Facebook Synthetic Modeling

Updated 2 weeks ago by Emily Shreero

Facebook is a walled garden, which impacts the ability for third parties to track view-based touchpoints or conversions directly. However as we all know, viewing an ad often has a significant impact in driving a user to convert. It's critical to credit Facebook views for driving conversions.

Rockerbox solves for this through a synthetic modeling process that inserts view-based touchpoints on a user's path to conversions. Read on for more detail!

Synthetic Modeling Methodology

The process of identifying users most likely to have arrived on-site from viewing a Facebook ad and selecting the specific ad that most likely drove the visit is outlined below.

  1. Facebook Data: Facebook Synthetic data begins with an understanding of campaign performance reported by Facebook from their hourly 1 day post-view conversion data. This allows Rockerbox to attribute aggregate view-based touchpoints to users most likely to have viewed a campaign with a specific targeting type (ex prospecting vs retargeting).
  2. First Party Data : your first-party data informs which users were most likely to have had a Facebook view on their path to conversion based on user properties. User eligibility for having a Facebook Synthetic touchpoint on their path is determined using:
  • Site Visit Data: determines when a customer arrives on-site and if they've visited previously
  • Device Graph Analysis: identifies the other marketing activity that may rule out Facebook as an eligible touchpoint using a device graph to expand the set of anonymous identifiers associated with a single user.
  • Multi-Touch Attribution (MTA) Data: helps establish eligibility criteria for each user by understanding their interactions with other marketing channels
  • UTM and Querystring Parameters: further refines eligibility by disqualifying visits from other attributable, non-navigational channels. For example, a user may be eligible for a Facebook Synthetic touchpoint if they arrives on-site from a brand search click, but disqualified if their site visit is tied to a click from a channel like Snap.
  1. Data Synthesis: this process involves synthesizing Facebook data and first party data to further assess which campaigns drove specific visits. In addition to the above, the synthesis also involves user ranking, through which we select the highest eligible user for each campaign until all Facebook-reported campaign conversions are filled in based on hourly data.

Output of Facebook Synthetic Modeling

The modeling process outlined above ultimately creates a Facebook view-based touchpoint on a user's path to conversion. This synthetic touchpoint is then treated like any other marketing touchpoint, meaning

1. Where to see clicks vs Facebook Synthetic views: conversions will be reported in all first-party reporting down to the ad level as an aggregate of views + clicks. View-based conversions vs click-based conversions will not be reported on separately.

2. How credit differs for Synthetic views vs clicks: it doesn't! The amount of credit given to a Facebook Synthetic touchpoint vs a click will not differ. Rather, it's the other touchpoints on a user's path that determine how much credit is given to the Facebook touchpoint.

Technical Setup

Synthetic modeling to account for view and mobile click based touchpoints in most cases can only be applied to 1 conversion event in Rockerbox. This is most often the primary conversion segment you optimize against (ex Purchase).

The corresponding Facebook conversion event must also be identified during setup. For example, you may apply Facebook Synthetic Modeling to your Purchase conversion segment in Rockerbox, using your Purchase - CAPI Facebook event as an input.

How did we do?