ML Models Drive Actions to Accelerate Customer Centric Transformation

In today’s business world, customer experience (CX) is more important than ever. CX encompasses every interaction with the customer at all points, from learning about products, through the buy, get, use, pay, and renew phases of the journey. For successful organizations, digital transformation begins with customer experience, and CX analytics play a key role in providing a comprehensive view of the full customer journey, as well as granular insights on what is driving customer experience. At Lumen, our customer centricity strategy focuses on four levels of analytics: 1. Descriptive Analytics help identify past and current trends, for example, what percentage of customers are unhappy and may churn. 2. Diagnostic Analytics help explain why things happened, for example, what pain points made customers unhappy and caused them to churn. 3. Predictive Analytics help determine what may happen in the future based on the diagnostic analysis, for example, how likely is a customer to be happy with us and will they stay or churn. 4. Prescriptive Analytics help us understand actions we can take to affect outcomes, for example, trigger actions that might turn an unhappy customer into a happy one. I will focus primarily on the predictive customer health score machine learning model, leveraging Azure Databricks orchestration from Azure Data Factory. I’ll share how Lumen CX data science improves the customer experience and delivers better prediction outcomes with ML analytics and prescribed actions. We create a unified data and analytical platform by leveraging Azure Databricks, Delta Lake, and Data Factory, as well as tools such as Snowflake, Cosmos DB, Microsoft Dynamics, PowerBI to ingest and process customer engagement transaction data. I will share best practices developed from real-world experience when the ML model is deployed as a black box model without explainability and using the SHAP to provide the global and local explanation for the prediction. This method helps build confidence with stakeholders and the users of the ML model to make business decisions based on the predictive outcomes. I will also delve into architectural best practices on how ML models are registered with MLOps and how to perform model validation with trustworthy AI evaluation for pillars such as fairness, explainability, bias, and model drift.