look finished?
- It should look credible, which is a lower bar than finished. Clear typography, one good image and no broken links are enough. Visitors forgive a plain page and abandon a confusing one.
- Can I run a smoke test without a price?
- You can, and it becomes a waitlist test measuring curiosity. That is fine when you want reach for launch day. It is a weak input to a build decision, because email is free to give.
Read next
- 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 conversion rate should I expect?Benchmarks by traffic type and ask, plus how to diagnose a bad rate stage by stage.
- How do I test an idea with Meta ads?Campaign settings field by field, three creatives worth testing, and the policy traps.
- What is a web2app funnel and what converts?Why apps sell on the web first, the screens in order, and the rate each one holds.
Get the page and the tracking in one step
The generated funnel carries your promise and your price across six screens, with visits, funnel starts, email capture and checkout tracked from the first visitor and benchmarked stage by stage.
One-time licence, 30-day money-back guarantee