The 3-2-2 Meta Campaign, Decoded: Testing Creatives, Offers, and Landing Pages in One CBO

Nine times out of ten, when I open a six-figure-a-month account that's stuck, the problem isn't the bid strategy or some exotic setting. It's that the account is testing one thing at a time, slowly, in a structure that fights the algorithm instead of feeding it.

The 3-2-2 is the structure I'd use to fix that. It's a clean way to test creatives properly inside a single CBO campaign. And then there's a layer most guides leave out, which is the part I actually get excited about: testing landing pages and offers inside that same campaign, not as a separate project six weeks later.

Let me decode the whole thing, then show you the extra layer.

Quick honesty first, because it matters. No structure saves bad ads. If your offer is weak, your creative is weak, and your landing page is weak, this won't rescue you. Account structure is a 10 to 20 per cent lift on top of good marketing, not a substitute for it. If you implement all of this and still can't scale, that's your signal the problem is the creative, the offer, or the page, not the campaign.

What the numbers actually mean

The 3-2-2 describes what goes inside one ad, built across a handful of variations.

• 3 creatives. Three different visuals or video concepts for the thing you're testing. Not three tiny edits of the same clip, three genuine swings at the idea.
• 2 primary texts. Two versions of the body copy. One is your best-performing copy for that product, the other is a fresh challenger.
• 2 headlines. Two versions of the headline under the creative.

So a single concept becomes three creatives, each paired with two copy options and two headlines, and you let the algorithm sort out which combination it wants to spend on. Primary text and headline matter more than people think, by the way. I've seen high-spending ads where the copy was clearly doing real work, so don't treat those twos as an afterthought.

One firm opinion here. Build this with manual single image or video ads, not the flexible-ad format. Flexible ads bundle your variations into a carousel and then hide which specific creative actually got the spend, which defeats the entire purpose of a creative test. You ran the test and Meta won't tell you what won. Build it manually so you can read the data. The half hour you save with flexible ads costs you the one thing you set the test up to learn.

The campaign frame: one CBO, broad, kept simple

Before the ad, the container. Keep it boring on purpose.

Run it as a single sales campaign with CBO turned on, so the budget flows to whatever's working instead of you guessing per ad set. If you're spending under roughly A$1,000 a day, you should have one, maybe two campaigns total. Spreading a small budget across many campaigns just starves the algorithm of the data it needs.

On targeting, go broad. I'd reframe broad as "unrestricted" rather than "everyone", because that's what it really is: you're letting Meta's algorithm find the buyer instead of boxing it in with interests it's going to spend outside of anyway. Set your location, leave the age wide unless the product genuinely demands otherwise, and the only exclusion I'd bother with is your existing purchaser list so you're prospecting, not paying to reach people who already bought.

On the bid strategy, highest volume or value is fine. Cost caps and bid caps are a whole rabbit hole, and they won't save weak ads either. I'd rather you spend the day making a better ad than the day fiddling with caps.

The three ad sets that make it run

Inside that one campaign, I'd run three types of ad set. Keep the settings identical across all three, because the moment you change exclusions or attribution between them, you're not running a fair test anymore.

1. The champions ad set. Your proven winners live here, dropped in as existing posts so they keep their social proof. It stays on. If you've got no winners yet, you don't need it yet, it fills up as you find them.

2. The creative testing ad set. This is the workhorse, where the 3-2-2 actually gets built. Every new concept goes in as a fresh ad set, named clearly so you can read reports later. This is the one you'll use most.

3. The page and offer testing ad set. This is the layer most setups never reach, and where the next section lives.

The layer most guides skip: pages and offers in the same campaign

Here's the part I think is underrated. You don't test creatives in one place and landing pages in some separate project months later. You test them in the same campaign, at the same time.

The mechanic is simple. When you build your 3-2-2, you upload the three creatives with their copy and headlines. Then you duplicate all of them inside the same ad set and change one thing on the duplicates: the destination URL. Now the same creatives are being tested against two different landing pages at once.

Why this is worth doing:

• It forces you to actually test pages, which most brands keep meaning to do and never get to.
• Meta crawls your landing pages and factors them in, so it will quietly push budget toward the creative-and-page combination that converts best. A great ad pointed at a congruent page will usually out-earn the same ad pointed at a generic one.
• You learn how specific ads behave with specific pages, which is the kind of thing brands burn months split-testing the slow way.

A worked example with invented numbers. Say a homewares brand sitting around A$2,000 a day in spend keeps getting cheap clicks on a particular video but soft conversions. Instead of killing it, you duplicate it in the page-testing ad set and point the copy at a proper advertorial instead of the bare product page. Same ad, different page. If it suddenly holds your target return, that lift was pure landing page, not media buying. That's the gap this layer exposes.

And the same ad set is where I'd test offers. Build an ad for a new offer, point it at a new offer page, and drop it in. If the offer's no good, it simply won't get spend, so you're not risking much to find out. The beauty of doing creative, page, and offer testing inside one CBO is that Meta only feeds what's actually working. Your weak new ideas quietly get starved instead of bleeding budget, and you keep your ego out of it.

Flexible vs manual, and when to force spend

Two guardrails people always ask about.

On flexible versus manual ads: I've said it, but it's worth repeating because it's the most common own goal. Manual, every time, for anything you want to read data from. Flexible is convenient right up until you need to know what won, and then it's useless.

On minimum budgets: CBO will sometimes refuse to spend on your shiny new ad, and that's usually Meta telling you the ad isn't as good as you hoped. Most of the time I'd let that happen, because forcing spend onto ads the algorithm doesn't rate rarely scales anything. But CBO does let you set a minimum budget per ad set if you genuinely need a test to get airtime, so you get the best of both worlds without splitting into separate ABO campaigns. I'd reach for that sparingly. What I'd avoid at low spend is a daily maximum, since capping your winners is the opposite of what you want.

How much to test, and when to kill

A simple sum keeps you from overwhelming the campaign. Take your daily budget and divide it by your target cost per acquisition. That's roughly how many tests the campaign can carry without spreading itself too thin. A$100 a day at a A$20 target is about five tests, not thirty. Above roughly A$1,000 a day you can test more freely.

In practice that's three to five new concepts a week, each one being a full 3-2-2 with its page variations, so a single concept is already several ads.

On killing: give a new ad set about seven days. If it hasn't earned meaningful spend and it isn't hitting your KPI in that window, turn it off. If you're the patient type and you've genuinely seen things take two weeks to wake up, then judge it at fourteen days. The rule underneath both: if it's working, leave it alone, and if it's clearly not, stop feeding it.

Try it on one concept

If you want to feel why this works, don't rebuild your whole account this week. Take a single concept you were about to test the old way, build it as a proper 3-2-2, then add the second layer: duplicate it against one alternative landing page inside the same CBO. Let it run a week and read which creative-and-page pairing Meta chose to spend on.

That one test usually teaches you more about your funnel than a month of changing ads in isolation. Run it, see what the data says, and let that decide what you build next.