Using multi-level decisioning to determine the next best action

From PegaWiki
Using multi-level decisioning to determine Next Best Action / This is the approved revision of this page, as well as being the most recent.
Jump to navigation Jump to search

Using multi-level decisioning to determine the next best action

Description How-to
Version as of 8.6
Application Pega Customer Decision Hub
Capability/Industry Area Artificial Intelligence



Introduction[edit]

Pega Customer Decision Hub has the ability to use more than one context at decision time to determine the next best action. This feature is called Multi-level decisioning. This capability allows you to run parallel decisioning using different contexts as the basis for making the decision. This gives greater scope and flexibility while keeping the complexity of engagement policies low.

Use case examples[edit]

In retail banking you want to consider the attributes of the customer (the person) to determine the next best action, but at the same time you want to look across all the customer's checking, credit, and investment accounts to see if there is a next best action for any one of those accounts.

In Insurance you want to consider the policy holder's attributes to determine the next best action and at the same time you want to look across all the policies to determine if there is a next best action for any of those policies.

Before you begin[edit]

It is critical that you have determined and configured your Context Dictionary. You may reference any of the applicable Industry templates as your starting point. In this step you will determine your primary context (generally customer) and up to two additional contexts.

Example context from Pega product documentation

While configuring your context is a required technical step, you should also consider the intent of your proposed multi-level actions. Some actions will have a clear decision context, while others may take a bit more consideration. The key question one should ask of each action:

"What am I trying to get customer to do? Is it related to the person or to a product / account / policy?"[edit]

Examples:

Customer Acquisition - A prospect has been browsing your website or you have purchased the lead from a third party. They do not own any of your products. You want to entice them to purchase one of your products. In this case, it is clear you should be making decisions against the primary customer context.

Service an Account - You have a customer that owns several of your products. It is beneficial for both you and the customer to entice the customer to set up for paperless statements on each account. In this case, you want to make a decision for each product they own and determine if setting up autopay for each is the right thing to talk about (and their maybe different priorities here). This is a good example of using Account / Product / Policy as the decisioning context.

Cross Selling - You have a customer who own one or more of your products. You would like to entice them to take an additional product. In this situation you are focused on making a decision in the context of the Customer, however, you may use information about the accounts / products / policies that they hold. This is an example of using the Customer context to make the decision, but using the associated Account / Product / Policy data.

Upsell a Product - You have a customer that owns one or more of your products. You would like to entice them to upgrade any or all of those products. In this scenario you could make the decision in the context of each Account / Product / Policy so that you can pick the best Upsell message for the best product.

Process steps to achieve objective[edit]

Let us walk through some specific examples inside the product. First assume you have the following context configuration in the Pega Customer Decision Hub instance:

This screen capture is taken from the Customer Profile Designer and indicates that Customer is our primary Context, Account is the Secondary context, and the Account table has also been added as Associated Data under the Customer Context.

Configuring the context for Customer Acquisition[edit]

We concluded earlier that this action needs to be created in the Customer context.

  1. On the Engagement policy tab of the new Silver Card action, select the Customer as the action context.
    A screenshot of the Engagement policy tab of an action.
  2. After saving this action, click Next-Best-Action > Designer > Engagement Policy > Acquire > Credit Cards.
The Silver Card shows up in the Customer Actions section, and not the Account Actions section.

To fine-tune the action's engagement policy using criteria, access the properties of the Customer context:

In this image you can see picking the Customer's age and seeing if it is 18 or greater

Configuring the context for Cross Sell[edit]

In this example we will place the action against the Customer context, but we will use properties from both the Customer context and the Account context.

  1. Pick a customer property.
    In the image above we pick a customer property as we did in the first example.
  2. We add a second criteria so we can target people that have deposit accounts. To do this we keep the Context as Customer and we browse the properties finding the Customer accounts reference to our Associated Accounts Data. By clicking the plus icon to the right, we are able to add the additional embedded logic we need.
A screenshot of the Customer accounts reference.

Now let's set up the finer details of what we are looking for inside the Customer Accounts page.

In the image above you see we have selected the Instances include option, and we are looking for at least 1 instance. Next we need to define what criteria defines an instance.
We select Create new condition so we can invoke the condition editor. In this example we are doing one criteria, but you can add additional criteria such as looking at Account Balance.
In the condition editor above, we have configured the condition to look for any account types with the type Deposit. This will look across all the customer's accounts and count those that are Deposit.
And since we have "instances include <=1" in the previous step, the entire criteria will be evaluated as true if there is at least one deposit account.

It is important to not be confused about the method described above, and for example, trying to pick the Account properties directly from within the Accounts page. In the image below, you see that if drill into the Customer Account page (not clicking the plus icon) you can indeed see all the Account properties and you could indeed look for Account Types = Deposit.

However this method would not allow you to select multiple criteria. If you attempted to add an eligibility criteria to look for account type, then a second criteria to do look at account balance, Pega Customer Decision Hub could not properly execute the logic because you want both these conditions executed at the same time as you iterate through the Account list looking for rows that meet all selected criteria. That is why you must use the embedded criteria as described previously.

Picking Account properties directly is not recommended for selecting multiple criteria

Configuring the context for Account Service actions[edit]

In this example we will entice a Customer to sign up for paperless account statements for each and every account where this is the next best action.

In the image above we are creating the Paperless Statement action and selecting Account as the Context.
And from here we can simply pick the Account context to the left and pick from the Account properties.
And we can add our criteria, for example looking for any Credit Card, Deposit, or Mortgage accounts

You may ask yourself why we were able to pick the Account type directly without having to perform the Instances contain and embedded criteria like we did in the earlier Customer context example. That is because we benefit from the fact that we have created this action in the Account context and we are looking at the Account context directly, as opposed to the previous action, which was created in the Customer Context. We are inside the Account context already, so criteria becomes more simple to execute.


To expand this example out a bit further, in addition to the Account properties, we may also need to reach back into the Customer data. For example, we only want customers who have deposit or credit accounts and have a credit score greater than 700 to get this action.

Here we have added a row of criteria, but notice we can select the Customer context because it is the primary context, and thus a parent of the Account context.
Here you see the criteria that uses both Customer and Account context on an action that is linked to the Account context.

Results[edit]

  1. You have seen how you should consider each action's context, ask yourself: "What am I trying to get the customer to do?" and choose the context that best achieves that goal.
  2. You have seen how to create actions in either the primary or secondary context.
  3. You have seen how to navigate to the properties you need within the context you have chosen for your action.
  4. You have seen how to use associated data in your primary context by using the instances within and embedded criteria.
  5. You have seen how to use the secondary context (e.g. Account) and also use the primary context.

At decision time, Pega Customer Decision Hub will evaluate all the next best actions for each context you have defined (e.g. Customer and Account). The strategy will use the data that you have configured in your engagement policy, arbitrate to find the next best action across both contexts, and deliver the true next best action.