Next Best Action — Optimizing Long-term Marketing Communications using Markov Decision Process

A reinforcement learning based approach for learning optimal policies (Part 2/3)

Generated by author using Bing Image Creator

Part 1: Generating synthetic customer journeys for mortgage payments.
Part 2: Next Best Action — Optimizing Long-term Marketing Communications using Markov Decision Process
Part 3: Next Best Action — Learn an optimal policy for maximizing mortgage collection

Welcome to Part 2 of the Next Best Action series. Curious about the cover picture? I chose it because the Next Best Action paradigm originated in military context.

Personalization, in customer-facing businesses, is often cited as having one single goal — tailoring the customer’s experience based on their needs. In many applications though, personalization has a second (sometimes dominant) objective — increasing revenues or inversely cutting losses over a longer timeframe.

Over longer timeframes (I’m talking years), customers go through complex journeys, their maturity increases over time, and their needs and circumstances change from time to time. Some of those changes are positive, in that, they may buy more products or extend existing ones with your company. Others are…