Skip to main content

Prioritizing actions using AI

4 Tasks

20 mins

Visible to: All users
Beginner
Pega Customer Decision Hub 8.7
English
Verify the version tags to ensure you are consuming the intended content or, complete the latest version.

Scenario

When customers log in to the U+ Bank website, they see the credit card offers for which they qualify based on the engagement policy defined by the bank.

When customers qualify for multiple credit card offers, AI decides which is the best offer to show.

The business understands how the Pega Customer Decision Hub™ AI works in the context of a web channel. When a customer sees an offer, but doesn't click it, the AI considers this as a negative behavior and lowers the propensity of the offer. When the customer clicks an offer, the propensity of that offer is increased.

To complete the assignment, use the following credentials:

Role User Name Password
Decisioning Architect DecisioningArchitect rules

Your assignment consists of the following tasks:

Task 1: Verify that propensity is enabled in arbitration.

In the Arbitration section of Next-Best-Action Designer, keep the Propensity check box enabled to ensure that only AI is used in the arbitration process to select the top action.

Task 2: Verify the negative behavior.

Successively log in to the U+ Bank website as Troy, but do not click the offer. Check that after multiple logins, the propensity of an offer decreases when you do not click it. Use the Customer profile viewer > Interaction history report to examine the propensity value. Troy's subject ID is 14.

Note: The Priority value is used to prioritize actions and select the top action. Because you enabled only Propensity, and context weighting is equal to 1, the Priority is equal to Propensity. The negligible difference is due to the adjusted propensity, which is not in scope of this exercise.

Task 3: Verify the positive behavior.

Log back in multiple times and click an offer each time to record a positive behavior and to prove that the AI learns, and that the propensity of the offer is increased for customers with a similar profile.

Note: The system decides when the actual learning happens. As a result, you might see some slight variations in the propensity and not a continuous increase after each click. However, multiple clicks do lead to an increased propensity after a while.

Task 4: Examine the AI model.

Finally, in Pega Customer Decision Hub, examine the AI model behind the action or treatment.

 

You must initiate your own Pega instance to complete this Challenge.

Initialization may take up to 5 minutes so please be patient.

Challenge Walkthrough

Detailed Tasks

1 Verify that propensity is enabled in arbitration

  1. Log in to Customer Decision Hub as the Decisioning Architect with User name DecisioningArchitect and the password rules
  2. In the navigation pane on the left, click Next-Best-Action > Designer.
    NBA Designer
  3. In Next-Best-Action Designer, click Arbitration.
  4. In the Propensity section, ensure that the switch on the right is turned on.
    turn on propensity
  5. Scroll down to the Business levers section.
  6. In the Business levers section, ensure that the switch on the right is turned off.
    turn off business levers

2 Verify the negative behavior.

  1. On the exercise system landing page, click U+ Bank to launch the U+ Bank website.
    exercise system
  2. In the U+ Bank website, in the upper-right corner, click Log in to log in as a customer.
  3. In the Username drop-down list, ensure that Troy is selected, and then click Sign in. An offer is displayed.
    Note: Do not click the offer. Each time you log out and then log in, Troy is presented with an offer, but do not click it.
  1. Perform successive logins without clicking the offer.
    Note: For this exercise, the system is configured to record a negative outcome after 60 seconds. During successive logins, ensure a 60-second time interval before logging out again.
  1. In the Pega Customer Decision Hub, click Explore arbitration results to open Customer profile viewer.
    Explore arbitration results
  2. In the Customer Profile Viewer window, in the Customer ID filed, enter 14 which is Troy's customer ID and click View.
    Customer 14
  3. In the profile of Troy, click Interaction history tab.
    Interaction history
     
  4. In the Interaction history tab, click Load all interactions.
  5. Check that the impression decreases with every customer interaction.
    Negative behaviour
    Note: The values you see in your system vary depending on the number of times you log in to U+ Bank app.
    The AI uses Thompson sampling to add some noise to the propensity values. Therefore, you might see some slight variations in the propensity, and not a continuous decrease in the values. If the models have received very few responses, a lot of noise is added. When the number of responses received increases, less noise is added and therefore the propensities returned are very close to the raw model propensity.
  1. Optional steps to see how the original propensity changes:
    1. In the profile of Troy, click Decision history.
    2. In the Decision history tab, click Load all decisions.
    3. Click Fields.
    4. Select Original model propensity from the list of fields.
      Original Propensity

3 Verify the positive behavior

  1. After you have tested the effects of the negative behavior, in the U+ Bank website login as Troy and click the Learn more link below the offer. Do multiple logins and click the Learn more link every time. This will increase the propensity of the offer. Click Load all interactions and note that after 3 logins and clicks on Learn more, the propensity has increased in the Interaction History report.
    positive behaviour
    Note: The system decides when the actual learning happens. Therefore, you might see some slight variations in the propensity, and not a continuous increase after every single click. However, after a while multiple clicks will lead to an increased propensity. If the models have received very few responses, a lot of noise is added. When the number of responses received increases, less noise is added and therefore the propensities returned are very close to the raw model propensity.
    For click-throughs: For this exercise, the system is configured to record a negative outcome after 60 seconds. Ensure that you click Learn more within the 60-second timeframe to avoid a negative outcome. Allow a minute for the positive click to be registered.
  1. Optional steps to see how the original propensity changes:
    1. In the profile of Troy, click Decision history.
    2. In the Decision history tab, click Load all decisions.
    3. Click Fields.
    4. Select Original model propensity from the list of fields.

4 Examine the AI model

  1. In the menu on the left, navigate to Intelligence > Prediction Studio to view the AI models. This will open Prediction Studio.
    open prediction studio
  2. In the Predictions window, click Predict Web Propensity > Open prediction
  3. Click Models tab to view the AI models.
  4. Click the Web_Click_Through_Rate of type Web_Click_Through_Rate_Customers to open the treatment level model.
    Web click through rate customer
    Note: Analyzing the AI models in detail is beyond the current scope of this Challenge. For now, just examine the list of predictors and the outcomes already configured by a Data Scientist.
  1. On the Monitor tab, check the overview of all the models in a given channel.

For example, you can see that the best-performing model is the Premier Rewards card, and that the model with the least responses, around 1500, is the Standard Card.

performance of actions
  1. On the Predictors tab, check the list of predictors configured at the time of creation of this model. Not all of the predictors can be used by the AI.
    predictors
  2. On the Outcomes tab, check that the action outcomes are mapped to positive and negative behaviors:
    1. Check that Clicked is mapped to positive and NoResponse to negative.
      outcomes of model
  3. On the Monitor tab, open the Model Report for one of the models. For example, the Rewards Plus model:
    Model repo
  4. On the Predictors tab, click on a predictor to open it. Examine the unique values or number ranges in the predictor, and the value or range with the highest propensity.
    Anual income
     
    Graph
    Note: The actual view of the predictors can be different depending on the amount of clicks that you made earlier.
  1. On the left pane, click Models.
  2. On the Models landing page, open the Omni_Adaptive_Model. The Omni_Adaptive_Model is the action level model.
    omni adaptive model
  3. In the lower-left corner, click Back to Customer Profile to return to Customer Profile Viewer in the Customer Decision Hub portal.
    Back to Customer profile
  4. In the navigation pane of the Customer Decision Hub portal, click Content > Actions and then open any of the credit card actions.
  5. On the Details tab of the action that you open, in the CUSTOMER PREDICTORS section on the right, check that the system displays the list of predictors used by the AI model, as in the following example for the Rewards Plus card:
    Customer predictors
    Note:

    The OmniAdaptiveModel AI model is an action-level model, which means that the model predicts customer behavior without considering any particular treatment.
  6. Click the Treatments tab to see the treatment-level AI model.
  7. In the Web section, click the Polaris icon to view the details of the treatment-level AI model.
    Polaris
  8. In the Analytical model for Rewards Plus card tile window, check the snapshot of the treatment-level AI model.
    You can see the current confidence level in predicting customer behavior and the predictors used to make the predictions, as in the following example:
    Analytical model

This Challenge is to practice what you learned in the following Module:


Available in the following missions:

If you are having problems with your training, please review the Pega Academy Support FAQs.

Did you find this content helpful?

Want to help us improve this content?

We'd prefer it if you saw us at our best.

Pega Academy has detected you are using a browser which may prevent you from experiencing the site as intended. To improve your experience, please update your browser.

Close Deprecation Notice