Doubling the budget is a new campaign

Something finally works. After a month of campaigns that spent money and returned polite nothing, one of them comes back with fourteen signups for six hundred dollars, and the arithmetic says forty-three dollars each, and forty-three dollars each is a price you would happily pay all year. The reflex that follows is almost universal and almost always wrong in its sizing: triple the budget on Monday. That reflex is where a large share of small ad budgets quietly go to die, and the reason has nothing to do with the campaign being fake. The campaign is real. The number is the problem.

Start with how thin the number is. Fourteen conversions is not a rate, it is fourteen events, and events at that count move around a lot on their own. Run the same campaign, unchanged, for another three weeks and you might get nine, which is a cost of sixty-seven dollars each, or you might get nineteen, which is thirty-two. Nothing about the campaign changed between those outcomes. If you scale on the good fortnight and then read the following month as a failure, you have drawn a conclusion about your budget increase from what was mostly the ordinary wobble of small numbers. The first honest move is to say out loud how wide the band around forty-three dollars actually is.

The second reason is structural, and it is the one that catches people who have already made peace with the first. Raising a budget does not buy more of the same thing. Ad platforms spend early money on the cheapest, most obviously interested people they can find for you, and there is a finite supply of those. When you double the budget, the extra money reaches further out — people who match less well, who are less ready, who need more convincing. So the honest question is never "what did this campaign cost per signup", it is "what will the next signup cost". Say the first five hundred dollars brought twelve signups and the second five hundred brings six. Your average now reads fifty-six dollars, which still looks fine on the dashboard, while the money you just added was actually buying signups at eighty-three.

the average cost tells you what happened. the marginal cost tells you what happens next.

Which points at the practical rule: raise in steps small enough that you can still read the result. Something like twenty or thirty percent at a time, then wait until the campaign has accumulated roughly as many conversions as the batch that convinced you in the first place, and only then look and decide again. This is slower than it feels like it should be, and the tradeoff deserves naming rather than hiding: if the campaign is a genuine winner, you left money on the table for a few weeks. That is the whole downside. The downside of the fast version is that you cannot tell whether the dip you are looking at came from the budget, the audience running thin, or the week — and not knowing costs you far more than a few weeks of underspending, because it makes the next decision guesswork too.

Change one thing while you do it. The temptation at the moment something works is to improve everything at once: raise the budget, add two audiences, swap in the new creative you have been sitting on, widen the countries. Do that and the campaign will move — up or down — and the movement will be uninterpretable. Keep the raise boring and isolated. Where the platform lets you edit a live campaign rather than rebuilding it, edit it, because a fresh campaign starts its learning over and the first days of a restarted campaign are not comparable to the last days of the old one.

While you wait for conversions to accumulate, the useful early signals are the cheap ones upstream. Cost per click and click-through rate move long before conversion counts do, and they tell you the same story sooner: if cost per click steps up right after the budget does, that is the platform reaching into a less responsive audience, showing up two weeks before it shows up in your cost per signup. It is the same reading discipline as [what the first $100 of ad spend can tell you](/blog/what-the-first-100-of-ad-spend-can-tell-you) — the shallow metrics resolve fast and cannot tell you about revenue, but they are the leading edge of the number that can.

There is also a ceiling, and it is worth knowing what hitting it looks like so you do not mistake it for a mistake you made. Check frequency: if the same people are seeing your ad four times instead of more people seeing it once, the extra budget is buying repetition, not reach. For a narrow niche this arrives much sooner than anyone expects — sometimes at a few thousand dollars a month, sometimes less. The answer at that point is not a bigger number in the same box. It is a different audience, a different offer, or a different channel, which is a new experiment with its own small starting budget and its own honest waiting period, not a continuation of this one.

The bookkeeping matters more here than anywhere else, because the whole thing turns on comparing a before and an after that are weeks apart. Write down the date you raised it, the old and new numbers, and what you expected to happen — that last one especially, because a prediction you recorded is the only thing that can later tell you whether you learned something or rationalised something. Our own answer to this is deliberately unexciting: campaigns launch paused so nothing starts spending without you seeing it, changes wait for your approval rather than happening quietly, and the metrics come in from the ad accounts you connected, so the before and after are sitting in one place when you go looking. None of that decides for you. It just means the comparison still exists in three weeks, when your memory of that good fortnight has smoothed into "it was working".

And when the raise does not work, resist the symmetric overreaction. A campaign that got worse after you doubled it has not been revealed as worthless; it has told you its efficient budget is closer to the old number than the new one. Turn it back down to the level that worked, let it run there, and take the finding for what it is — a ceiling, discovered cheaply, which is genuinely useful information and much better than the alternative of discovering it at four times the spend. That is a different judgment from [deciding whether to kill it altogether](/blog/when-to-kill-a-campaign-and-when-to-wait), and it is worth keeping the two apart in your head.

So: scale at the rate you can measure, not the rate you can afford. And before you touch the number, answer one question — what result would make you turn this back down, and by when do you expect to see it? If you cannot say, you are not scaling a campaign. You are just spending more, and hoping the good fortnight was the real one.