Adding behavioral data to a customer profile

This demo shows you how to use the accelerator component to leverage industry-specific best practice clickstream summaries and extend the customer profile with behavioral data to improve the performance of the predictive models.

The U+ Bank technical team wants to enhance the out-of-the-box best practices of Customer Profile Designer to capture clickstream data to extend a customer's profile and introduce powerful predictors to their existing adaptive models.

To capture clickstream data, download the Customer Profile Designer Accelerator component from Pega Marketplace. In Dev Studio, navigate to the application definition. Here, you can install the new component.

 

You can browse through the predefined components from Pega Marketplace to download the accelerator component.

 

Search for the Customer Profile Designer Accelerator component, and then download it to your computer.

 

Now, install the downloaded component and view the underlying rules.

 

After you successfully install the component, confirm that the Enabled checkbox selection is the default status.

 

In Dev Studio, confirm that the component is displayed in the Enabled components section of the application rule, and then save your changes.

 

The component includes several important artifacts:

A sample Clickstream Rest service that the system can invoke from a website.

 

A data structure to support the service payload.

 

A Clickstream data set for generic clickstream data.

 

Three example summaries for aggregation of clickstreams on industry-specific web pages.

 

The number of summaries and aggregate conditions that the system allows have limits to ensure that there is no impact on system performance. You apply these limits through the dynamic system settings.

The max_summary_associations_number dynamic system setting shows the number of summaries that you can create in Customer Profile Designer. The default value is 3.

 

The max_aggregates_number dynamic system setting is the number of aggregate conditions each summary can hold, which has a default value of 120. You can change these limits if necessary, but analyze the performance impact.

 

As a best practice, block the summary rules that you do not intend to use. In this case, you can block the Insurance and Healthcare summaries. Save the HealthcareClickstream summary to your Rules ruleset, and then mark it as Blocked. Save your changes.

 

Follow the same steps to block the InsuranceClickstream summary.

 

Navigate to the Profile Data Sources landing page in the Customer Decision Hub portal. The Clickstream data set that the component provides to store customer web click activity is now available as a data source. It is important to note that the system cannot directly associate streaming data with the primary context and needs to summarize the data into meaningful aggregations.

 

On the Summaries tab, the Financial services clickstream summary is available. Note that the data source for this summary uses the Clickstream data set. This setting means that every event that flows through the Clickstream data set factors into the calculation of the aggregations in the Financial services clickstream summary rule.

 

Launch the Financial services clickstream summary rule, and then click Save. The system begins a background process for creating the aggregate properties, dataflow scheduler, the corresponding class structure for the generated artifacts, and marking the properties as relevant for use in Next-Best-Action Designer.

 

This Financial services clickstream summary includes 90 aggregates, each with a unique aggregation condition. For example, select the Card page visits last 1 day aggregate to browse its condition. You can see that this aggregation counts the number of events received as PageView with a PageType of Card that occurred in the last one day.

 

With everything in place, associate this summary with your primary context in the Profile Designer.

 

Add a new summary to the associated data of the primary context. In this case, the only available data source is Financial services clickstream.

 

The Financial services clickstream summary rule has a single data source key, which is automatically mapped to the context key of the primary class.

 

Once you add the summary as a new association, save your changes.

 

You are ready to test the configuration. In Dev Studio, search for the Clickstream service REST rule for testing.

 

Run the service with a sample JSON payload as if the web page invokes the service for a customer who is viewing the Cards page.

{
"CustomerID":"14",
"InterestedIn":"",
"InterestLevel":"",
"Event":"PageView",
"PageType":"Card",
"DeviceType":"PC",
"PageViewActiveTime":"",
"CookieID":""
}

Now, in the Financial services clickstream summary rule that you associated with the customer, click the Records tab. Browse the aggregated data record that the system generates for the event.

 

You can now share the JSON structure and service endpoint with web developers.

The U+ Bank technical team has developed the integration between the website and Customer Decision Hub. Now, the customer's web activity streams in through the Clickstream service that you previously tested.

Launch the U+ website and log in as Troy.

 

At the top of the screen, click Credit Cards to open the Cards page.

 

By logging in to the Account overview for Troy and navigating to the Credit Cards page, you create some web activity for Troy.

In Customer Decision Hub, review the records in the Financial services clickstream to verify that the events streamed and processed successfully.

Confirm that the aggregate CardPageVisitLast1Day has increased. Note that the system began calculating additional aggregates, such as the average active seconds spent on the page. After reviewing the aggregates, launch Prediction Studio.

 

After completing the end-to-end testing, it is time to empower the adaptive models by leveraging the attributes that come with the summary rule. This task is typical for data scientist, but it is demonstrated to complete the use case.

 

In the Predict Web Propensity prediction, launch the Web_Click_Through_Rate adaptive model.

You can add new predictors on the Predictors tab of the adaptive model. Click the Add field list, and then select Add multiple fields to add more than one predictor.

 

The FSClickstream page in the Customer class contains all the predictors that the system calculates in the FSClickstream summary rule. Select all the properties, and then save your progress.

 

You have reached the end of this video. What did it show you?

• How to install the Clickstream component.
• How to associate a Summary rule with the primary context.
• How to test the Clickstream service.
• How to use the Behavioral data source as predictors.