Meta creative testing: one ad set with many ads or a separate ad set per concept?

Plenty of people still put every hook in its own ad set so each gets an equal slice of budget. I get the logic, but on a small budget it usually slows you down. Say you've got five concepts at $20 a day each. Every ad set needs roughly 50 conversions a week to leave the learning phase, and it probably won't hit that, so you end up running five slow tests instead of one fast one. Put all five ads in a single ad set with a clear conversion event and the algorithm shifts spend toward the winner within a few days. You lose the clean per-concept read, but you get an answer much sooner, and the creative is doing most of the targeting anyway. At what budget or conversion volume do you start splitting them out?

Budget logic is fine, but the first few days are where this goes sideways. Meta tends to pour most of the spend into whichever ad picks up cheap early clicks, so your "winner" can just be the one that got lucky on day one. If one ad has 80% of spend and a handful of conversions while the rest barely delivered, you haven't learned which hook works better. You've learned which one started first. So look at spend share per ad before you call it, and give the quieter ones a few days of real impressions.

Agreed on the spend share check. A simple rule that keeps you honest: don't call a winner until every ad has around 1,000 impressions and the top ad's share has held steady for a couple of days. I'd weight impressions over spend, because a cheap ad can rack up impressions while an expensive one only gets a few clicks. Early CTR and hook rate tell you more about the hook than early purchases do, since conversions come in too slowly to judge a creative in its first week. Save conversions for the final check, once there's enough volume behind them.

Hitting 1,000 impressions per ad inside a single ad set is tough, since Meta tends to starve whatever it rates as weaker. So an ad that barely serves isn't necessarily a loser, it's a delivery problem. For judging the creative itself, hook rate (3-second views divided by impressions) is cleaner than CTR, because CTR also picks up how much the offer is pulling.

Hook rate works when the opening is the only thing changing. If all five ads push the same offer and landing page and only the first few seconds differ, 3-second views tell you about the hook and not much else, so I'd rank on that. How much it tells you depends on how different the concepts are, though. If one leads with a discount and another with a testimonial, a hook can look strong while the offer drags down clicks, and hook rate won't show that. Then I'd read hook rate and CTR together and check cost per landing page view before deciding anything.

On starvation, I wouldn't sit around waiting for 1,000 impressions per ad in a single ad set. Ads that barely served just don't have enough data yet, so they shouldn't be called losers, and rotating them out on a fixed schedule only reflects delivery, not performance.

Reading hook rate next to CTR and landing page views makes sense. Before comparing ads, though, make sure they share the same conversion event and attribution window. Meta can credit a view-through conversion to a 3-second hook, which makes a weak offer look like a strong creative.

View-through is the sneaky one. For e-commerce I'd compare ads on click-through purchases only, because a view window can credit a hook with sales that would have happened anyway. Does your reporting break out view conversions separately, or do they get lumped in?

Agreed, and local services work a bit differently. A plumber's conversion is usually a call or a form fill, so a view-through can claim credit for calls that would have come in anyway. I'd compare on clicks only, and check Meta's call numbers against your call tracking, since they often disagree.