Citizen Data Scientist

Overview

Citizen Data Scientist (CDS) builds upon the platform's Machine Learning capabilities, allowing you to customize any of Cheetah Digital's existing model types to suit your unique predictive use cases.
The Citizen Data Scientist feature is highly flexible. The feature supports the utilization of standard and custom EDP data attributes, all Event data, and the most common Activity types. You can deploy models against any target outcome variable of interest to detect all relevant "signals" needed to maximize marketing potential.
The output of Citizen Data Scientist varies depending on the model type used, but is typically a score or some other value written to each consumer record. This value can be then used for a variety of marketing purposes, such as building Segments for targeting in Campaigns, analyzing consumer behavior in reports, and personalizing message content.

Note: Citizen Data Scientist is an optional feature that must be enabled in your Customer Engagement Suite account. Please speak to your Client Services representative for more details.

Access

The Citizen Data Feature and capabilities are accessible from the Machine Learning screen for clients that have CDS enabled. 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

Citizen Data Scientist Overview Guide

Prerequisites

Before you can begin creating your custom models with Citizen Data Scientist, you may need to create the following supporting assets in EDP.

Segment: You can use a Segment to limit the consumers who are scored by the model.
Business Unit: You can use Business Units in two ways. First, consumers can be assigned to a specific Business Unit, and you can limit the consumers who are scored by the model based on their inclusion in a Business Unit. Second, Attributes in your People table can be assigned to a specific Business Unit. You can limit the Attributes used as inputs in the model process based on their inclusion in a Business Unit.
Attributes: You must define the Attributes where you want to store the results of the model scoring process. For the different model types, you must have the following Attributes defined:

Send Time Optimization

STO Preference: Integer field
STO Preference Known: Boolean field
STO Control Group: Boolean field

Propensity

Propensity Score: Decimal field
Propensity Group: String field
Propensity Decile: Integer field

Clustering

Cluster ID: Integer field
Cluster Group: String field

Note: You must define the above Attributes for each custom model that you intend to create. So, for example, if you create two custom Propensity models, you would need to create the score / group / decile Attributes for each model.

Features

Using Citizen Data Scientist, you can build custom models off of any of the standard model types (Sent Time Optimization, Propensity, or Clustering) to support your specific business strategy. CDS starts by selecting one of the above model types. The feature then walks you through the process of customizing the inputs to the model type, the scoring logic to use, and the destination of the final output score.

Create a Custom STO Model

Click hereClick here

In the Enter Display Name field, enter a name for your new custom model.
From the drop-down menu, select Send Time Optimization.
Click create. The Create ML Model pop-up window is displayed. This window is organized around several sections which walk you through the process of defining your custom model.

Basic Properties

The "Display Name" field is populated with the value you entered above; optionally edit this value.
The "Internal Name" is automatically populated based on the Display Name value. This field is disabled by default. To edit the Internal Name, check Edit Internal Name. A confirmation dialog box is displayed; click ok. Edit the Internal Name value.
In the "Description field," optionally enter a description of the custom model.
If you're using a Segment to define the audience of consumers to be scored, select it from the "Segment" drop-down menu.
If you want to use a Business Unit to limit which consumers are scored, select it from the "Business Unit" drop-down menu.
If you want this model to appear within the left-hand navigation menu, check show in navigation.
The Member Attribute Mapping section is used to select the Attributes where you want to store the output of the model. Next to the model output field, select the Attribute where you want to store that value:

STO Preference: The hour of the day that this customer prefers to be contacted.
STO Control Group: Yes / No flag indicating whether this customer has been placed in the control group used to calculate the performance lift associated with an STO Campaign.
STO Preference Known: Yes / No flag indicating whether the preferred send time for this customer is known.

The Advanced Options section allows you to override various default parameter values used in the model. To enter a different value for a parameter, check raw input, then enter the desired value in the text field.
Click next or select the Define Outcome tab.

Define Outcome

The Define Outcome section is used to configure the desired results for this model by identifying what you want the model to predict or quantify. To define the outcome for the custom model:

Select the Event type (such as opens or purchases) you want to use to determine a consumer's preferred sending time.
Define the time period to evaluate.
Click next or select the Review tab.

Review

The Review section is intended as a final summary of all the options that you selected for this custom model.

Review the settings for the custom model. If you identify something that needs to be changed or fixed, you can navigate back to that section using the tabs across the top of the window.
If you're satisfied with the model settings, click save.

Create a Custom Propensity Model

Click hereClick here

In the Enter Display Name field, enter a name for your new custom model.
From the drop-down menu, select Propensity.
Click create. The Create ML Model pop-up window is displayed. This window is organized around several sections which walk you through the process of defining your custom model.

Basic Properties

The "Display Name" field is populated with the value you entered above; optionally edit this value.
The "Internal Name" is automatically populated based on the Display Name value. This field is disabled by default. To edit the Internal Name, check Edit Internal Name. A confirmation dialog box is displayed; click ok. Edit the Internal Name value.
In the "Description field," optionally enter a description of the custom model.
If you're using a Segment to define the audience of consumers to be scored, select it from the "Segment" drop-down menu.
If you want to use a Business Unit to limit which consumers are scored, select it from the "Business Unit" drop-down menu.
If you want this model to appear within the left-hand navigation menu, check show in navigation.
The Member Attribute Mapping section is used to select the Attributes where you want to store the output of the model. Next to the model output field, select the Attribute where you want to store that value:

Propensity Score: The score assigned to this consumer.
Propensity Group: Customers are assigned a Propensity Group -- High, Moderate, or Low -- which represents a simple classification scheme for how likely that customer is to perform the selected action.
Propensity Decile: Customers are organized into deciles for easy grouping, each of which represents a tenth of the total population of the database

The Data section controls how you want to define the custom model. Select one of the following options:

Walk me through the process: Using the platform's user interface (described below), configure the model parameters.
Use custom query: Select a custom ML query for Build and Scoring.

The Advanced Options section allows you to enter a custom end-date for determining engagement. By default, the platform uses "yesterday" as the last date to determine engagement. To enter a different date, flip the Raw Input toggle to yes, then enter the desired date in the text field.
Click next or select the Define Outcome tab.

Define Outcome

The Define Outcome section is used to configure the desired results for this model by identifying what you want the model to predict or quantify. To define the outcome for the custom model:

Select one of the following options:

Future Marketing Event: Select the Event Type and marketing channel. Optionally, to filter the Events by Campaign type, click into "Campaign Types" and select one or more Campaign types. Use the "Time Period" section to define the time horizon for the predicted behavior (for example, "within the next three months"). Use the "Condition" section to define a mathematical condition indicating the number of predicted Events (for example, "greater than five"). As you define the Event, the platform updates the description.
Future Activity: Select the Activity Type and optionally a Sub-Activity Type. Depending on the selected Activity Type, you may need to select a Currency. Use the "Time Period" section to define the time horizon for the predicted behavior (for example, "within the next three months"). Use the "Condition" section to define a mathematical condition indicating the number of predicted Activities (for example, "greater than five"). As you define the Activity, the platform updates the description.
Member Attribute: Select an Attribute, then select how you want to use that Attribute. To predict future occurrences, enter the time horizon for the predicted behavior (for example, "within the next three months"). To find similar Members (i.e., "lookalike" modeling), enter the condition and time period to use.

Click next or select the Feature Selection tab.

Feature Selection

The Feature Selection allows you to configure the "inputs"to the Propensity model.

The Member Attributes section is used to select the Attributes used in the model. The Attributes are organized by Data Type (Numeric, Boolean, etc.). Within the section for a Data Type, the platform provides the following options for selecting the desired Attributes:

Add an Attribute: To manually add an Attribute, click into the selection field, then select the desired Attribute from the drop-down menu. Repeat this step as needed.
Remove an Attribute: To manually remove an Attribute, click the "X" icon next to the Attribute name.
Select All Attributes: Click "Select All Attributes" to add all standard and custom Member Attributes of this Data Type to the model.
Clear Selection: Click "Clear Selection" to remove all Attributes of this Data Type.

The Marketing Events section is used to select the Events used in the model. Select one of the following:

All Event Types: Use all standard Event types in the model. Also select a marketing channel. Optionally, to filter the Events by Campaign type, click into "Campaign Types" and select one or more Campaign types.
Custom Event Types: This option lets you select only the specific Event types you want to include in the model. Click into the selection field, then select the desired Event Type from the drop-down menu. Repeat this step as needed. Or, click "Select All Event Types" and then remove the ones you don't want. Also, select a marketing channel. Optionally, to filter the Events by Campaign type, click into "Campaign Types" and select one or more Campaign types.

The Activities section is used to select the Activity Type(s) used in the model. Click the Add button (plus-sign icon). From the first drop-down menu, select the desired Activity Type. Optionally select a Sub-Activity type from the second-drop-down menu. To add another Activity Type, click the Add button again and repeat this step as needed. To remove an Activity Type, click the trash can icon.
When finished, click next, or select the ML Configuration tab.

ML Configuration

The ML Configuration allows you to configure various rules and parameters within the model type.

Select an option:

Standard: Use the standard model configuration. This option is suitable for most use cases.
Custom: Customize the model configuration. The available parameters for this model type are displayed. Make any necessary changes to these parameters.

Note: The Custom option is intended for use by analysts and data scientists who need to customize the standard modeling process. The parameters available in this section are highly technical, and should not be modified without a clear understanding of their intended use

When finished, click next, or select the Review tab.

Review

The Review section is intended as a final summary of all the options that you selected for this custom model.

Review the settings for the custom model. If you identify something that needs to be changed or fixed, you can navigate back to that section using the tabs across the top of the window.
If you're satisfied with the model settings, click save.

Create a Custom Clustering Model

Click hereClick here

In the Enter Display Name field, enter a name for your new custom model.
From the drop-down menu, select Clustering.
Click create. The Create ML Model pop-up window is displayed. This window is organized around several sections which walk you through the process of defining your custom model.

Basic Properties

The "Display Name" field is populated with the value you entered above; optionally edit this value.
The "Internal Name" is automatically populated based on the Display Name value. This field is disabled by default. To edit the Internal Name, check Edit Internal Name. A confirmation dialog box is displayed; click ok. Edit the Internal Name value.
In the "Description field," optionally enter a description of the custom model.
If you're using a Segment to define the audience of consumers to be scored, select it from the "Segment" drop-down menu.
If you want to use a Business Unit to limit which consumers are scored, select it from the "Business Unit" drop-down menu.
If you want this model to appear within the left-hand navigation menu, check show in navigation.
The Member Attribute Mapping section is used to select the Attributes where you want to store the output of the model. Next to the model output field, select the Attribute where you want to store that value:

Cluster ID: Cluster identifier
Cluster Group: Cluster name.

The Data section controls how you want to define the custom model. Select one of the following options:

Walk me through the process: Using the platform's user interface (described below), configure the model parameters.
Use custom query: Select a custom ML query for Build / Scoring.

The Advanced Options section allows you to enter a custom end-date for determining engagement. By default, the platform uses "yesterday" as the last date to determine engagement. To enter a different date, flip the Raw Input toggle to yes, then enter the desired date in the text field.
Click next or select the Feature Selection tab.

Feature Selection

The Feature Selection allows you to configure the "inputs"to the Clustering model.

The Member Attributes section is used to select the Attributes used in the model. The Attributes are organized by Data Type (Numeric, Boolean, etc.). Within the section for a Data Type, the platform provides the following options for selecting the desired Attributes:

Add an Attribute: To manually add an Attribute, click into the selection field, then select the desired Attribute from the drop-down menu. Repeat this step as needed.
Remove an Attribute: To manually remove an Attribute, click the "X" icon next to the Attribute name.
Select All Attributes: Click "Select All Attributes" to add all standard and custom Member Attributes of this Data Type to the model.
Clear Selection: Click "Clear Selection" to remove all Attributes of this Data Type.

Define the period of time over which the platform will evaluate Event and Activity data. Enter an interval in the text field, then select a time period from the drop-down menu (for example, "6 months").
The Marketing Events section is used to select the Events used in the model. Select one of the following:

All Event Types: Use all standard Event types in the model. Also select a marketing channel. Optionally, to filter the Events by Campaign type, click into "Campaign Types" and select one or more Campaign types.
Custom Event Types: This option lets you select only the specific Event types you want to include in the model. Click into the selection field, then select the desired Event Type from the drop-down menu. Repeat this step as needed. Or, click "Select All Event Types" and then remove the ones you don't want. Also, select a marketing channel. Optionally, to filter the Events by Campaign type, click into "Campaign Types" and select one or more Campaign types.

The Activities section is used to select the Activity Type(s) used in the model. Click the Add button (plus-sign icon). From the first drop-down menu, select the desired Activity Type. Optionally select a Sub-Activity type from the second-drop-down menu. To add another Activity Type, click the Add button again and repeat this step as needed. To remove an Activity Type, click the trash can icon.
When finished, click next, or select the ML Configuration tab.

ML Configuration

The ML Configuration allows you to configure various rules and parameters within the model type.

Select an option:

Standard: Use the standard model configuration. This option is suitable for most use cases.
Custom: Customize the model configuration. The available parameters for this model type are displayed. Make any necessary changes to these parameters.

When finished, click next, or select the Review tab.

Review

The Review section is intended as a final summary of all the options that you selected for this custom model.

Review the settings for the custom model. If you identify something that needs to be changed or fixed, you can navigate back to that section using the tabs across the top of the window.
If you're satisfied with the model settings, click save.

Back to Machine Learning Models