Supply chain analytics¶

Project Title: Supply Chain Analytics for reducing understocking
Client: Walmart (via Mu Sigma)
Timeline: 2018 – April 2020
Location: LATAM, Global (US, Canada, Mexico, Argentina, Chile)
Team Size: Cross-functional collaboration with Walmart’s tech and business units
Technologies: R, Python, SQL, CPLEX, Google Cloud Platform

Problem Statement¶

Walmart faced frequent out-of-stock scenarios across its global retail network, leading to lost sales and customer dissatisfaction. The objective was to identify and quantify the root causes of understocking, predict supply chain failures, and optimize inventory levels across stores and warehouses.

Solution¶

The project involved a multi-pronged approach:

Root Cause Analysis: Used hypothesis testing and statistical modeling to identify key drivers of understocking, pinpointing fill rate and lead time as the most impactful among 14 factors.
Predictive Modeling: Built a gradient boosting classifier to predict supplier delivery risks (fill rate) with 75% accuracy and over 50% specificity. Also developed a random forest regression model to forecast vendor lead times with 85% SMAPE accuracy.
Optimization: Applied integer programming and Monte Carlo simulations to suggest optimised Economic Order Quantity (EOQ) and reorder points at both store and warehouse levels based on modified lead time and fill rates.
Deployment: All models and workflows were deployed on the cloud to enable daily risk and lead time predictions.

Impact¶

The solution had the potential to save Walmart an estimated $12 million per month by reducing inventory costs and out-of-stock losses. It also improved supply chain visibility and decision-making across multiple geographies.