Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Machine learning and Empirical Bayesian Approach for Predictive Buying in B2B E-commerce

Tuhin Subhra De, Pranjal Singh, Alok Patel

TL;DR

The paper addresses predicting buyer orders in B2B e-commerce by merging a short-horizon ML predictor (XGBoost) with an empirical Bayesian model (MPG) that captures repeat purchasing dynamics. A stacking framework with logistic regression combines outputs, yielding superior performance (AUC-ROC up to $0.8753$; AUC-PR up to $0.4262$) and a real-world conversion uplift (from $4.19\%$ manual to $13.39\%$ with the stacked model) over a three-week telecalling pilot. The approach demonstrates substantial business value by efficiently targeting high-potential buyers and improving resource utilization for telecallers. It also highlights future directions, including customer-product level predictions and richer data sources for real-time decision support.

Abstract

In the context of developing nations like India, traditional business to business (B2B) commerce heavily relies on the establishment of robust relationships, trust, and credit arrangements between buyers and sellers. Consequently, ecommerce enterprises frequently. Established in 2016 with a vision to revolutionize trade in India through technology, Udaan is the countrys largest business to business ecommerce platform. Udaan operates across diverse product categories, including lifestyle, electronics, home and employ telecallers to cultivate buyer relationships, streamline order placement procedures, and promote special promotions. The accurate anticipation of buyer order placement behavior emerges as a pivotal factor for attaining sustainable growth, heightening competitiveness, and optimizing the efficiency of these telecallers. To address this challenge, we have employed an ensemble approach comprising XGBoost and a modified version of Poisson Gamma model to predict customer order patterns with precision. This paper provides an in-depth exploration of the strategic fusion of machine learning and an empirical Bayesian approach, bolstered by the judicious selection of pertinent features. This innovative approach has yielded a remarkable 3 times increase in customer order rates, show casing its potential for transformative impact in the ecommerce industry.

A Machine learning and Empirical Bayesian Approach for Predictive Buying in B2B E-commerce

TL;DR

The paper addresses predicting buyer orders in B2B e-commerce by merging a short-horizon ML predictor (XGBoost) with an empirical Bayesian model (MPG) that captures repeat purchasing dynamics. A stacking framework with logistic regression combines outputs, yielding superior performance (AUC-ROC up to ; AUC-PR up to ) and a real-world conversion uplift (from manual to with the stacked model) over a three-week telecalling pilot. The approach demonstrates substantial business value by efficiently targeting high-potential buyers and improving resource utilization for telecallers. It also highlights future directions, including customer-product level predictions and richer data sources for real-time decision support.

Abstract

In the context of developing nations like India, traditional business to business (B2B) commerce heavily relies on the establishment of robust relationships, trust, and credit arrangements between buyers and sellers. Consequently, ecommerce enterprises frequently. Established in 2016 with a vision to revolutionize trade in India through technology, Udaan is the countrys largest business to business ecommerce platform. Udaan operates across diverse product categories, including lifestyle, electronics, home and employ telecallers to cultivate buyer relationships, streamline order placement procedures, and promote special promotions. The accurate anticipation of buyer order placement behavior emerges as a pivotal factor for attaining sustainable growth, heightening competitiveness, and optimizing the efficiency of these telecallers. To address this challenge, we have employed an ensemble approach comprising XGBoost and a modified version of Poisson Gamma model to predict customer order patterns with precision. This paper provides an in-depth exploration of the strategic fusion of machine learning and an empirical Bayesian approach, bolstered by the judicious selection of pertinent features. This innovative approach has yielded a remarkable 3 times increase in customer order rates, show casing its potential for transformative impact in the ecommerce industry.
Paper Structure (14 sections, 12 equations, 6 figures, 5 tables)

This paper contains 14 sections, 12 equations, 6 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Proposed Approach
  3. XGBoost Approach
  4. Poisson Gamma Model
  5. Modified Poisson Gamma model (MPG)
  6. XGBoost + MPG approach
  7. Results
  8. Feature Dataset
  9. Standalone XGBoost performance
  10. Stack model performance
  11. Comparative studies
  12. Theoretical comparison
  13. Comparison by applying in real-time business
  14. Conclusion

Figures (6)

  • Figure 1: Flow of data in XGBoost model with $n$ trees. The next tree is trained on the residuals of the previous tree, in other way it corrects the mistakes made by the previous one.
  • Figure 2: Stacking XGBoost and MPG model with Logistic regression as meta learner
  • Figure 3: Top 15 input features in decreasing order of importance obtained from standalone XGBoost prediction.
  • Figure 4: The AUC-ROC and AUC-PR of XGBoost model
  • Figure 5: The AUC-ROC and AUC-PR of Stacked model
  • ...and 1 more figures