ABO vs CBO: Pick by Whether You Can Fund the Test

Under ABO, you decide: each ad set gets the budget you assign and Meta spends it on that ad set only. Under CBO, Meta distributes one campaign budget in real time to the ad sets with the best opportunities. That difference decides which job each setting suits.

A test compares creatives, so it needs equal, forced spend. CBO is built to do the opposite: find a leader and fund it. That makes it a fit for ads that have already won, where handing allocation to Meta saves daily management.

There are now three options, not two. As of October 2026, Ads Manager offers campaign budget, ad set budgets, and ad set budget sharing, where you keep a budget per ad set and Meta may move up to 20% of it to other ad sets in the campaign (Meta Business Help Center). Check which one is selected before you publish, because a new campaign may come preselected with campaign budget.

Each ad set needs a daily budget of cost per purchase × 50 ÷ 7. This applies the learning phase target of about 50 optimization events in seven days from Meta's learning phase guidance, as of October 2026. The formula is our model, not a minimum Meta enforces. Our learning phase guide works the same sum with more examples.

We calculated the table below from that formula. Four ad sets is an example, and figures are rounded from unrounded values, so four ad sets at $214 appear as $857.

Run this check before you launch:

1. Get your cost per purchase. Use the cost per purchase on your prospecting campaigns (campaigns aimed at people who have not bought from you) over the last 30 days in Ads Manager.
2. Compute the required daily budget per ad set: cost per purchase × 50 ÷ 7. This step is arithmetic only, with no pass or fail.
3. Compute expected weekly purchases: planned daily budget per ad set × 7 ÷ cost per purchase. Pass: 50 or more. Fail: fewer than 50. Within a fail, 40 to 49 is close, 25 to 39 is short, and under 25 is structurally short.
4. Compute how many ad sets you can fund: total daily test budget ÷ the step 2 figure, rounded down. Pass: the result is at least the number of ad sets you planned. Fail: it is lower.

Our budget calculator does the same sum from concepts, variants, duration, assumed cost per purchase and spend.

A common starting budget of 2 to 3 times cost per purchase per ad set per day (our observation, not a measured rate) funds only 28% to 42% of the target. The sum is easy to run and the shortfall is large.

At $30 per purchase, 2 times is $60 a day, which buys 14 purchases a week ($60 × 7 ÷ $30), or 28% of 50. Three times is $90 a day, which buys 21 purchases, or 42%. Both land in our "structurally short" band.

Our view is that an ad set with 14 purchases rarely separates two creatives of similar quality, because a gap of three or four purchases can be luck. The same weakness shows up under CBO. Meta moves money toward early leaders, and an early lead built on a handful of purchases can starve the other ad sets before they have shown anything. A test fails the same way under either setting.

Run fewer ad sets. Divide your daily test budget by the required daily budget per ad set and round down. With $300 a day and a $30 cost per purchase, $300 ÷ $214.29 is 1.4, so one ad set is properly funded. With $450 a day, $450 ÷ $214.29 is 2.1, so two are.

In this test design, one creative sits in one ad set, so the number of ad sets you can fund is the number of creatives you can compare at once. Four options follow:

• Test in rounds. Fund the number of ad sets you can afford, such as two. Run each round for a week, keep the winner as the control and replace the loser with a new challenger.
• Put two or three ads in one funded ad set. When only one ad set is fundable, this is the fallback. It relies on Meta's delivery split between the ads, and Meta may favor one ad early, so it is a weaker test than separate funded ad sets.
• Move to an earlier event. Optimizing for add to cart, which happens more often than purchase, gives each ad set more events for the same money. The cost is that the result predicts sales less well, so confirm any winner on purchase.
• Test bigger differences. Two creatives with different angles separate at low volume far better than two with different first lines. A creative brief built around one test keeps each ad set to one variable.

Pick ad set budgets when you are testing with funded ad sets, campaign budget when you are scaling winners, and budget sharing only when you want flexibility more than a clean read.

When a creative wins, move it into a separate campaign-budget campaign, and keep testing in its own ABO campaign so a scaling decision never changes a test.

Take the top ranked tests, up to the number of ad sets your budget check says you can fund, and brief each around one variable. Our creative intelligence platform combines your ad-performance data, competitor activity, customer signals and past ad learnings into ranked weekly creative test plans and execution-ready creative briefs for Meta.

The budget check tells you how many ad sets you can fund. A ranked plan tells you which tests deserve them. See Deepsolv pricing.