Data-driven Demand Forecasting and Assortment Optimization for Indian Omnichannel Retai...

Objective: To develop and validate a data-driven model for accurate demand forecasting and optimal assortment planning across online and offline retail channels in India.

Problem & Context: Indian omnichannel retailers face significant challenges in predicting product demand and optimizing assortment due to diverse consumer preferences, regional seasonality, fragmented supply chains, and the rapid growth of both online and offline sales channels. Inaccurate forecasts lead to overstocking, stockouts, and lost sales opportunities, especially in high-SKU environments like fashion, grocery, and consumer electronics.

Approach & Methodology: The project involves collecting and analyzing historical sales, inventory, and promotional data from both online and offline stores. Students will apply time-series forecasting, machine learning (e.g., ARIMA, Prophet, Random Forest), and clustering techniques for demand prediction. Assortment planning will use optimization frameworks (e.g., Mixed Integer Programming) to recommend SKU mixes per channel and location based on forecasted demand and operational constraints.

Deliverables & Analysis: Key outputs include a cleaned and visualized dataset, a validated forecasting model, optimized assortment recommendations for at least two categories, a Power BI dashboard for real-time monitoring, and a business report quantifying potential inventory cost reduction and sales uplift.

Business Impact & Decision: The project informs critical supply chain and merchandising decisions, enabling Indian retailers to minimize lost sales, reduce excess inventory, and tailor assortments to local demand, thus improving profitability and customer satisfaction in a highly competitive market.