How to test creative without burning your whole budget on losers

Here’s a number that should take some pressure off. The very best marketers and media buyers in the world only have something like a 15 to 20% success rate, and that includes me. If I make ten new ads, maybe one or two of them are going to do really well.

So losers aren’t a sign you’re bad at this. They’re the cost of finding out. The question is how much you pay per answer.

That’s what this is about. Not avoiding losers (you can’t), but making each test cheap, clear and actually worth the money.

Step one: test concepts, not cosmetics

The fastest way to burn money when you test Facebook ad creative is to test things that can’t teach you anything.

Blue background versus green. Man in the photo versus woman. Twenty images of the same idea. Those feel like tests, but they’re mostly the same ad in a different outfit, aimed at the same person. You can spend a month on that and come out knowing nothing new about your market.

A concept is different. It’s a picture of one type of person at one point in time: what they believe, what they’re living through, and a guess at what message will move them. Different concepts reach different people, and that’s where the real learning is. I’ve written more about the difference between a concept and a hook if the line between them is blurry for you.

The test I use before anything goes live: does this ad require a different type of buyer to respond, or just a different visual? If it’s a different buyer, it’s a new concept and it’s worth testing. If the same buyer would respond to either, it’s a variation, and it doesn’t deserve its own budget yet.

Step two: you need fewer ads than you think

I’m not of the belief that you need to make hundreds and hundreds of ads. I think that causes more harm than good, especially at lower spend. Even at $1,000 to $3,000 a day, I might make 20 or 30 ads, and they’re more targeted and precise.

For a first round, I’ll start with maybe five to ten of the best ads I can write, spread across three to five really strong concepts. That’s enough to learn which ideas the market cares about without spreading the budget so thin that nothing gets enough spend to tell you anything.

And each ad doesn’t have to be expensive to make. One solid body of copy can carry several different openings (rewriting the first two lines is the cheapest new ad there is), and a simple image with a line of text costs you minutes, not a photoshoot.

Step three: match the structure to your budget

This is where people copy a creative test setup built for a much bigger Facebook ad account and wonder why nothing gets out of the gate.

My general advice, kept simple: if you’re spending under about $200 a day, run one campaign with the budget set at the campaign level (CBO), a single ad set, and your ads inside it, maybe three or four per concept. If you’re at $200 a day or more, split it out so each concept gets its own ad set with its own budget, four to six ads each.

The logic is just math. At low spend, splitting into lots of ad sets gives each one too little money to learn anything. At higher spend, separating concepts lets you see which idea is working without one ad set swallowing the budget.

A couple of years ago I also ran cheap awareness campaigns to pre-test elements before putting them in the sales campaign. I don’t do that anymore. I start with the real campaign, optimized for the result I actually want.

Step four: give every ad a fair look

When you test Facebook ad creative, under two full days of data is too early to judge anything. The system is still figuring out who to show your ads to. At that stage you’re looking for signal, not winners. More on that in how long to leave an ad set running before you judge it.

After the first few days, I go back and study the results. If a concept is working, I make more ads inside it. If things are struggling, I check the technical side first, then look at whether the concepts are out of step with what this audience expects, make a new batch and run it again.

For what it’s worth, when I ran one big campaign with ten to fifteen ads in it, what usually happened was that three to five of them did really well and the rest either fizzled out in spend or I paused them. That’s the other reason not to panic early. In that setup, a lot of the weaker ads never ate much of the budget in the first place.

Step five: only cut what’s genuinely broken

This is the step that saves or burns the most money, and it’s the opposite of what most people expect.

The instinct is to kill anything above target the moment you see it. But what I’ve found is that Meta sequences ads. People see several before they buy, and the credit usually lands on one of them, often the first or last. An ad that looks expensive can be doing supporting work for the one that looks cheap.

So I have a watching zone. Say your target is $30 per purchase and an ad is sitting at $40 to $50. I pay attention, but I don’t turn it off right away, because it could be ordinary fluctuation or it could be playing a supporting role. I look for signs of life: comments, likes, a decent click-through rate. If there aren’t many, I give it another day or two.

My actual kill rule when I test Facebook ad creative is strict. The ad’s CPA has to be three to four times target, and its engagement (CTR and cost per click) has to be weak too. That’s a real loser. Everything else is either working or still telling you something.

Anyways. Killing too early throws away what you already paid to learn.

What to do this week

List every ad you’re currently testing and write one sentence next to each describing the person it’s for. Group the ones with the same sentence together. However many groups you end up with is how many concepts you’re really testing. If it’s fewer than three, stop making more variations of those same ideas and put your next testing dollars into two ads aimed at a type of buyer none of your current ads speak to.

How much should I spend testing Facebook ad creative?

How many ads should I test at once?

When should I turn off a losing ad?