Meta Ads has had a rough few years in the press. iOS privacy changes, rising CPMs, attribution that no longer works the way it used to. The narrative became "Meta Ads is broken." That narrative is wrong, but the channel has genuinely changed, and companies running it the same way they did in 2020 are getting worse results for reasons they do not fully understand. Here is how I think about it now.
The core problem is signal loss. Before ATT, Meta could see a large fraction of the conversion events that happened after a click or view. It used that data to train its bidding algorithm. After ATT, reported conversions in Ads Manager dropped — not because fewer people converted, but because Meta could no longer see all of them.
The advertisers who adapted early did two things: they set up the Conversions API server-side, and they stopped over-relying on reported ROAS in Ads Manager as a measure of campaign health. The CAPI setup sends conversion data directly from your server to Meta rather than relying on browser events. It does not fully recover the lost signal, but it closes a meaningful part of the gap.
If you are not running CAPI alongside your pixel and you are spending meaningful budget on Meta, you are making decisions based on an incomplete picture of your own performance.
For DTC, Meta remains one of the most powerful acquisition channels available, primarily because of creative. The algorithm is good at finding buyers — the limiting factor is almost always the quality and variety of creative. If you are only testing two or three creatives per month, you are leaving most of the channel's potential on the table. The brands winning on Meta right now are running aggressive creative testing frameworks: multiple concepts, multiple formats, clear creative learning loops where winning ideas get iterated and losing ones get killed fast.
For B2B, the story is different. Meta can work well for lead generation and top-of-funnel awareness, but the targeting has weakened since iOS changes removed a lot of behavioural signal. Broad targeting with strong creative and a well-defined ICP message tends to outperform narrow interest-based targeting for most B2B accounts now. The instinct to be very precise with targeting often backfires because you give the algorithm less room to find relevant people.
Meta's own guidance is toward simpler structures: fewer campaigns, broader ad sets, and more creative variation within those ad sets. Advantage+ Shopping Campaigns for DTC ecommerce have worked well for many advertisers. For lead gen, Advantage+ Audiences are worth testing. The general principle is to give the algorithm room rather than fighting it with over-segmentation.
This does not mean blind trust. You still need to monitor where budget is going, watch for audience overlap, and build a creative pipeline that keeps the algorithm fed with fresh material. Simplifying structure is not the same as reducing scrutiny.
Reported ROAS in Meta Ads Manager is not a reliable standalone metric. It does not account for view-through attribution inflating numbers, it does not tell you what is incremental, and it competes with other channels for credit in ways that are opaque. The advertisers I see making the best decisions use a blended measure: reported numbers in-platform as a directional signal, combined with holdout tests or geo experiments to understand true incrementality, and overall business metrics (revenue, new customer acquisition) as the ground truth.
If you are making budget decisions purely on Meta-reported ROAS, you are probably over-investing in some areas and under-investing in others.
If you want to work through any of this for your specific account, get in touch.