The effectiveness of adaptive models
Applying simple business rules to a sales strategy enables you to identify eligible actions for a customer. However, business rules alone will not enable you to select the best action for a customer, or the action the customer is most likely to accept. As a result, action acceptance rates can be low when only business rules are used to make sales decisions.
Note: Please be aware that actions can be renamed and are sometimes referred to as propositions or offers.
To improve acceptance rates, augment the business rules in a decision strategy with analytics.
Applying adaptive analytics to your decision strategies enables the strategies to detect changes in customer behavior in real time so that you can act on the changes immediately.
Pega Adaptive Decision Manager
Pega Adaptive Decision Manager (ADM) is a component that allows you to build self-learning adaptive models that continuously improve predictions for a customer. ADM can automatically detect changes in customer behavior and act on the changes in real time, which enables business processes and customer interactions to adapt instantly to the changing interests and needs of customers.
Adaptive decisioning continuously increases the accuracy of its decisions by learning from each response to an action. For example, if a customer is offered and then accepts a product, the likelihood that customers with a similar profile also accept that offer increases slightly. The mathematical expressions of these probabilities in the model are regularly updated.
ADM is a closed-loop system that automates the model creation, deployment, and monitoring process. The component can manage a large number of models without human intervention.
In contrast to predictive analytics, which requires historical data and human resources to develop a reliable predictive model, adaptive decisioning can start to calculate who is likely to accept or reject an offer without using any historical information, learning on the fly. Adaptive decisioning captures and analyzes response data in real time, which is useful in situations where the behavior itself is volatile. A typical use case is predicting customer behavior following the introduction of a new offering.
You can use predictive models as an alternative to, or in conjunction with, cases where data is available for offline modeling.
Adaptive decisioning creates binary models and uses these models for predictions. The full adaptive modeling cycle consists of the following steps:
- Capture response data in real time from every customer interaction.
- Use sophisticated auto-grouping to create coarse-grained, statistically reliable numeric intervals or sets of symbols.
- Use predictor grouping to assess inter-correlations in the data.
- Use predictor selection to establish an uncorrelated view that contains all relevant aspects of the action.
- Use the resulting, statistically robust adaptive binary model for scoring customers.
- Whenever new data is available, update the data model.
Adaptive decisioning can also build channel-specific models that account for differences in customer responses to outbound versus real-time inbound offers.