Meta's New Incremental Attribution Optimization – Friend or Foe?

Meta's latest rollout, optimization with incremental attribution, could be one of the most powerful unlocks in performance marketing right now. But there’s little clarity out there on what it actually is (and what it’s not), and it’s causing confusion.

Let’s first break down incremental attribution. What are the implementation specifics and what’s proprietary to Meta?

Incrementality is NOT just an attribution tool.

Separately, inside Ads Manager, you can now view incremental performance, i.e., attribution results. Meta’s new feature is an optimization lever. 

Here’s what’s happening:

Meta’s ad auction now bids on your behalf based on the incremental value of each impression.

The system calculates the delta between the probability of a user converting with ad exposure vs. without ad exposure. Then, the auction decides whether to bid (and how much) based on that incremental lift, not just the raw conversion probability.

Now, contrast that with traditional, rules-based attribution:

Meta bids based on overall conversion probability within your set attribution window (1-day-click, 7-day-click, etc.), without factoring in incrementality.

The statistical backbone? This is known as uplift modeling or incremental modeling. These models have been around for years – telecoms and subscription companies have used them to answer questions like: Should I contact my active subscriber if the touchpoint will be truly incremental?

Imagine there are 2 users:

Under the standard attribution model, Meta will bid more for user A than B, everything else being equal, because user A has higher chances of conversion, hence higher estimated action rate.

‍

Under incremental attribution optimization, however, Meta is incentivized to bid more for the user B over user A, as user B has more incremental conversion likelihood, if treated. 

The catch? Clean exposure data (meaning, who saw the ad and how did they interact) is critical—and that’s been a major blocker for wide adoption.

But walled gardens like Meta, that’s their superpower. Meta can build these uplift models at scale because of logged-in users and granular ad logs.

Unfortunately, for marketers outside these walled gardens (and including all MTA, MMM vendors), we’re out of luck unless we gain access to the same depth of ad log data. This is also why you’re likely seeing data discrepancies between Meta and your MTA provider.

This is the new frontier of performance marketing, and it’s going to reshape how we think about omnichannel optimization and growth.

So where do we go from here? Angler’s predictive CAPI (P-CAPI) improves signal quality back to the platform, identifying the moveable middle — audiences who aren’t necessarily in-market shoppers, but have a high propensity to convert within the attribution window (say 7-days post click) once they discover the brand on Meta. As incremental value of such audiences higher than in-market value, schedule a call with us to learn how Angler's predictive signals can be directly used in the incremental attribution.