Machine Learning Models

Overview

Marketers are experiencing increased business demands, increased expectations from consumers, and increased volume, velocity, and variety of data. The Machine Learning feature in the Engagement Data Platform (EDP) is designed to handle all the "heavy lifting," and to present marketers with clear, meaningful results that can be used to improve targeting, and to derive new insights about their customers. The statistical models in the Machine Learning (ML) feature automate many of the necessary -- but non-strategic -- choices, thereby allowing marketers to focus on messaging strategies and content. 

The Cheetah Learning portal provides access to a wide range of training videos on how to use Cheetah Digital's products. Click here to view a training video on how to use Machine Learning. Please note that you must have a free Cheetah Learning user account in order to access this video.

Standard Models

The Engagement Data Platform offers a set of standard, "out-of-the-box" models available for all clients. Because these models use existing logic and analytical methods, they're quick to setup and begin using. These standard models can be used in their existing form, but they can also be thought of as templates, that can be modified to meet your unique business requirements and marketing strategy. 

The Machine Learning feature is integrated directly into the EDP application, and uses native data from your EDP database to calculate and score customers across the following three model types:

• Propensity: Propensity modeling attempts to solve a complex problem for marketers -- predicting future customer behavior. The outcome of a propensity model is a score for each customer from 0 to 100% that indicates how likely that customer is to engage in the predicted behavior in the future. For example, if you are predicting unsubscribe behavior for the next month, and a customer has a score of 0.9 (or 90%), then that customer has a 90% likelihood of unsubscribing in the next month, unless action is taken. The platform supports the following standard Propensity models:

• Propensity to Unsubscribe: Likelihood to unsubscribe in the next month.

• Propensity to Click: Likelihood to click a link in the next month.

• Propensity to Open: Likelihood to open a message in the next month.

• Propensity to Unengage: Likelihood to stop engaging in the next month. 

• Propensity to Purchase: Likelihood to make a purchase in the next month (this model requires purchase activity data to be loaded into EDP). 

• Send Time Optimization:  Marketers know that not all their recipients are most likely to engage with their communications at any one time of day, but tools for determining individualized optimal send times have been elusive. Send Time Optimization (STO) modeling focuses on when to target promotions to reach the maximum possible audience. The platform supports the following standard STO models:

• STO for Clicks: Optimize best time of day to send to generate click activity.

• STO for Opens: Optimize best time of day to send to generate open activity.

• Clustering: Within every customer-base, there exist natural groupings of customers based on their common engagement levels, demographics, and preferences. Marketers want to know how they can target these different groups as effectively as possible, but they don't always know how to identify those groups and commonalities (i.e. their makeup and behavioral characteristics) in order to tailor marketing programs to address them more uniquely. Cluster modeling creates, identifies, and explains the commonalities between groups of customers, so marketers can better understand and target them. The platform supports the following standard Clustering models: 

• Engagement Recency-Frequency: Segment customers based on how recently they responded, and how often they engage. 

• Personas: Segment customers based on their demographic attributes and engagement behavior.

• Recency-Frequency-Monetary (RFM): Segment customers based on their spend recency, frequency, and amount. This model type requires purchase activity data to be loaded into EDP. 

Citizen Data Scientist

Citizen Data Scientist (CDS) builds upon the platform's Machine Learning capabilities, by simplifying the complexities of data science tools, and delivering an intuitive modeling workflow accessible to marketers. The feature's user interface guides you through the process of creating your own Machine Learning models, thereby allowing you to customize any of Cheetah Digital's existing model types to suit your unique predictive use cases. See Citizen Data Scientist for more details on creating your own ML models. 

Access

To access the Machine Learning screen, select Machine Learning from the Main Navigation Menu, then select Machine Learning Models from the Sub-Category menu.

Additional Resources

• Machine Learning Capabilities

Features

The ML Models screen provides the following features related to managing your ML Models:

Searching and Sorting Click hereClick here Search The search feature allows you to search for a specified text string anywhere within the following ML Model fields: Display Name, Internal Name, or Tag.  Optionally, in the search field near the top of Machine Learning Models screen, enter the desired text string.  Press Enter, or click the search button (magnifying glass icon). The Machine Learning Models screen is refreshed to show only the models that meet your search criteria.  To remove all search criteria, click the remove button ("X" icon). Navigation Once you've found the desired model, you can navigate to the following other screens: Click the Machine Learning Model name to navigate to the ML Model Details screen.

 

Set Model Status Click hereClick here A Machine Learning Model must have a status (referred to as a "mode") of Published to be considered live and in use. To change the Model status: In the Action column, next to the desired ML Model, click the Change ML Model Mode icon.The "Edit Machine Learning Model" pop-up window is displayed.  Select the desired status for this Model -- Draft or Published.  Click save.

 

View or Edit Model Properties Click hereClick here To view or edit the properties of a Machine Learning Model: In the Action column, next to the desired ML Model, click the Edit icon.The "Edit Machine Learning Model" pop-up window is displayed.  Make any necessary changes to the model properties.  Click save.  To view the Model results: Click the Model name to navigate to the ML Model Details screen.

 

Copy a Model Click hereClick here To copy a Machine Learning Model: In the Action column, next to the desired ML Model, click the Copy icon. The platform creates a new Model with the same properties as the copied Model.

 

Delete a Model Click hereClick here To delete a Model: In the Action column, next to the desired ML Model, click the Delete icon. A confirmation dialog box is displayed. Click confirm.

 

Last Updated: April 2022

© Cheetah Digital    Engagement Data Platform Online Help 22.0  April 2022