EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

2704. Enhancing B2B customer retention: a streamlined profit-driven approach using fairness-inspired pre-processing techniques

Invited abstract in session TA-28: Fairness and responsible AI, stream Advancements of OR-analytics in statistics, machine learning and data science.

Tuesday, 8:30-10:00
Room: 065 (building: 208)

Authors (first author is the speaker)

1. Shimanto Rahman
Economics and Business Administration, Ghent University
2. Bram Janssens
Ghent University
3. Matthias Bogaert
Marketing, Innovation and Organization, Ghent University

Abstract

In the realm of Customer Relationship Management, customer retention is pivotal for sustained growth. The positive correlation between customer satisfaction, retention, and profitability guides traditional Customer Churn Prediction models to rank high-value customers lower on average than their low-value counterparts. Ranking based on churn propensity without considering the profitability of retaining specific customers may lead to suboptimal decision-making. This is especially pronounced in Business-to-Business (B2B) industries. Various profit-driven metrics and algorithms address this, but their integration and extension demands considerable effort, hindering widespread adoption.

This paper proposes a streamlined approach, incorporating pre-processing techniques inspired by fairness literature, including Massaging, Reweighting, and Resampling. These techniques rectify the underrepresentation of high-value customers within the targeted group, improving churn campaign profitability. Preliminary results on two B2B datasets show these techniques outperforming baselines.

The research contributes by proposing techniques to amplify the often-underrepresented high-value customer fraction within the targeted group. Framing the challenge as one involving the sensitivity of high-value customers provides a simpler interface for improving churn campaign profit on par with existing profit-driven techniques.

Keywords

Status: accepted

Back to the list of papers