Article Details
  • Published Online:
    July  2026
  • Product Name:
    The IUP Journal of Computer Sciences
  • Product Type:
    Article
  • Product Code:
    IJCS030726
  • DOI:
    10.71329/IUPJCS/2026.20.3.31-42
  • Author Name:
    Kaushik Bar
  • Availability:
    YES
  • Subject/Domain:
    Engineering
  • Download Format:
    PDF
  • Pages:
    31-42
Volume 20, Issue 3, July-September 2026
Counterfactual Demand Forecasting for Retail Promotions Using Bayesian Causal Forests
Abstract

The paper proposes a Bayesian causal forest (BCF) framework to estimate the heterogeneous treatment effects (HTEs) of promotions, enabling counterfactual demand forecasting that is both personalized and uncertainty-aware. Unlike deterministic machine learning (ML) models or fixed-effects regressions, the approach leverages a Bayesian treatment of uncertainty, allowing for credible interval estimation around individual uplift predictions. The model is benchmarked using three publicly available datasets (RetailHero sales uplift dataset, Rossmann store sales dataset, and Criteo uplift prediction dataset) against widely used baselines, including S-learner, T-learner, X-learner, causal forests, and standard demand forecasting models. The results show that the proposed BCF model consistently outperforms existing methods in uplift accuracy, Qini coefficient, and uncertainty calibration error, while also providing interpretable insights into segment-level responsiveness to marketing interventions. This study demonstrates the feasibility and value of integrating Bayesian causal inference into business analytics pipelines for targeted promotional optimization.

Introduction

In retail, promotional campaigns are pivotal levers for boosting short-term sales and influencing long-term customer behavior.