Last week, Facebook published an information sheet to advertisers detailing three upcoming changes to its Ads Manager platform related to web campaigns. While all three changes will be of interest to advertisers, one, in particular, stands out as potentially the most impactful: starting on June 30th, Facebook will now model conversions on a 7-day click attribution window.
This change is fairly monumental. Back in December, Facebook updated its advertiser guidance around App Tracking Transparency to clarify that mobile web advertising campaigns run on its platform would be governed by the new privacy restrictions in the same way as mobile app campaigns. Accordingly, Facebook introduced a new mechanic, Aggregated Event Measurement (AEM), to facilitate the measurement of those campaigns. AEM functions similarly to Apple’s SKAdNetwork mechanism for mobile app campaign measurement, but AEM can only observe advertiser events for one day after an ad click. In other words: the click attribution window for opt-out traffic is just one day via AEM, versus the standard of 28 days in the pre-ATT environment.
This abbreviated observation window obviously represents a susbtantial reduction in measurement capability relative to what was available to advertisers before ATT. What Facebook has announced with its expansion of conversion modeling to a 7-day window is that it will attempt to giver advertisers more context around how its cohorts are projected to behave after clicking, even though that behavior isn’t observed directly. The value of this longer, modeled conversion attribution window is obvious: if Facebook is very good at modeling mid-term conversions from just one day of observed behavior, then it will better optimize ad spend across campaigns on the basis of those predicted mid-term conversions, and advertisers can therefore invest more money into those campaigns.
Of course, Facebook’s modeled conversion predictions aren’t auditable and can’t be verified. But most advertisers don’t have the analytical competence or bandwidth to build these types of prediction mechanisms internally, and so Facebook’s estimates will be the only ones available for use. The scope of data available to advertisers is so limited in the ATT environment that the advertising platforms have been given powerful license to grade their own homework: to own the feedback loop between ads delivery and measurement that drives budget allocation. And it’s not just Facebook that is doing this: Google is utilizing conversion value modeling to inform campaign spend allocation, too.
ATT created this opportunity for ads platforms to force advertisers to rely on modeled conversions: the entire point of ATT is to limit the amount of third party data that ads platforms can access, which by definition creates the need for projections in order to derive ROI estimates.
To Apple’s credit, the company recently announced that SKAdNetwork postbacks can be sent directly to advertisers, which prevents ad networks from withholding the accurate counts of conversion values that have been generated by cohorts. And further, ad networks have always modeled conversions, so conceptually, the notion of reporting projected conversions isn’t totally new. But ATT allows ad networks to dramatically reduce the amount of data upon which modeled output is derived, which gives ad networks an incredible amount of agency in terms of claiming conversions.
Ultimately, performance defines budget in the long term, even if reporting defines budget in the short term: ad networks can falsely claim conversions up to a point, but ad fraud, or faulty metrics, or advertiser incompetence at some point must contend with the harsh reality of ROI. Ad networks can plausibly inflate conversions that are modeled from limited data in the ATT environment for some amount of time, but they can’t do it forever.