Predictive Modeling (PREM)

Predictive Modeling is an enhanced matching service unique to Branch.

Overview

Predictive Modeling (PREM) is a probabilistic recognition system, that cross-references past user interactions across the Branch Link Graph, to more accurately attribute conversion events. PREM analyzes patterns in anonymous event metadata to uncover the most likely connection only used where allowed under platform policies These matches are a significant improvement over point-in-time FingerprintingFingerprinting - Fingerprinting is the process of matching device snapshots, made up of details like IP address and operating system, from an ad click to an app install in order to deep link a user. However, it can be an unreliable method of attribution with an error rate of 30-40% or greater. The best way to mitigate this inaccuracy is to employ a cross-platform attribution system., and extremely reliable for most use cases, but should not be relied upon for authentication or other critical failure use cases. The probabilistic estimation is based on the uniqueness of a combination of device characteristics, the presence of web cookies, and the number of devices seen frequenting the same IP address over time.

Benefits of using PREM

  • User-level metrics
  • Visibility of ongoing app activity
  • Support for different accounting and value-based methods like top-of-the-funnel assists, re-engagement campaigns, and MTA
  • Real-time optimization for continuous campaign optimization

How does PREM work?

  1. User clicks on any Branch Link
  2. PREM observes key traffic characteristics like:
    • Device IP
    • Timestamp
    • OS
  3. Branch attempts to match that device that has completed a conversion to an upstream advertisement click.
  4. PREM estimates the probability that the device is the same as a previously seen device based on the uniqueness of a combination of device characteristics, presence of web cookies, and number of devices seen frequenting the same IP address over time)

SafeTrack™

With the introduction of Apple's AppTrackingTransparency policy in iOS 14.5, Branch enabled SafeTrack™ for all customers using Branch. In compliance with privacy and industry standards, SafeTrack™ disables Predictive Modeling for ad attribution. With SafeTrack™, you outsource the technical complexity of keeping up with ever-changing compliance policies to Branch, and you'll have peace of mind knowing your marketing ops is in compliance with all known platform privacy norms.

SafeTrack™ Options

  1. Default: Disable probabilistic matching for ad links.
  2. Opt-in Expanded: Disable probabilistic matching for all links (available for Enterprise clients).

Please reach out to your Branch account manager to find out more about SafeTrack™ and its configuration options.


Did this page help you?