Einstein Engagement Scoring for Mobile Model Card

Person or Organization

Salesforce Einstein for Marketing Cloud Engagement

Model Date and Version

• August 2020
• Minor changes can occur throughout the release
• Major changes can occur and are communicated via release notes

Model Type

Forecasting, Time-series analysis, Clustering

General Information

• Einstein Engagement Scoring for Mobile predicts five types of engagement scores for each individual mobile consumer contact within each customer’s apps.
  • Likelihood to tap on a push notification
  • Likelihood of an inferred open
  • Predicted spending time in the app over the next 7 days
  • Predicted app sessions over the next 7 days
  • Predicted overall, aggregated app engagement
• Push direct open cards show the likelihood of your engaged contacts to open an app directly from a push notification. The trend shows the change in average likelihood for the past 7 days.
• Push indirect open cards show the likelihood of your engaged contacts to open an app after receiving a push notification without tapping the app from the push notification. The trend shows the change in average likelihood for the past 7 days.
• The model further classifies each score into one of four categories based on customer-defined thresholds or machine-inferred thresholds.
• The model analyzes an individual contact’s historical events and engagement patterns to forecast future engagement scores. More recent behavior and engagement patterns are emphasized more heavily in the predictions.
• The model clusters predicted engagement scores to derive local threshold boundaries to classify contact engagement levels.
• Salesforce trains each org’s model using only that customer’s app data. If an org opts in to the Global Models feature, they get the benefit of a wider pool of probabilistic contact matching with a greater range of training data.
  

  Tip Learn more about Einstein Global Models in Marketing Cloud Engagement.

Constraints

• A minimum of 1 push or open event is required for a contact to have prediction scores.

License

Mobile Engagement Scoring is available to use for Marketing Cloud Engagement customers with any of these editions.

• Corporate Edition
• Enterprise Edition
• Enterprise+ Edition

Primary Intended Uses

The Einstein Mobile Engagement Scoring model offers five application-specific, predictive engagement scores for each mobile subscriber. These scores help marketers conduct further segmentation or improve targeting on future mobile campaigns and sends.

Out-of-Scope Use Cases

Anything other than the primary use case is out of scope and isn’t recommended.

Model Input

Einstein Engagement Scoring for Mobile analyzes up to 90 days of historical engagement patterns of the subscribers. The engagement history includes these factors.

• Push events and reactions, such as push notifications and taps on push notifications, and associated timestamps
• App engagement behavior, such as app opens and time spent in app, and associated timestamps
• Data and metadata about customer sending patterns, including how campaigns are executed
• Categorization thresholds defined by customers when applicable

The engagement history that Einstein Engagement Scoring for Mobile analyzes excludes these factors.

• Data purchased or collected from third parties
• Demographic data that is typically stored in SFMC as data extensions or subscriber or contact attributes
• Specific content within the push notification template or rendered message body

Model Output

• Up to five types of scores are predicted for each subscriber for each app.
• For each app, the model produces an average score across all subscribers and associated model confidence level.
• The model produces data extensions that contain each subscriber’s predicted scores and the corresponding category.
• The model recommends local thresholds based on your data for each score type.

Groups

The model doesn’t include demographic data or other data purchased from third-party data providers.

Environment

The model is trained and deployed in the Marketing Cloud Engagement environment.

Model Performance Measures

Model performance metrics include metrics like model fitness score (root mean squared error, mean absolute error) and data richness score.

Salesforce customers are responsible for their interpretation of results generated by the model. Customers must be aware of any assumptions they make when acting based on scores generated by the model. There can be adverse outcomes in certain circumstances. For example, if the model predicts low engagement, a marketer can simply choose not to send a marketing communication. If that prediction were erroneous, certain audiences or individuals can be excluded from access to benefits or opportunities.

Scores and Models

Einstein Engagement Scoring scores and models for email are updated approximately weekly. The refresh cadence varies by one to several days based on a customer's individual business unit.