Powered by AI and machine learning, Adobe Campaign’s Send-Time Optimization and Predictive Engagement Scoring can analyze and predict open rates, optimal send times, and probable churn based on historical engagement metrics.
Adobe Campaign offers two new Machine Learning models: Predictive Send Time Optimization and Predictive Engagement Scoring. These two models are machine-learning models that are specific to designing and delivering better customer journeys.
This capability is not available out of the box as part of the product. It is only available for Adobe Campaign Managed Cloud Services customers running Adobe Campaign Classic v7 or Adobe Campaign v8.
The implementation requires Adobe Consulting to be engaged. To find out more, reach out to your Adobe representative.
Predictive Send-Time Optimization predicts which is the best send time for each recipient profile for email opens or clicks and for push-message opens. For each recipient profile, the scores indicate the best send time for each weekday and which weekday is the best to send for best results.
Within the Predictive Send-Time Optimization model, there are two sub-models:
Predictive send time for open is the best time a communication must be sent to the customer to maximize opens
Predictive send time for click is the best time a communication must be sent to the customer to maximize clicks
Model Input: Delivery logs, tracking logs and profile attributes (non-PII)
Model Output: Best time to send a message (for opens and clicks)
Predictive Send Time Optimization is stored at profile level:
The model needs at least one month of data to produce significant results. These predictive capabilities apply only to email and push channels.
Predictive Engagement Scoring predicts the probability of a recipient engaging with a message as well as the probability of opting out (unsubscribing) within the next 7 days after the next email send. The probabilities are further divided into buckets according to the predicted level of engagement with your content: high, medium, or low. These models also provide the unsubscribe-risk percentile rank for the customers to understand where the rank of a certain customer is in relation to others.
The predictive engagement scoring lets you:
This model uses multiple scores to indicate:
These predictive capabilities only apply to email deliveries.
The model needs at least one month of data to produce significant results.
Model Input: Delivery logs, tracking logs and specific profile attributes
Model Output: A profile attribute that describes the profile’s score and category