+1 vote
by (15k points)

1 Answer

0 votes

I have absolutely seen performance improvements when campaigns with different objectives were siloed into different advertising accounts. Facebook doesn't provide any transparency on this, but my suspicion is that Facebook distributes impressions based on account-wide as well as campaign-wide performance metrics, meaning that it'll look at the performance across an entire account and not just that specific campaign when deciding whether to allocate impressions to an ad or not. More background on how Facebook's algorithm works in this answer: How does Facebook’s advertising targeting algorithm work?

The practical impact of this is that losers in an ad account will cap the performance of the account's winners. If you're running a lot of new ads or testing new formats, or if you're running MAI campaigns in the same ad account as VO and AEO campaigns, then the overall mean performance of that ad account will be used in determining the expected value of any given ad being served in an impression and the "winners" (outperforming ads) will be stifled by the global ad account performance mean.

As an example: suppose an advertiser is constantly testing new creatives with MAI campaigns, most of which end up underperforming. If the advertiser has VO campaigns in the same ad account, the low CTR of the test MAI campaigns could end up impacting delivery for the VO campaigns because Facebook sees the low CTR at the ad account level as reducing the expected value of the VO campaigns. If the VO campaign was split out into its own ad account, the MAI campaigns wouldn't affect it and it might potentially see increased delivery.

Obviously CTR is just one component of the overall expected value calculation that Facebook performs when serving ads: targeting, bids, min-ROAS targets, etc. will vary from campaign to campaign and potentially muddy Facebook's interpretation of an account's performance. Advertisers split these activities out into separate ad accounts to avoid that.

by (15k points)