How to Run a Smoke Test to Prove Demand
A smoke test puts a real offer in front of real people and measures whether they act, before you build anything. Done well, it saves months. Done lazily, it tells you nothing. Here is how to run one you can trust.
What a smoke test actually measures
The goal is action, not opinion. Asking people if they like your idea is worthless, because most will say yes to be polite. A smoke test asks them to do something that costs a little effort or commitment: click a buy button, enter an email to join a waitlist, or fill out a request form.
The action is the signal. If people will not take a small step now, they will not take a bigger one later. You are looking for proof that the problem is real enough to make someone move.
Build the page in an afternoon
You do not need the product. You need a single page that describes it as if it exists.
A good smoke test page has:
- A clear headline naming the problem and who it is for.
- Three or four lines on what the product does and why it helps.
- One price, even if you have not finalized it. Price changes behavior, and you want real behavior.
- A single call to action. Buy now, join the waitlist, or request access.
Use any landing page builder or a simple site. Keep it honest. Describe what you genuinely plan to build, not a fantasy you cannot deliver.
Decide what counts as a pass before you start
Pick your success threshold in advance, or you will rationalize whatever number you get. A common honest setup looks like this:
- Define the action. For example, clicks on a buy button that lead to a "coming soon, reserve your spot" step.
- Set a target. Out of 100 visitors who match your audience, how many need to act for the idea to be worth building? A few percent for a paid action is a reasonable starting bar.
- Write it down. "If at least 4 of 100 targeted visitors reserve a spot, I proceed."
The exact number depends on your price and channel. The point is committing to it before the data tempts you to move the line.
Send real traffic, not random traffic
The test only works if the visitors match your buyer. A hundred people who would never buy will sink any idea, and a hundred friends will float a bad one.
Ways to get matched traffic:
- A small paid ad campaign aimed at the audience, with a tight interest or keyword target.
- Posting in communities where your buyers already gather, where it is welcome and not spammy.
- Direct outreach to people who fit, sending them to the page.
- A relevant newsletter or partner with the right readers.
Aim for at least a few hundred targeted visitors so the result is not noise. Track where each visitor came from, because a buy click from a paid stranger means far more than one from a friend.
Read the result honestly
When the traffic lands, three things can happen.
- People act at or above your threshold. The demand signal is real. Move to pre-selling or building a thin first version.
- People click the buy button but bail at the reserve step. Interest exists, but something about price, trust, or the offer is off. Worth investigating, not abandoning.
- Almost no one acts. The problem is not painful enough, the audience is wrong, or the message does not land. Better to know now.
Watch for false positives. Email signups are cheap and can flatter you. A click on a payment button, even one that stops at "coming soon," is a stronger signal because it sits closer to a real decision.
What to do next
A passing smoke test is permission to keep going, not a guarantee. The next step is to turn soft interest into harder commitment: a pre-order, a deposit, or a paid pilot. Each step you can get people to take raises your confidence and lowers your risk.
A failing test is a gift. You spent an afternoon and a small ad budget instead of three months of building. Change one variable, the audience, the price, or the message, and run it again.
Before you build the page, it helps to know which pain to put in the headline and which audience to send. A DemandSonar scan surfaces the real complaints and the people feeling them, so your smoke test points at demand that already exists instead of a problem you hope is there.