I love Meta earnings calls. As an advertiser, it’s so important to understand the backdrop, which is this: cost of advertising is up.
And Q4 this year is likely to be about 8% higher than the same quarter last year.
A huge amount of your Q4 performance has already been determined in the last 10.5 months of the year. But fear not, there are always things to do in these final weeks, new experiments to run.
Today, I wanted to share 5 surprising things working on Meta right now.
1. Lookalikes are back – and incremental
This one is our big surprise of the last few weeks. I had a call with a Meta rep a while back and they recommended testing lookalikes as a suggested audience. I was a bit shocked when they did, but am always keen to test these things.
We’ve got three accounts now running LALs in some regard. We’re testing a 2% setup and a 10%. All LALs are outperforming broad on CPA.
But perhaps most surprisingly, they’re winning on net new reach too. Almost all the 2% and 10% reach is net new. Really puts the mental into incremental.
2. Carousels are back
Carousels – like lookalikes – used to be dominant back in the pre-2024 days, but quietly they all faded out.
Not now though.
Of the 2.4k ads we’ve launched over the last 30 days, carousels are outperforming on spend per ad more than 2x against video and 3.7x compared to statics.
Comparative volumes are much lower which can lead to novelty bias. But still – start producing carousels asap.
3. More ad sets, more winners?
In the long scale of “trust meta” through to “heavy interventionist” we’re broadly quite Meta-leaning.
But, the last month or so we’ve seen success by breaking out campaigns into more granular ad sets. Whether by promotion, or new creative grouping, we’re seeing that forcing spend through is improving overall account CPA.
4. Bid caps on the up
For a long time, the Bid Cap camp on X ruled supreme. Then they got quieter, and now balanced accounts seem to be in favour.
But we’re seeing an improvement in bid cap campaigns recently.
The best from an efficiency perspective seems to be a true bid cap only approach to a region, though for volume, we’re running hybrids with some volume going into lowest cost/max volume and then accelerated setups for the bid caps.
5. Catalogues driving new (not engaged) customers
And finally, perhaps the big surprise is that catalogue weighting towards new customers is on the up.
Historically, I’ve often seen DPAs/DABAs great at converting engaged audiences, aka acting as bottom of funnel, but rarely great for building demand itself.
But recently, we’re seeing weighting towards new users increase in multiple accounts proving they’re great hooks for acquisition all by themselves.
Note the exception here is for super broad catalogue (10k+ SKUs) where they’ve always worked for new customer acquisition as well.
What are you seeing? Hit reply or leave a comment below.




On the lookalike finding, seed size and overlap are what decide whether it holds. Meta's documented floor is 100 matched people, but a seed in the low hundreds drifts toward broad fast, and a 2 per cent and a 10 per cent off the same thin seed end up the same audience. Worth running the audience overlap report between the LAL ad sets and broad before calling the reach incremental. Reach that is new to an ad set is not always new to the account.