ABO vs CBO vs Advantage+: The Two-Campaign Meta Structure, Worked Through

Most arguments about ABO vs CBO on Meta ads are really arguments about two different jobs. One job is finding out which new creative works. The other is spending as much as possible on the creative you already know works. Put both jobs in one campaign and they fight: the algorithm keeps pouring money into last month's winner, and the new ads you paid creators to make never get enough spend to prove anything either way.

The way out is arithmetic, not opinion. Take one DTC brand, one daily budget, one target CPA, set up as two campaigns from scratch: a testing campaign on ad set budgets (ABO) and a scaling campaign on Advantage+ campaign budget (what Meta used to call CBO). The numbers are illustrative, so swap in your own. Keep the formulas.

ABO, CBO and Advantage+: what each one is in 2026

The names have moved around since Meta renamed CBO as Advantage+ campaign budget.

Two rows in that table deserve a closer look. First, Meta's own help page on Advantage+ campaign budget says it "may not spend your budget equally for each ad set" and might put most of the budget into one ad set if that gets the best overall result. That is exactly what you want when scaling, and exactly what ruins a test. Second, an Advantage+ sales campaign is not a separate thing from CBO. Meta's developer documentation describes it as the updated, streamlined version of the old Advantage+ shopping campaign, and a campaign only counts as Advantage+ when it runs Advantage+ campaign budget, Advantage+ audience and Advantage+ placements together. In other words, Advantage+ is CBO with the audience and placement decisions automated too.

That gives the short answer to the ABO vs CBO question: use ABO to test and CBO (in practice, an Advantage+ sales campaign) to scale, and keep them in separate campaigns so one job can't starve the other.

The worked example: the brand and the numbers

Here is the account we'll build. It's a made-up DTC skincare brand, but the shape is common.

The 80/20 split is a starting assumption, not a law. It's a sensible default when an account has a handful of proven ads and a steady supply of new creative. If you have no proven ads yet, testing gets a far bigger share until something works. If creative supply is thin, testing gets less, because a testing budget with nothing new to spend on just re-tests old ads.

Step 1: size the scaling campaign against the learning phase

Meta says an ad set exits the learning phase once it can deliver stably, which "usually occurs after about 50 results in the week after the ad set's last significant edit". That one sentence decides how many ad sets your scaling campaign can afford.

For our brand: the scaling budget is 80% of $600, so $480 a day. That's $3,360 a week. At a $40 CPA, that buys about 84 purchases a week. Divide by 50 and you get 1.68. So the scaling campaign can support one ad set that reliably gets out of learning. Split the same $480 across two ad sets and each one expects about 42 purchases a week, under the 50 mark, and both are likely to sit in learning or show as learning limited.

I'd bet this explains most of the accounts with five or six scaling ad sets that feel jumpy week to week. There's enough money. It's just spread across more ad sets than the conversion volume can feed. One broad ad set inside an Advantage+ sales campaign, holding the proven ads, is the right build at this budget.

Flip the formula round and you get a useful threshold for any account. To feed one ad set, the scaling campaign needs a daily budget of at least 50 × CPA ÷ 7, or roughly 7.1 times your target CPA. At $40 that's about $286 a day. Below that, no amount of restructuring gets a scaling ad set out of learning, and the honest move is to accept a longer learning period or optimise for an event that happens more often.

Step 2: build the testing campaign on ABO

The testing campaign gets the other 20%: $120 a day. With three concepts a week, that's one ad set per concept at $40 a day each, and each ad set holds that concept's three hook variations.

The settings that make or break this campaign:

• Ad set budgets, not campaign budget. The whole point is that Concept C gets its $40 even if Concept A starts strong on day one.
• Budget sharing off. Meta's ad set budget sharing lets each ad set share up to 20% of its daily budget with the others. That's a useful nudge for proven ad sets and a leak in a test, where a fair read depends on spend staying put.
• The same audience and optimisation event in every ad set: broad targeting, purchase optimisation, nothing else different. If the ad sets differ in anything except the creative, you're testing two things at once.

Notice what the maths says about learning here. At $280 a week and a $40 CPA, each testing ad set expects about 7 purchases. None of them will exit learning, and that's fine. You aren't trying to make a test ad set stable. You're trying to get enough spend on each concept to judge it against your kill and keep rules, which is a separate calculation covered in our guide to how much to spend per creative to get a real read.

Step 3: graduate winners on a schedule

A concept that clears your keep rule moves from testing into scaling. How you move it matters, because Meta lists "adding a new ad to your ad set" and "any change to ad creative" among the significant edits that send an ad set back into learning.

That gives you a simple rule: graduate once a week, in one batch, rather than dripping winners in as they appear. Every graduation resets the scaling ad set's learning clock, so three separate additions in a week cost you three resets. One batch on a Monday costs one.

When you graduate an ad, consider using the existing post (the same post ID) rather than uploading the video again, so the likes and comments it earned in testing travel with it into scaling. And pause the losing ads in the scaling ad set at the same Monday edit, so the ad set isn't carrying dead weight.

Step 4: scale the budget without breaking learning

Budget changes can count as significant edits too, and Meta's significant edits page is clear that size matters. Its own example: going from $100 to $101 is unlikely to restart learning, while going from $100 to $1,000 may. Meta doesn't publish an exact threshold, so the safe habit is small steps, made at the same weekly edit as your graduations, so the account takes one hit of instability a week rather than several.

Here's how the structure stretches as our brand grows, keeping the 80/20 split and the $40 CPA.

Two things fall out of that table. At $300 a day the scaling campaign can't feed even one ad set to 50 purchases a week, so at that level you'd either accept a slower learning period or test fewer concepts and push more into scaling. And at $1,500 a day the budget could feed four scaling ad sets, but it doesn't have to. Extra ad sets only earn their place when they hold something genuinely different, such as a separate country or a product line with its own CPA. Splitting the same broad audience four ways just to use the room adds nothing.

What about retargeting?

Older setups ran a third campaign just for retargeting past visitors and buyers. With an Advantage+ sales campaign, that's often no longer needed. A broad or Advantage+ audience includes people who already know the brand unless you exclude them, and a separate retargeting campaign aimed at the same people splits delivery across them, the kind of audience fragmentation Meta recommends combining away. A dedicated retargeting campaign earns its place when it carries something different, such as a catalogue ad or an offer only past visitors see. If it's running the same ads to the same warm people, fold it back in.

Where this structure goes wrong

• Testing inside the scaling campaign. Dropping new ads into the Advantage+ sales campaign looks efficient. In practice the proven ads keep most of the spend and the new ones get too little to judge, which is the exact problem the second campaign exists to solve.
• Too many scaling ad sets. If the formula says one, run one. Each extra ad set divides the purchases that get the others out of learning.
• Graduating daily. Every addition is a reset. Batch them.
• A testing budget with nothing to test. The 20% only works if there are new concepts every week. When the creative pipeline dries up, the testing campaign quietly turns into a second, worse scaling campaign. Keeping that pipeline full is the work we do for brands on done-for-you UGC, and it's why we plan creative volume around the account's spend, as covered in how many creatives to test per week.

If a graduated ad starts to slide in the scaling campaign, the 1.5x CPA kill rule tells you when to pull it, and the slot it leaves is what next week's tests are for.

Frequently asked questions

Is ABO or CBO better for Meta ads?

Neither is better overall; they do different jobs. ABO (ad set budgets) is better for testing, because each ad set spends its own budget and every new concept gets a fair read. CBO, now called Advantage+ campaign budget, is better for scaling, because Meta moves the budget in real time towards the ad sets with the best opportunities. Most DTC accounts run both, in separate campaigns.

Is Advantage+ the same as CBO?

Not quite. An Advantage+ sales campaign always uses Advantage+ campaign budget (CBO), but it also automates the audience and placements. Meta only treats a campaign as Advantage+ when all three are switched on together. So every Advantage+ sales campaign uses CBO, but a CBO campaign with manual targeting isn't Advantage+.

How many ad sets should a scaling campaign have?

Only as many as your conversion volume can feed. Meta says ad sets usually exit the learning phase after about 50 results in a week, so divide your weekly scaling budget by your CPA, then divide by 50. A $480 a day scaling budget at a $40 CPA gives about 84 purchases a week, which supports one ad set.

If you'd rather not run the testing side yourself, we write, produce and ship the weekly concepts this structure needs. Tell us about your account.

Researched and drafted with AI assistance, then edited and fact-checked by Chris Bristow.