The model in this card analyzes how open events are calculated and the likelihood that each open event is real.
- Model Details
The Einstein Probabilistic Opens model in Marketing Cloud Engagement analyzes message open likelihood using these training algorithms, parameters, fairness constraints, features, and other applied approaches. - Intended Use
The Einstein Probabilistic Opens model in Marketing Cloud Engagement is intended for these use cases. - Relevant Factors
These factors are associated with the Einstein Probabilistic Opens model in Marketing Cloud Engagement. - Metrics
Einstein evaluates and monitors model performance metrics to ensure and improve the quality of the model. These performance measures are associated with the Einstein Probabilistic Opens model. - Training Data
You have a customized version of the model that’s trained on your data alone. Data from one Salesforce customer doesn’t affect the behavior for another Salesforce customer. While model training happens for each customer on their data, the initial development of the model is validated with a representative set of pilot customers’ data. - Ethical Considerations
Review the ethical factors associated with the Einstein Probabilistic Opens model in Marketing Cloud Engagement
Model Details
The Einstein Probabilistic Opens model in Marketing Cloud Engagement analyzes message open likelihood using these training algorithms, parameters, fairness constraints, features, and other applied approaches.
Person or Organization
Salesforce Einstein in Marketing Cloud Engagement
Model Date and Version
- December 2021
- Minor changes can occur throughout the release.
- Major changes can occur and are communicated via release notes.
Model Type
Time-series analysis, Predictive modeling
General Information
Einstein Probabilistic Opens assigns each email open event a confidence score that represents the event being a human-initiated real open. The score ranges from 0% to 100%. Higher scores represent higher confidence that the open is real. Lower scores represent a higher chance that the open event is machine-generated.
The model analyzes a subscriber’s historical events and engagement patterns to predict the chance that an open is real. The model performs event association and attribution to monitor behavior changes. Most recent behavior changes and engagement patterns are emphasized more heavily in the predictions.
Compared to a typical filtering-based approach, the probabilistic open model provides more accurate results by reducing the risk of false positive and false negative results while minimizing information loss. Probabilistic opens are foundational to all other Einstein modeling and products. Einstein trains each customer’s model using only that customer’s data.
Constraints
There are no discernible constraints.
License
Einstein Probabilistic Opens is available to Marketing Cloud Engagement customers with any of these editions.
- Corporate Edition
- Enterprise Edition
- Enterprise+ Edition
Intended Use
The Einstein Probabilistic Opens model in Marketing Cloud Engagement is intended for these use cases.
Primary Intended Uses
- The Einstein Probabilistic Opens model provides a layer of information that helps Einstein calculate how trustworthy each email open is. This model is in response to Mail Privacy Protection implemented by some email client providers. As a result, Einstein can build more accurate models and provide better recommendations.
- Marketers can gain insights from Einstein Probabilistic Opens that provide a more realistic view of subscriber engagement levels. Using this insight, marketers can conduct further segmentation or better targeting on future campaigns.
Out-of-Scope Use Cases
Anything other than the primary use case is out of scope and not recommended.
Relevant Factors
These factors are associated with the Einstein Probabilistic Opens model in Marketing Cloud Engagement.
Model Input
Einstein Engagement Frequency analyzes up to 90 days of historical engagement patterns of the subscribers. The engagement history includes these factors.
- Email sends, bounces, and engagement events including open, click, and unsubscribe, spam complaints, and associated timestamps
- Data and metadata about customer sending patterns, including how campaigns are executed
The engagement history that Einstein Probabilistic Opens analyzes excludes these factors.
- Data purchased or collected from third parties
- Demographic Data, which is typically stored in SFMC as data extensions or subscriber or contact attributes
- Specific content in the email template or rendered email body
Model Output
For each open event, a probabilistic score is produced to measure the likelihood that the open is real.
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.
Metrics
Einstein evaluates and monitors model performance metrics to ensure and improve the quality of the model. These performance measures are associated with the Einstein Probabilistic Opens model.
Model Performance Measures
Model performance metrics include data like model mean absolute error score.
Training Data
You have a customized version of the model that’s trained on your data alone. Data from one Salesforce customer doesn’t affect the behavior for another Salesforce customer. While model training happens for each customer on their data, the initial development of the model is validated with a representative set of pilot customers’ data.
Ethical Considerations
Review the ethical factors associated with the Einstein Probabilistic Opens model in Marketing Cloud Engagement
As you interpret your data, be aware of making assumptions based on the scores generated. Scenarios with potential adverse outcomes can occur in certain circumstances.
