Meta creative testing: ABO for a clean read, or CBO and let the algorithm pick?

Say you're testing three new creatives on Meta with around 100 a day. The ABO crowd says give each ad set a fixed share so you can compare them head to head. The CBO crowd says let Meta push money toward the winner and stop babysitting budgets.

For pure creative testing I think ABO still gives the cleaner read. CBO tends to starve two of the three ad sets within a few days, so the verdict ends up resting on whichever one got lucky early. Once you know the winner, move it into a CBO campaign to scale.

Has anyone seen CBO give a fair read on creatives at lower budgets, or does it almost always collapse onto one ad set?

Mostly agree on ABO for the read. But at 100 a day, a third of the budget per ad set is often too little to get a verdict on purchases at all. If your CPA sits around 40 and you want roughly 50 conversions per ad set before trusting it, you're looking at weeks. So at this budget I'd judge creatives on a cheaper signal first, like hook rate or outbound CTR, and only look at purchases once each one has real volume. Otherwise ABO and CBO both just tell you which ad got lucky on a handful of clicks. The starvation you describe is real, but it's partly a sign the test is underfunded, not proof that CBO can't handle creative testing.

Iris, the purchase math is worse than "weeks." Fifty conversions at a 40 CPA is 2,000 spend per ad set, and at about 33 a day that works out to roughly 60 days. At this budget, a purchase verdict is basically off the table.

I wouldn't lean on hook rate as the main filter, though. It's cheap to read, but it rewards whatever stops the scroll, including a loud or misleading first second that loses people right after. Picture a dental clinic creative with strong 3-second views where almost nobody who clicks reaches the booking page. Hook rate alone would call that a winner. Outbound CTR or landing page views sit closer to intent, so use hook rate to cut the obvious losers, then rank the survivors on landing page views.

Agreed on the math. How you get out of it depends on how different the creatives are. If they're three distinct angles, a higher-funnel event like add to cart or landing page views can give you enough volume in a couple of weeks, and you can judge the ads on whether visitors actually stay. If they're three versions of the same pitch, nothing you can measure at this budget will separate them cleanly, so the test is really about which concept to explore next.

One caveat on landing page views: Meta counts them on page load, so a slow page inflates them. Check them against GA4 sessions before you trust the ranking.

I'm not sure add to cart holds up as the deciding metric here. Two weeks at about 33 a day per ad set is roughly 460 spend, and if CPC is around 1 that's maybe 400 landing page sessions. Add to cart on a cold audience often lands somewhere between 3 and 8 percent of sessions, so you'd end up with 10 to 30 adds per ad set. That's too thin to rank three ads, because one bad week can flip the order. Landing page views fill in faster, but I'd still treat that ranking as a rough sort, not a verdict.