Meta testing at small budgets: ABO per concept or let CBO pick the winner?

At around $30 a day with three new concepts, the usual advice is to run CBO and let the algorithm pick a winner. The problem is that at that budget CBO tends to pour most of the spend into whichever ad set got a lucky early start, so the other two never get a fair read. ABO with a fixed budget per concept gives each idea a real shot, which is the whole point of a test. I'd run ABO during the test phase, then move the winner into a CBO campaign once each concept has a handful of conversions. Creative does most of the targeting work anyway, so I'd rather spend the budget finding out which hook lands than trust the algorithm to guess. Has anyone actually seen CBO split a small budget evenly across concepts?

ABO fixes the split, but at $10 a day per concept you can end up with none of them ever leaving learning. Meta wants roughly 50 optimization events per ad set per week, and a $10 budget usually can't get there on purchases. If a concept can't get close to that, test on landing page views first and move to purchases once you've found a winner.

Landing page views help, but I don't think they fix the real problem. At $10 a day per concept, cost per landing page view can swing a lot with audience and placement, so a cheap-looking LPV winner may just be the concept that got lucky on a cheap placement.

For a creative test, the cleaner first read is the hook itself: 3-second video views as a share of impressions, then hold rate at around 15 seconds. Those numbers show up within a day or two, long before you have any conversion volume, and they measure what you're actually testing, which is whether the first two seconds stop the scroll.

Once one concept clearly holds attention better, that's when it makes sense to put the conversion learning budget behind it. Before that, you're asking a $10 ad set to answer a question it doesn't have enough data for. Are the three concepts far apart on hold rate yet, or still too close to call?

Hook rate has the same placement problem you raised for LPV. Reels tend to get cheaper impressions and more 3-second plays than Feed or Stories, so a concept can look stronger just because of where it was delivered. Define hold rate as 15-second views divided by 3-second views too, otherwise the numbers won't line up across concepts.

Agreed on the placement issue. My rule of thumb is that a hook only counts as a winner if it leads on hold rate in the placements where all three concepts actually delivered. Open the placement breakdown and compare Feed and Reels separately. If concept A only pulls ahead because Reels gave it cheaper, longer plays, call it a tie and let the test run longer.

At $10 a day, delivery can skew hard toward one placement, and that shifts every average. So the overall number mostly tells you where each ad was shown, not how the hook performed.

Splitting the comparison by placement and calling close results a tie doesn't hold up at $10 a day per concept. The Feed and Reels cells each get only a few hundred impressions, so their hold rates are mostly noise. You'd be trading one skew for a sample-size problem, and "let it run longer" can mean the test never resolves.

A cleaner option is to lock all three concepts to the same placement for the test, Reels only or Feed only, so they compete on identical inventory from the start. Then the hook comparison reflects the hook, not delivery. You give up some reach, but you get a read you can act on. Once one concept clearly leads, widen the placements again and check that it still holds.