Delete Your Testing Campaign: How We Cut Wasted Ad Spend to Under 5%

Nine times out of ten, when I open a new account and look at how the budget's carved up, there's a campaign sitting there labelled something like "Testing" eating 20 to 30% of the spend.

And nine times out of ten, that campaign is the single biggest leak in the account.

I know that's not the popular view. Ring-fencing a fixed slice of budget for testing is just how most people were taught to do it. It feels disciplined. It feels safe. You're protecting your winners from the chaos of new creative, and you've got a tidy line item that says you're "always testing."

Here's my take though, after watching this play out across a lot of accounts: a dedicated testing campaign is usually a permission slip to waste money. So in most accounts, I'd delete it.

Let me explain what I mean, and then I'll be honest about the cases where I wouldn't.

What that test budget is actually buying you

Walk through the maths on a ring-fenced test campaign and it gets uncomfortable fast.

Say you're spending A$1,000 a day and you've parked 25% of that in a testing campaign. That's A$250 a day, A$7,500 a month, pointed at creative you already suspect is unproven. That's the whole point of it being a test.

Now, some of those test ads are duds. Most of them, if you're honest. So a big chunk of that A$250 a day is going to ads that were never going to make it, in an environment you've specifically built to keep them away from your good ads.

You've created a cost centre and called it discipline.

The bit that gets missed is the opportunity cost. Every dollar stuck in the test campaign is a dollar not sitting behind a proven winner that's already returning above your target. If even half of that A$250 a day is effectively wasted, that's roughly A$3,750 a month you could have pushed into ads you already know work.

To put that in perspective: clawing back even 10 to 15% of total spend from waste and redirecting it to proven creative is the kind of swing that quietly changes an account's whole trajectory over a quarter. It rarely feels dramatic week to week. It compounds.

The thing a separate test campaign can't do

Here's the structural problem, and it's the heart of why I'd rather not run one.

When you isolate testing in its own campaign, you've taken your new ads out of the competition. They're not fighting your incumbents. They're fighting each other, in a little sandbox, for a budget you've pre-decided to spend whether the ads deserve it or not.

That's backwards. The whole reason Meta's algorithm is useful is that it's far better and far faster than any human at working out which ad deserves the next dollar. It's sitting on an enormous pile of data points built precisely to stop you spending money on bad ads. When you ring-fence a test budget, you're overriding the one thing the machine is genuinely good at.

What I'd do instead is let new ads fight the incumbents directly, inside the main prospecting CBO.

You drop the new creative into the same campaign as your proven ads, and you let them compete for the same budget on the same terms. No protected pool. No guaranteed spend.

And here's the part people find counterintuitive: a good new ad won't get much spend on day one, and that's the system working, not failing. The algorithm spends a little, watches, spends a little more, and only really opens the taps once it's convinced the new ad can beat what's already in there. Your duds quietly starve. Your winners earn their way up.

So the "test" still happens. It's just that the algorithm runs it, the budget follows performance instead of a calendar, and your wasted spend on unproven creative drops toward the low single digits instead of sitting at a fixed 25%.

A worked example (invented, but it's the shape I see)

Let me make it concrete with made-up but very typical numbers.

Picture a homewares brand spending A$1,000 a day. Old structure: A$250 in a dedicated test campaign, A$750 across the evergreen winners.

The test campaign was running at maybe a 1.4 blended return. The evergreen sat closer to 2.8. On paper the owner felt fine, because "you have to test to find the next winner." Fair enough as far as it goes.

We collapsed it. The test campaign got deleted, and new creative started going straight into the main prospecting CBO to fight the incumbents. Within a few weeks the spend that used to sit in that 1.4-return sandbox had been reallocated by the algorithm. Most of it landed behind ads already clearing 2.8, and the genuinely good new ads still climbed, just on their own merit rather than on a handout.

The wasted-spend share on unproven creative went from that fixed 25% down to under 5%. Same daily budget. The difference was simply where the money ended up. When the recovered spend lands behind proven winners instead of duds, the blended number moves up without spending a dollar more.

I want to be careful here, because this isn't a guaranteed result and I'm not selling you a number. The point isn't "you'll make exactly this much more." The point is the structure stops bleeding budget into ads you'd already flagged as unproven, and that money has to go somewhere better.

When I'd actually keep a separate test structure

Now the honest part, because if I just told you to nuke every test campaign I'd be doing the same lazy thing I'm criticising. There are real cases where a separate structure still earns its keep.

Brand-new angles or mechanisms. When you're testing something genuinely different from anything in the account, a totally new angle, a new story, a new way of framing the product, it can struggle to ever get spend inside a CBO stacked with established winners. The algorithm keeps feeding the proven ads and the new idea never gets a fair look. For something that different, a small dedicated pocket gives it room to prove itself before you throw it into the main fight.

A genuinely new avatar. Same logic. If you're going after a customer the account has never spoken to, the existing winners aren't a fair benchmark, because they were built for a different person. A separate structure lets you learn on the new audience without that learning getting strangled by ads optimised for the old one.

Tiny accounts. This one's important. Everything I've said assumes the account has enough daily budget for the algorithm to actually make decisions. If you're spending A$50 or A$80 a day, there isn't enough volume for the machine to sort winners from losers inside one competitive campaign. At that scale you sometimes do need to be more deliberate and hand-hold the testing, because the algorithm simply doesn't have the data to do it for you yet.

So the rule isn't "testing campaigns are always wrong." It's narrower than that: a permanent, ring-fenced 20 to 30% test budget, running by default at real scale, is almost always wrong. A small, deliberate test pocket for new angles, new avatars, or a small account is a different thing entirely, and it earns its place.

The mindset shift underneath all this

If there's one idea to take from this, it's that testing isn't a place in your account. It's something that happens continuously, everywhere, when you let proven and unproven ads compete on equal terms.

The dedicated test campaign feels like control. In reality it's the opposite. You've decided in advance to spend a fixed amount on uncertainty, and you've protected your winners from exactly the competition that would prove how good they are.

Letting your new ads fight the incumbents feels riskier. It's actually the safer bet, because the budget only follows performance. The worst ads get a few dollars and die. The best ones earn everything they get.

So I'd look hard at any account carrying a permanent test campaign and ask the blunt question: is this finding us winners faster, or is it just a comfortable place to lose 20% of the budget every month? For most accounts at scale, I think you already know the answer, and the fix is mostly a matter of having the nerve to pull the trigger.

The interesting question isn't whether to test. It's whether you trust the algorithm to run the test better than you can, and where the handful of exceptions to that actually sit for your account.