AI-Powered Business Intelligence

Chapter 8. AI-Powered Prescriptive Analytics

Now that you have learned how to leverage AI to analyze data from the past and make predictions about the future, it’s time to discuss recommended actions! In this chapter, you will learn how to support data-driven decision making by letting an algorithm suggest the best option out of a range of possible actions. Let’s go!

Use Case: Next Best Action Recommendation

For this use case, we are building on the suggestions of a predictive model and selecting the best option for a specific customer.

Problem Statement

We are working on the BI team of a large telco provider. The company sells various products, such as monthly or yearly cable and cell phone subscriptions, and has millions of active customers each year. The company is facing an increasing problem of churning customers.

The data scientists on the team have successfully built a customer churn model that predicts the likelihood of single customers to churn by the end of the current quarter. The churn predictions have been calculated as a churn score: 100 means highest churn probability, and 0 means lowest churn probability. While the churn predictor has proven quite accurate and useful in identifying those customers who are likely to cancel their subscription with the company, the business still struggles with the right measures to counter churn. The BI team has incorporated a churn prevention dashboard, shown in Figure 8-1.

Figure 8-1. Baseline churn prevention dashboard

The ...