CBO vs ABO on TikTok: Test With One, Scale With the Other

The choice between campaign-level and ad-group-level budgets is not a preference. It maps directly onto what you are doing: finding out what works, or pouring money into what already does. Using the wrong one is how buyers convince themselves that nine out of ten creatives failed.

What each one actually does

ABO puts the budget on each ad group. You decide that this group gets $20 a day and that one gets $20 a day, and TikTok honours it regardless of which performs better.

CBO puts one budget on the campaign and lets the algorithm distribute it across ad groups in real time, pushing money toward whatever is winning.

Described that way CBO sounds strictly better. On a testing campaign it is strictly worse, and the reason is worth understanding properly.

Why CBO ruins tests

CBO reallocates budget within the first hours, long before any group has statistically meaningful data. If creative B happens to get three cheap conversions in the first ninety minutes, CBO concludes B is the winner and starves the rest.

You then look at the report and see that creative B got $80 and four conversions while creatives A, C, D and E got $5 each and nothing. The natural conclusion — B works, the others do not — is unfounded. The others never got a chance to be measured.

This gets worse the smaller the budget. On $50 a day across five groups, CBO effectively picks a winner by coin flip and commits.

The structure that works

Testing layer: ABO

• One campaign, 3 to 5 ad groups
• $20 to $30 per day on each group, set at group level
• One variable per group. If you are testing creatives, every group has identical targeting and one different creative. If you are testing audiences, the creative is identical and only targeting differs.
• Let it run 3 to 5 days without touching anything

The one-variable rule is what most buyers skip, and skipping it makes the whole test meaningless. If group A has a different creative and a different audience than group B, a difference in result tells you nothing about which change caused it.

Scaling layer: CBO

• Build a new campaign, do not convert the old one
• Put in only the proven winners, ideally 3 to 5 of them
• Start the campaign budget at roughly the sum of what those groups were already spending
• Raise by 20 to 30% every 2 to 3 days, not in jumps

Here CBO does exactly what it is good at: shifting money between options that are all genuinely viable, reacting faster than you could manually.

When to move a winner up

A group earns its place in the scaling campaign when it has produced at least 25 to 30 conversions at an acceptable cost. Below that, what you are looking at is still partly noise, and you risk scaling an accident.

Two things to keep in mind when you promote it:

• Rebuild, do not convert. Switching an existing campaign from ABO to CBO resets learning across everything inside it. Build the scaling campaign fresh.
• Keep the testing campaign running. Creatives fatigue, and on TikTok they fatigue fast. The test campaign is not a phase you complete, it is a permanent supply line feeding the scaling one.

Where Smart+ fits

Smart+ automates targeting, bidding and placement, and it is genuinely useful for getting a stalled group to deliver or for scaling something already proven. It is not useful for testing, for the same reason CBO is not: you give up the control that makes a comparison valid.

The 2026 version lets you switch individual automation modules on and off, which makes a reasonable middle ground — automate delivery, keep control of creative.

The mistake this all comes down to

Buyers who cannot produce creative volume try to compensate with structure. They test three creatives, get an ambiguous result, and start rearranging budgets, bids and audiences hoping to find performance in the settings.

It is not there. On TikTok, structure protects a good test from being ruined; it does not turn a weak creative into a strong one. The buyers who win are the ones who put twenty angles through a clean ABO test, not the ones who put three through a clever one.

Summary

• ABO to test, CBO to scale, never the reverse
• One variable per ad group or the test proves nothing
• $20-30 per group, 3-5 days, hands off
• Promote at 25-30 conversions, into a new campaign
• Scale in 20-30% steps every few days
• Never stop testing — the supply line matters more than the structure