Overview

Measuring attribution is becoming increasingly challenging and complex. As the privacy landscape evolves, shifting guidelines and regulations become more and more difficult to keep up with. As these rules grow and change, attribution solutions are becoming more intricate and harder to maintain.

On top of all this, there is a growing number of opted-out users in marketing programs, lowering the accuracy of campaign performance measurements. This leaves you guessing as to how to optimize user engagement, especially among iOS users.

Predictive Aggregate Measurement (PAM) is an innovative approach to mobile campaign attribution, designed for performance marketers navigating today’s complex, privacy-focused mobile measurement world.

PAM leverages attribution modeling mechanisms and differential privacy controls to help ensure user data is protected, while still providing accurate, granular insights regarding campaign performance.

What Is PAM?

PAM is Branch's new approach to mobile campaign attribution!

The latest advancement in our suite of mobile measurement solutions, PAM is designed to navigate the complexities of today's changing privacy landscape while expanding attribution coverage.

By utilizing our predictive modeling tools alongside strict privacy protocols, PAM helps safeguard user privacy while also providing marketers with the most granular and effective reporting methodology available for each unique situation.

Our dynamic models adjust to the data available, employing the most precise attribution methods for each situation. We do this with user privacy in mind, so marketers can focus on analyzing performance and optimizing their strategies.

Data Precautions

PAM maximizes attribution coverage while still remaining privacy-focused.

Below are the methods PAM uses to help you meet your privacy compliance requirements.

Data Collection

PAM collects campaign data in a way that prioritizes user privacy by using aggregated identifiers and signals for users who have opted out of tracking.

These users' conversions will still be annotated with campaign data. However, this campaign data will not include individual identifiers (such as device IDs).

Data is gathered from various linking touchpoints across multiple platforms, offering a comprehensive view.

Thresholding

PAM applies thresholding techniques to the data collected, meaning individual user actions cannot be traced across mobile apps and the web. In this way, PAM helps maintain user anonymity.

These techniques allow marketers to access valuable insights through aggregated data, without exposing individual user-level information.

Data Aggregation

To further help protect privacy, PAM aggregates campaign data, bringing individual data points into a broader dataset.

This provides a comprehensive view of campaign performance but doesn't expose data about the individual user.

The aggregated data is presented in a way that marketers can get actionable insights while preventing user-level identification.

Together, these three data precautions help ensure that your user's privacy is respected, while still giving you valuable insights.

Implementation

Prerequisites

To enable PAM, you first need to add our Performance product to your Branch account. To learn more about our products, visit our packaging page.
You also need to have access to make changes at the org-level or app-level, depending on what level you want to enable PAM on.

Org-Level Enablement

To enable PAM at the org-level:

Go the Branch Dashboard and select the org that you want to enable PAM for.
Once you've selected the org, use the left-side navigation bar to click on Configuration, which is in the Configure section.
Next, navigate to the Attribution tab. There, you will see a switch that enables PAM.

Be careful! This enables PAM for the entire org.
Flip the switch to enable PAM for the org, then make sure to click Save at the bottom of the page.

App-Level Enablement

To enable PAM at the app-level:

Go the Branch Dashboard and choose a specific app you want to enable PAM for.
Use the left-side navigation bar to click on Configuration, which is in the Configure section.
Next, navigate to the Attribution tab.
Check to see if the switch inside the Inherit Organization Attribution Settings section is enabled. If it is, then your app is inheriting attribution settings from the org it lives under.

Decide if you still want to make PAM changes at the app-level.
Also note that the Attribution Windows section will be greyed out if your app is inheriting from org-level attribution settings.
If you've taken note of the org-level settings and you still want to enable PAM at the app-level, flip the PAM switch and then click Save at the bottom of the page.

Watch the video below for an illustrative demonstration of how to turn PAM on at the app-level:

FAQ

Does Branch do probabilistic matching?
Branch provides customers with both Predictive Modeling (PREM) and Predictive Aggregated Measurement (PAM).
What is required to implement PAM?
Implementing PAM doesn't require any engineering resources and is an easy lift for marketers. Simply follow this guide to enable PAM at the org-level or app-level with a switch.
Is Branch reverting to modeled conversions?
Modeled conversions from Ad Networks will flow into the attribution waterfall of PAM.
How will Branch dedupe this aggregate data against the log-level owned channel data?
PAM will utilize the deduplication logic we have with opt-in conversions and SKAN. The data will live within
our Event-Level Reporting suite and be deduplicated against all other sources within our Unified View Analytics.
What do “Differential Privacy Controls” entail?
Differential Privacy Controls include various end-user privacy practices that neither rely on nor expose individual identifiers across apps and websites. Examples of Differential Privacy Controls include (but are not limited to): noise introduction, thresholding, etc.
Can you consider a conversion deterministic if you’re using an aggregate ID?
Yes, if the aggregate ID is not individually identifying. This implies a many to one relationship, which can be pulled deterministically and referenced deterministically.
Is fingerprinting forbidden?
Yes, fingerprinting is forbidden. Branch does not use fingerprinting - we utilize probabilistic matching techniques.