CBO vs ABO in Meta Ads: when does each budget work best?

ABO when you test; CBO when you scale — this is the rule I work by. ABO (budget at ad set level) fixes how much each set gets, so every idea receives a comparable volume of impressions. CBO (budget at campaign level) lets Meta redistribute the money in real time and, with proven winning sets, allocates faster than a person can. The threshold around which I move to CBO is 50 conversions a week per campaign. At €10–20 per day CBO usually sends almost everything into one set and the others do not leave the learning phase. Keep 3 to 5 ad sets and do not mix retargeting, lookalike and broad targeting in one CBO campaign.

My name is Hristiyan Panov — a performance marketing consultant with over 5 years’ experience and over €2 000 000 Managed advertising budgets for over 90 companies on Google and Meta. Below, you’ll find out when each option works best and what goes wrong most often.

What does the CBO actually do?

It looks for the cheapest conversion at the moment, regardless of which ad set it is in. If one set shows a better early signal, the money moves there within hours. The allocation is not a one-off decision in the morning — it is recalculated continuously, at auction level, against the expected value of each impression.

This is an advantage in a campaign with uniform audiences and proven creatives. There all the sets compete for the same type of buyer, and the faster movement of money wins.

And it is a drawback in a test — because the algorithm rules on 20 results and kills an idea that needed 200. The early difference between two sets is often noise: a different time of actual start, different inventory, a different density of the auction at the moment of display.

You can tell from the spend by set in the first days. If by the second day one set holds most of the budget, the campaign has already picked a favourite before it had any grounds.

When is ABO the better choice?

• Creativity test. Every set with a fixed budget receives a comparable quantity, and the comparison is fair.
• Audiences of varying value. Retargeting will always appear cheaper than cold audience targeting, and in CBO it will eat into the budget — even though it doesn’t bring in new customers.
• A small budget. At €10–20 a day, the CBO often sends almost everything to a single set, whilst the rest never progress beyond the training phase.
• When you need to ensure spending on a specific audience — for example, a product range designed for a specific purpose.

When does the CBO come out on top?

When you have several proven winning sets and enough data volume. Then manual redistribution between them is slower and less accurate than the algorithm, and human intervention restarts the learning for no reason. “Proven winning” here means a set that has left the learning phase and has kept the cost per result within an acceptable corridor over several consecutive weeks — not a set with one good day.

CBO also wins when results fluctuate strongly from day to day — the algorithm reacts within the day, a person reacts the next. Holidays, paydays, competitive pressure in the auction and the difference between weekday and weekend move the cost per result without anything in the account having changed.

The practical sign that the moment has come: the creatives have already been sifted, the audiences are similar in value, and the only remaining question is how much each should get today. While that question is still being decided by you by hand, you react with a delay of one day.

What is the most common mistake made with CBO?

Mixing incompatible audiences in a single campaign. Retargeting, lookalike audiences and broad targeting within a single CBO mean that almost the entire budget goes towards retargeting, because it appears to be the cheapest option.

The result is a campaign with an excellent return on investment but one that stifles growth — you’re paying to buy again from people who would have bought anyway. This is also why the overall return on investment looks good, whilst turnover remains static.

How do you properly test the transition?

Do not switch an existing campaign. Duplicate the structure in a new campaign using CBO, run it in parallel with the same creatives, and compare the results for at least 14 days, assuming an equal total budget.

Changing the budget type in a live campaign puts it back into learning and distorts the comparison — you see the difference between “before and after a restart”, not between CBO and ABO.

For the comparison to be fair, keep the other variables identical too: the same optimisation event, the same placements, the same audiences and the same attribution window. During the test do not touch budgets, do not add creatives and do not pause sets — every edit puts the campaign in question back into learning and makes the days before it meaningless.

• Exclude the first days from the report — there the result comes from the learning phase, not from the structure.
• Compare cost per result and return, but also look at the number of new customers.
• Keep the audience overlap in mind: the two campaigns bid against each other, which makes the test itself more expensive.

What should you keep an eye on after launching the CBO?

• The distribution of spend by set — if one takes over 80%, the campaign effectively has one set, and the others exist only on paper.
• The cost per result by set separately, not just the overall figure. The overall figure hides the set that eats the budget with cheap but low-value results.
• Is the number of new customers increasing, or is it just the total number of orders that is rising?.
• The learning status by set — a set that has been in learning for weeks is not getting enough volume and only dilutes the campaign.

Look at these rows on a weekly, not a daily, basis. The daily picture under CBO swings by design: shifting budget between sets within the day is the function itself, not a problem to solve.

If the skew towards one set is persistent, the choice is between two moves — you take the non-comparable audience out into a separate campaign, or you limit the spend at ad set level. The second works, but every limit takes away from the freedom for which you launched CBO.

FAQ — Frequently Asked Questions about the CBO and ABO

Is the CBO now compulsory?

No. Meta periodically pushes for campaign-level budgeting, but ABO remains fully available and is the right choice for testing.

Can I set a minimum cost per set in CBO?

Yes, there are limits on the minimum and maximum consumption per ad set. Use them sparingly — the more limits there are, the less point there is in using CBO.

How many hell sets are optimal in CBO?

Between 3 and 5 on a typical budget. Any more than that means that everyone receives too little to move beyond the training phase.

Does learning restart when the CBO budget is changed?

For minor changes — no. The rule of thumb is to increase up to 20% in one go, then wait.

Which is better for a cold audience?

ABO at the start, until it becomes clear which creative works. Then CBO with the survivors.

Conclusion

CBO and ABO are not a matter of better or worse. ABO buys information; CBO buys efficiency. An account that is always on CBO stops learning new things; an account that is always on ABO spends more than it needs to.

If you’re not sure which part of your business is losing money, start by commissioning an independent review.

Request a quote for a digital audit →

Related articles: How much does advertising on Facebook cost? · ROAS and margin thresholds · advertising account audit