Is Meta really a performance-driven channel?

Meta Ads are a popular choice for digital marketers. The platform offers incredible reach and quick results and is suitable for B2B and B2C advertising. However, Meta has lost some ground over the last few years regarding conversion tracking and attribution.

So let’s explore exactly how performance-driven Meta Ads are and whether you can — or should — rely on its metrics when evaluating campaign effectiveness and deciding where to invest your advertising capital.

Meta Ads: A quick review

There was a time, only a few years ago, when Meta (then called Facebook) provided the highest return on advertising spend (ROAS) for US retail advertisers. However, the intervening years have been challenging for the social media giant.

Around the beginning of 2021, Meta’s attribution and conversion tracking took a hit. This situation was, uncoincidentally, around the same time as the iOS 14 update.

As Meta transitioned to cookieless attribution, things started to get a little funny. Marketers who had once been convinced that Meta’s ad metrics were rock solid began to notice some inaccuracies. For example, there was a sharp rise in Organic conversions.

Another big issue concerned advertisers were the differences between Meta Ads Manager and Google Analytics conversions. Quite often, Meta showed comparatively better results when analyzing the same data.

So what is going on here? Have privacy changes broken Meta’s ability to provide reliable data for advertising teams?

How does attribution work on Meta?

To delve deeper into what is going on, we must consider how attribution works on Meta. To do that, we need to answer some key questions.

  • How does Meta Attribution actually work?
  • How does Meta register conversions?
  • How does Meta attribute conversions?
  • Is Meta performance-driven?

Meta Ads before and after iOS 14

To get to the bottom of what’s happening, it’s worth looking at Meta before and after the iOS 14 update.

Pre-iOS 14

Everyone used the Facebook pixel before iOS 14. This pixel was excellent at tracking performance data. Some of the things it recorded were:

  • Conversions
  • Landing page views
  • User data that you could leverage for targeting

Crucially, it tracked the full conversion path, and the platform used all that data to help users optimize campaigns and drive better results.

Post-iOS 14

Once the iOS 14 update happened, Meta lost some of its tracking capabilities. Where it was once able to record the entire conversion path and feed that to its machine learning (ML) algorithms, now it was suffering from an absence of data from iOS users.

In essence, Meta went from being able to track the full conversion path to only being able to track the last event in that same path for iOS devices. Losing this information caused big problems for how information is sent to Meta Ad Manager and its ML.

So, what did the platform do to compensate for this missing data? They build a new attribution model.

Meta’s new attribution model

Meta’s new attribution model was built to survive in a post-cookie world. To overcome the loss of information that we mentioned above, the platform turned to data modeling and built a model from the information it had from iOS 14 and other users.

As you might note, this signaled a shift away from the rock-solid numbers we associated with performance-driven channels. Instead, the idea here was that they needed to fill in the gaps because they were working with an incomplete data set (less information from iOS 14 users).

Now, there is nothing inherently wrong with data modeling. The process is used across business and software development to define relationships between objects and gain actionable insights. It works perfectly well in the right situations.

However, an issue arises because Meta has commingled iOS 14 user data with their other user’s data. As such, this has changed the combined output to something more probabilistic. For many advertisers, it’s not the sort of data you would feel confident measuring campaign performance with or directing where you should allocate your marketing budget.

In short, instead of giving you data that accurately measures your attribution and campaigns, Meta now gives you an estimation of the effects of your ads.

There are around 1 billion iPhone users. iOS has over 26% of the global market share. However, that number is higher in certain regions, e.g., Europe (35%). Meta’s model can be somewhat unreliable for advertisers with a large number of iOS 14 (and beyond) users.


The problem with Meta’s attribution model

Meta can no longer construct a precise picture of how your ad campaigns are performing. Without enough data, they have been forced to fill in the gaps using data modeling. However, that’s not the only problem you need to think about.

The data that is sent to your Meta Ads Manager dashboard is probabilistic. That’s not a perfect scenario, and it could lead to over or under-attribution, which could lead to suboptimal choices about campaigns and budgets.

But the situation gets more complicated when we consider that this same incomplete data is being used by Meta’s ML algorithms that are supposed to optimize your ads and campaigns.

The performance-driven precision that people love about digital marketing has been reduced. While data modeling can perform quite well, it’s based on inferences and patterns that aren’t the source of truth that people have grown to know and love.

When measuring campaigns and using that to dictate where you invest your marketing budget, a probably right model is like an educated guess.


Summary: So, is Meta performance-driven?

Having considered the model that Meta uses to drive its reports and optimization, we’d have to say that we can’t describe it as performance-driven. Yes, performance does influence the Meta platform and algorithms, but the accuracy of its modeling is undermined by incomplete information.

Amanda AI uses an impact-based model to help you optimize your ads and budget. Like Meta, our machine analyzes your ad channels and improves performance by feeding the platform’s algorithms with better data. Amanda AI considers the effects of your creatives and continuously optimizes this information to produce a more accurate picture of which ads are driving conversions.

Customer acquisition costs (CAC) have increased significantly over the last few years. Meta’s switch from a data-driven model to a probabilistic model may be costing your ads, leads, and sales. Amanda AI can help bridge the gap and get you back to where you were before the iOS 14 update.


Related content

The best marketing campaigns are a mix of art and science. But how can brands augment their hunches with data? Let’s explore the concept of data-driven marketing campaigns and how they can help you get better results.
The AAARR framework allows brands to break down the buyer’s journey into distinct phases. Because each phase has its own metrics, this process increases the accessibility of the audit process and helps brands remedy any friction that is hurting revenue.
High-performing retargeting strategies of the past are being affected by the removal of third-party cookies. What does that mean for marketing teams, and can they find ways to recover?

In this article

Stay updated

Get on the (email) list!

Stay updated

Join our newsletter