Run the ad before you build the thing

The cheapest version of a feature is the one you never build. Not the scrappy weekend version, not the half-finished branch you abandon in October — the one you cancel on a Tuesday afternoon because the page you put up for it got a hundred and forty visits and two people who cared enough to leave an email. That afternoon costs you a lunch. The alternative costs you a quarter.

The move is old and deeply unglamorous: write the ad, build a one-screen page that makes the promise, spend a small fixed amount, count what comes back. You are not testing headlines or button colours here. You are asking one question — does this promise pull anyone at all — and paying a small amount of money to get an answer from strangers instead of from your own optimism.

What you are actually measuring

The number that matters is not the number of clicks. Clicks tell you the ad worked. What you want is the fraction of people who arrived, read the promise, and then did something slightly inconvenient about it.

Here is the shape of it with made-up arithmetic, to keep it concrete. You put a hundred and twenty dollars behind one ad over eight days. At eighty cents a click that buys roughly a hundred and fifty visits. Your page makes a single promise, in plain words, with one field: leave your email and we will tell you when it exists. Nine people do. That is six percent, and six percent of strangers volunteering their inbox for a thing that does not exist yet is a real signal. Now run the same test and get one signup out of a hundred and fifty. That is not a weak yes. That is a no with a receipt.

The reason both numbers mean anything is that you have something to compare them to. You already know roughly what your existing pages do — your homepage, your best landing page, the last thing you launched. If your normal page converts visitors to signups at four percent and this new promise pulls at one, the promise is the weak part, not the traffic. Without that baseline you are staring at a percentage with no opinion about it. This is the same reason a single number you steer by beats a wall of charts: comparison is what turns a measurement into a decision.

a landing page is a question, and the ad is just how you get someone to hear it

The three ways this test lies to you

The first is volume. A hundred and fifty visits and nine signups feels like data, but shift two people either way and your percentage moves by more than a third. This is the same trap as trying to A/B test on small traffic — the arithmetic of small numbers does not care how badly you want a clean answer. A demand test survives it only because you are not looking for a two-point difference. You are looking for the gap between "several people did this" and "essentially nobody did", which is a much coarser and more honest question.

The second is that you may have tested the ad rather than the idea. One creative, one audience, one clumsy first line, and a perfectly good promise dies of a bad sentence. If the click-through rate was miserable, you never got to test the promise at all — almost nobody read it. Look at the two rates separately before you conclude anything: people who clicked, then people who acted. A low click rate is a story about the ad. A healthy click rate with nothing after it is a story about the promise.

The third is the sneakiest: you measured curiosity, not intent. An email field costs a person nothing. If you want a signal with weight behind it, ask for something that costs a little — a fifteen-minute call at a specific time, a reply describing what they would use it for, a card on file for a launch discount. You will collect far fewer of them, and every single one will be worth more than a dozen idle addresses. Choose deliberately which of those two you are buying before you write the page.

Write the decision rule before you spend

Do this part first, while you are still capable of being disappointed. One sentence, written down before the ad ever runs: if fewer than eight people leave an email by Sunday, I do not build this before winter. Once the results are in, you will be able to argue yourself into anything — the season was wrong, the audience was off, the wording was clumsy — and every one of those arguments will feel reasonable, because some of them will be true. The rule you wrote on Monday is the only version of you that had no stake in the outcome.

And be precise about what a no actually means. It is not proof that nobody wants the thing. It is proof that this promise, to this audience, at this price, this week, did not pull. That is worth one more attempt with a genuinely different angle — a different buyer, a different problem framed first, a different price implied. Two honest failures with two different stories is a shelf. One is a coin toss.

Our part in this is deliberately small. We will draft the ad and the page copy from what you tell us the thing does, and the campaign arrives paused, so nothing spends a cent until you have read every word of it and pressed the button yourself. The signups land next to the rest of your numbers, so the six percent you just measured sits beside the four percent your homepage has been doing all year rather than floating in a spreadsheet with nothing to lean on. What we will not do is tell you the answer was good. Eight was your line, not ours.

Build the page first. It is the cheapest thing you will ever ship, it takes an evening, and it is the only part of the feature that can tell you whether to build the rest.