Predictive Machine Learning Model for Customer Purchase and Lapse Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Financial services entities face challenges in identifying customers ready to purchase additional products and determining suitable products to offer, as well as predicting policyholder lapse likelihood, which is labor-intensive and time-consuming.
Innovation Solution
A processor-based method using a predictive machine learning model that analyzes customer purchase history and profile data to classify customers into target and non-target groups for product offerings and predicts lapse likelihood, utilizing regression models and gradient boosting for accurate product recommendations and retention strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to identify customers ready to purchase additional products through needs analysis, then product recommendations can be made, but the process requires customers to answer dozens of questions which is time-consuming and labor-intensive
Solution Approach 1:
The system performs preliminary analysis by continuously monitoring customer transaction data, profile information, and behavioral patterns in advance. Machine learning models pre-calculate purchase likelihood scores and product affinity metrics before a needs analysis is triggered, so when identification is required, the system can quickly retrieve pre-computed insights rather than gathering all data from scratch during the interaction
Solution Approach 2:
Instead of directly analyzing raw transaction data and customer profiles during the needs analysis process, the system creates simplified copies or representations in the form of pre-computed customer segments, purchase propensity scores, and product recommendation rankings. These copied insights are then used to rapidly identify ready-to-purchase customers without re-processing the entire data set
2Adaptability or versatility
If a broad product offering is provided to customers, then more product options are available, but it becomes difficult to navigate and identify suitable products for existing customers
Solution Approach 1:
The system applies local quality by providing different product recommendations and information to different customer segments based on their specific characteristics, purchase history, and inferred needs. Rather than presenting the same broad catalog to all customers, the system tailors the product subset and presentation style to each customer's profile, making the overwhelming broad offering manageable through personalized filtering and prioritization
Solution Approach 2:
The machine learning model serves multiple functions simultaneously: it segments customers, ranks products, predicts purchase likelihood, and generates personalized recommendations all in one unified system. This multi-functional approach allows the system to handle the complexity of broad product offerings while maintaining ease of use across different customer types and scenarios
3Measurement precision
If manual methods are used to determine policyholder lapse likelihood and value, then assessments can be made, but the process is labor-intensive and time-consuming
Solution Approach 1:
The system implements self-service by automatically monitoring payment patterns, policy characteristics, and customer behavioral data to compute lapse likelihood scores without human intervention. The machine learning models continuously self-update as new data becomes available, and the system autonomously identifies at-risk policies and computes retention recommendations, eliminating the need for manual assessment while maintaining high accuracy through automated data processing
Data Source
AI summary
A processor-based system and method retrieve customer purchase history information from an internal customer purchase history database for a plurality of customer records representing customers that previously purchased products of an enterprise, and retrieve customer profile information for each customer record. The processor executes a predictive machine learning model to determine a set of product purchase scores for each of the customers by applying a logistic regression model utilizing gradient boosting to the customer purchase history information and the customer profile information. The processor classifies the customers into a target customer group and a non-target customer group by applying a classification criterion to the set of product purchase scores, and generates a report of customers in the target customer group including highest product purchase scores and products recommended for cross-sale. In some embodiments, the predictive machine learning model is configured to forecast likelihood that given customers will lapse in payment.