Adjusting Next-Best-Action catalogue to support changed business priorities

From PegaWiki
This is the approved revision of this page, as well as being the most recent.
Jump to navigation Jump to search

Adjusting Next-Best-Action catalogue to support changed business priorities

Description Adjusting Next-Best-Action catalogue to support changed business priorities
Version as of 8.1
Application Pega Customer Decision Hub
Capability/Industry Area Next Best Action



This article is part of a series about using Pega Customer Decision Hub to communicate with customers during emergency events. For an overview of the process, see Rapid Response to Emergencies. For the previous article in the series, see Adjusting arbitration to reflect changed business priorities.

Adjusting Next-Best-Action Catalogue

With the Next-Best-Action Designer in Pega Customer Decision Hub you have already created a taxonomy for various actions, treatments, predictive analytics (adaptive modeling or offline model scores) and engagement policies to manage your existing Next-Best-Action strategy. In the previous steps you have expanded it to include an Information issue and related groups.

Take this opportunity to expand your action taxonomy's mix of actions for education, loyalty, fraud, collections to address these issues within the context of an emergency situation or a “new normal”. Re-assessing your taxonomy allows you to think outside in, from a customer perspective.

Here are some examples of actions Pega customers are deploying:

  • Pro-actively alerting customers about services such as deferred loan repayments, refunded fees, waived interest, and volume alerts from their contact center.
  • Using the branch channel to reach out and reassure customers personally, and creating mobile customer experiences to showcase payment holidays or credit limit extensions as they search on digital channels.
  • Reaching out to customers who may needs support to keep learning or working during the crisis and offering free high-speed data, waivers for service fees, and prioritization for 1st responders.

Next best action versus next best offer[edit]

Perhaps your taxonomy is heavily weighted with next best offers (NBO) – acquisition, upsell, cross-sell and retention offers.  Now that we have created an issue for Information-type actions and specific groups to further break those out, we can use these to differentiate between next best offers and next best actions.  We have done this in the previous steps setting business levers, business value, etc. for Information actions.

In a crisis there may be a single best action to do - not just a set of actions with varying levels of relevance – this is what has been achieved so far in addressing mandatory issues. Take the example of addressing a crisis, churn or bad payment situation versus making a cross-sell offer. Typically, one of these issues will be more important than others to address, for example information > churn > bad payment > cross-sell.

In this stage, context and relevance of the actions is taken into account, beyond priority.  In the Pega next-best-action taxonomy there should be a mix of cross sell actions, educational actions, loyalty actions, etc. - not just explicit sales offers (next best offer). Therefore, the final next best action isn't necessarily the top next best offer.  At the top level of the decisioning framework (the NBA rule) we check what issue to address. If according to the Information\Mandatory strategy there is an emergency issue to address, this will override all the NBOs: all NBOs are NBAs, but not every NBA is an NBO.

The more actions a customer is eligible for, the higher the potential outcome/payoff of the (top) action could be - the arbitration strategy has more options to choose from. Inbound, outbound, paid and event triggers could all prompt the NBA strategy, so the sum of outcomes is maximized. It is better to have good coverage of business issues that contribute to overall outcomes in terms of customer value and experience.

Implementation approach[edit]

Implementing next best action requires an iterative model, and it must be perpetually adapted and updated as business needs change. Consider:

Issues and groups: As detailed, create an Information issue and related groups.  It is easier to add new issues and groups than to move groups between issues or rename groups or issues.  Also, many things that the system does, such as reporting, depend on having the right categorizations, so you need to consider reporting requirements while adopting the issue hierarchy.

Issue hierarchy

Actions: Your catalog of actions will change over time, so issues and groups should accommodate the current actions and those that may be introduced in the foreseeable future.  The mix of Information Actions should reflect current needs and accommodate future, yet to be determined, actions. Getting a good handle on the Information actions is also very important because data requirements are derived from these new actions. If you know the type of action you want to make, you know the rules that calculate customer’s eligibility for that Action, and from that you can infer what any additional, incremental data requirements will be.

Likewise, you may want to consider adding mandatory actions within your existing taxonomy, not just within your Information issue.  For example, deferred loan repayments, refunded fees, waived interest actions may be best managed within product groups (e.g., Credit Cards, Unsecured Loans) in a Retention issue, where similarity in eligibility rules and arbitration are maintained.

Prioritization and arbitration:  Prioritization can lean towards the customer or the business as needed by way of business rules, propensity, business levers and action value. Customer objectives are typically met by emphasizing the propensity model, and as a result, the probability to accept. This is a natural way to measure the action's relevance to the customer. The company objectives are normally met by weighting actions in favor of product types, operational channels, or business issues and groups. Review simulation results and ensure that you understand the implications of these settings and how the new actions will be evaluated alongside your existing actions in all issues and groups.

Adaptive models: In order to predict the customer's propensity, and therefore the relevance of a proposition, adaptive predictive models should be used for Information actions and treatments. The best practice is to implement adaptive models from the outset of the utilizing these new Information actions. They enable the system to begin learning and adapting to customer behavior immediately.  Use business levers instead of artificially manipulating model propensities to affect Information action priority.

Business rules: By design, Eligibility, Applicability and Suitability rules and contact policies for Information groups and actions will be more explicit, hard rules than you design for more sales or relationship-oriented offers.  Again, review simulation results and ensure that you understand the implications of any business rules for any new actions alongside your existing actions in all issues and groups.

Weights and levers: Weights and levers are settings that enable you to change the balance of the prioritization to favor one objective, issue, group or action over another. For example, the business will probably want to prioritize some business issues over others, such as upselling over cross-selling in normal times and downplay both in case of an emergency. Use business levers as your best means to affect outcomes, as they are easy to change and maintain. Avoid making the arbitration logic too complicated and overly prescriptive. Try to keep the usage of components reasonable and realistic considering your overall arbitration needs, not just for this emergency use case.

Context weighting:  The final consideration is the context weighting (customer intent) which is only applicable on an inbound channel, via real-time containers. Context weighting is used to provide an additional increase or reduction in Action weighting based on the real time context of the request into Pega for an NBA.  An example is when a customer calls into the call center and wants to talk about a specific offer or customer service issue. In this case, the calling intent could be used to up-weight a particular issue or group and down-weight others. Use care in invoking customer intent; the same result may be achieved using business levers, which provide a more universal effect. Review simulation results and ensure that you understand the implications of these settings for your existing actions in all issues and groups.

Context weighting

Next steps[edit]

Manage your customer communications on an ongoing basis by leveraging self-learning models to ensure optimization in the new situation.