How to Choose Between CBO and ABO for Your Campaign

Write down whether this campaign's job is to isolate a variable (creative, audience, offer) or to buy the most efficient conversions from a set of already-proven pieces. If you need a fair read, you need ABO or a single ad set. If you need Meta to put spend where it converts, you need CBO — or Advantage+ at a later stage. Put that job in the campaign name so a teammate cannot 'optimise' the setting without noticing the mismatch.

Most CBO-versus-ABO arguments are actually arguments about campaign job, not about a platform setting. A creative test with six videos in one ad set does not care about CBO; there is nothing to reallocate. An audience test with three interest stacks in three ad sets does care, because CBO will starve the slower-starting audience and you will call it a loser. A scaling campaign with two proven geo ad sets and a retargeting ad set cares the other way: you want Meta to shift toward whatever is converting today, which is CBO's job.

If you cannot name the job in one sentence, default to ABO with as few ad sets as possible until the sentence exists. Mixing jobs in one campaign — a test ad set, a scale ad set, and a retargeting ad set under one CBO budget — produces a blended CPA you cannot act on. Split labs from factories. The rest of this tutorial is only useful after that split. One campaign, one job, one budget type — mixing those three is how CBO versus ABO becomes a religious argument instead of a setup choice you can defend.

Choose ABO whenever a fair comparison requires each ad set to actually spend. That is the rule for audience tests, offer tests split by ad set, and any structure where you would not trust a result if one cell got 80% of the budget by day two. If you would not trust the ranking when spend is uneven, you are still in ABO territory, regardless of what a blog says about CBO being 'modern'.

Equal spend is the only way an audience test is real. CBO will give more budget to the ad set that exits learning first, which is often the warmest or broadest audience, not the one you needed to evaluate. You then pause the 'loser' that never got impressions and congratulate the audience that was always going to spend. ABO at the same daily budget per ad set, left alone for 72 hours, is slower and much harder to misread. Write the equal-budget figure on each ad set before launch so nobody 'helps' the lagging cell on day two.

Creative tests usually should not be split by ad set at all. Put the videos in one ABO ad set and let the auction rank ads inside it. The exception is when you are testing something that cannot live in one ad set — different landing pages with different pixels, different countries with different currencies, or a placement split you have a specific reason to force. Even then, equal ABO budgets and a written 72-hour rule still apply. Do not 'help' a lagging ad set with a mid-test budget bump; a change of more than about 20% restarts learning and voids the comparison.

Choose CBO when every ad set in the campaign has already earned its place — proven creative, a viable audience, a landing page you trust — and the remaining question is how to allocate spend among them as conditions change. CBO is a delivery tool, not a discovery tool. If any ad set in the draft is still a hypothesis, it is not ready for CBO and should stay in the lab.

Once two or three ad sets are each capable of producing conversions at or below target CPA, locking equal budgets becomes a drag. One geo will have a cheaper day, one audience will fatigue, one offer will take over on a weekend. CBO moves budget toward the efficient set without you making a 20% edit that resets learning. That is the point of campaign-level budgets, and it is why scaling campaigns are usually CBO (or Advantage+ shopping, which is the more automated cousin). CBO is not a shortcut around proving the cell; it is a way to let proven cells share oxygen.

CBO still needs guardrails. Do not drop an untested audience into a CBO scaler 'to see what happens' — it will either steal spend from winners or starve and look like a failed idea. Do not combine prospecting and retargeting under one CBO budget if you need to cap remarketing; retargeting audiences are smaller and often cheaper, and CBO will over-serve them until frequency climbs past roughly 2.5–3. Keep retargeting in its own campaign with its own cap, typically 10–15% of total spend. If you need remarketing and prospecting in one account, that is two campaigns, not a clever CBO mix.

Each ad set needs roughly 50 optimisation events a week to exit learning. Divide your weekly conversion volume by 50 and that is the maximum number of ad sets this account can honestly support right now. ABO or CBO will not save a structure that has more ad sets than events. Write that number on the campaign brief before you add 'just one more' audience split.

This is the rule that should kill most 'one ad set per angle' accounts. If you see about 50 purchases a week, you can support about one learning ad set, not five. ABO will simply run five underfed tests forever; CBO will pick one ad set and starve the rest, which looks like optimisation and is actually censorship of the test. Consolidate until each living ad set can actually finish learning, then choose ABO or CBO based on job, not based on how many ideas you wish you could run. When volume is seasonal, the honest ad-set count changes; a Q4 structure can be too fat in January.

Advantage+ shopping complicates the picture because it hides ad sets, but the same math applies at campaign level: you still want on the order of 50 purchase events in a week, and you still should not edit budget, creative or audience every other day. If weekly purchase volume is thin, run a simpler manual conversion campaign — one ad set, ABO, broad, six creatives — until the pixel has signal. Switching to CBO or Advantage+ on a starving pixel just automates the noise. Automated budget movement cannot create purchase events the pixel has not seen.

When an ad or ad set wins a fair ABO test — 72 hours, several conversions, 15–20% better cost per result than second place — duplicate the winner into a new CBO campaign. Do not flip the test campaign from ABO to CBO in place, and do not raise the test budgets around the winner. The test campaign stays ABO and untouched so next week's matrix still has a clean room.

Flipping a live campaign from ABO to CBO (or the reverse) is a meaningful structural edit. It reassigns how spend moves and typically restarts learning, which is how a 4x-looking test winner comes back as a noisy 1.8x scaler. Duplicate instead: the lab keeps its ABO structure and equal budgets for the next matrix; the factory is a clean CBO campaign that only contains proven ads. Raise the scaler 20–30% every 48 hours rather than doubling overnight. Treat the duplicate as a new campaign with a new learning clock, not as a continuation you can edit freely on day one.

Bring only the winning creative, or the winning ad set's audience plus that creative, into the CBO campaign. Leave losing hooks in the lab so they cannot steal scaler spend. If you need a second geo or a lookalike, add it to the CBO campaign only after the first ad set is stable — not on day one of the duplicate. Adding ad sets is how CBO scalers quietly turn back into unreadable tests. A scaler that also contains last week's losers is not a scaler; it is a CBO test in disguise and it will spend like one.

Once a week, open the CBO campaign and check whether spend landed where you intended: new versus existing customers, prospecting versus any leftover remarketing, and whether one ad set is taking all the budget because of a tracking quirk rather than true efficiency. Do this on a fixed weekday, not whenever CPA looks spiky on your phone, or you will edit the campaign back into learning.

CBO will sometimes 'optimise' into a hole: an ad set with inflated conversion reporting, a broken landing page that still fires a pixel, or a tiny retargeting pool with high frequency. A weekly allocation review catches that. If one ad set has more than a clearly dominant share of spend and its blended numbers look too good relative to Shopify or your store, stop scaling and reconcile events before you add budget. Platform ROAS is for ranking ads inside a channel; blended return on ad spend is for whether the business can afford the total.

ABO campaigns need a different weekly check: is each ad set still actually spending its budget, and has any ad set gone into learning-limited because volume dropped? If an ABO ad set cannot spend, the audience is too narrow or the bid is too constrained — that is a structure problem, not a reason to switch the whole campaign to CBO mid-week. Fix or pause that ad set, and batch the change so you are not resetting learning every afternoon. Batch the fix: one complete change, then another no-touch window, rather than a drip of daily tweaks.

If the team is still arguing CBO versus ABO, you probably have too many ad sets. Collapse to one prospecting ad set with several ads, keep retargeting separate, and the budget-type decision mostly disappears. Creative volume from a tool like Klip Kanvas belongs inside that ad set as ads, not as new ad sets. Creative belongs in ads; audiences and offers that still need a read belong in ABO tests; proven delivery belongs in CBO.

The highest-leverage simplification in Meta is almost always 'fewer ad sets, more ads.' Six hooks in one ad set teach you something; six ad sets with one hook each teach you which ad set got lucky on delivery. Budget type cannot fix that. Once the structure is one lab campaign (ABO, one ad set) and one scale campaign (CBO, proven ads only), the CBO/ABO choice is obvious and boring, which is the goal. If you need more volume, add ads to the proven ad set before you add ad sets to the campaign itself.

Revisit the choice when the job changes: a new country, a new offer that needs isolation, or a retargeting sequence that must not compete with prospecting. Those are reasons to add an ABO test or a separate CBO campaign, not reasons to turn one campaign into a Swiss army knife. Write the job on the campaign name — TEST ABO hooks, SCALE CBO US, RTG 3-stage — so nobody 'optimises' the wrong one at 6 p.m. on a Friday. A naming scheme that states TEST versus SCALE is cheaper than a Slack argument after someone switches CBO on a lab.