Letsgooo!

A discord would be fantastic

You are now only testing 2 creatives per ad set. That in itself is okay, but I would personally always recommend at least 3 ads per ad set, especially with CBO. Why? - CBO tends to push all the budget to one ad too quickly - With 2 ads, there is little room to correct that or recognise patterns - With 3 ads per angle, you can see quicker whether the problem is with the angle, or with the creative itself - It also helps to test with multiple formats: UGC, Problem/Solution and then as a third a Before/After or Founder-style ad If you then see after 2-3 days that all ads within a certain desire are performing poorly, you are much more certain that it is at the angle, not one underperforming ad.

I would be a little more careful about this anyway. The fact that Facebook sends the most budget to one ad does not necessarily mean that those other 9 ads are worse , just that the algorithm got an early signal of performance faster with that one. But that doesn't mean it's the right “winner” at ROAS level. Sometimes an ad just gets clicks faster or a lower CPM, but later turns out to have worse conversion quality. Also Meta optimises for engagement or low-cost events (such as clicks or ATCs), not necessarily for ROAS unless you have a lot of conversion data. An ad with better creatives or targeting may initially under-optimise because the algorithm simply does not yet have enough data points to accurately optimise. In short: if you're on $100 spend and the “favourite” performs poorly, that's definitely a sign, but I'd dive into the adset breakdown first. Sometimes it pays to convert the CBO to ABO and still manually give some spend to those other creatives. Especially if you do a larger test with 10 creatives. Otherwise, you stop the test based on half information.