Skip to content

Supply chain analytics

Project done at MuSigma (Client: Walmart)

Technologies used: R, Python, SQL, CPLEX, Google Cloud Platform Timeline : 2018 - April 2020

Problem Statement: Reducing out-of-stock scenarios in stores by identifying and quantifying the different factors, predicting the failures due to various factors, and optimizing inventory based on them.

Team: Collaborated with the technology and business units of Walmart Supply chain and market POC's in the US, Canada, Mexico, Argentina, and Chile. Worked end-to-end from ideation to POC development to production

Quantified savings: The potential average cost savings from a reduction in inventory and out-of-stock costs would be $12 Million per month

Quantify the reasons for under-stock scenarios

  • Quantified the reasons causing under-stock scenarios in a store utilizing hypothesis testing and statistical modeling pinpointing the two main factors among 14 with the most significant impact (fill rate and lead time).

Identify the risk of a supplier not delivering an order

  • Designed classification model (gradient boosting) predicting the risk of a supplier not delivering an order in full (fill rate) with 75% accuracy and 50+% specificity
  • Deployed the solution on the cloud and created workflows to predict the risk daily

Forecasting inbound lead time of vendors

  • Forecasted lead time applying a tree-based ensemble regression model (random forest) with 85% (SMAPE) accuracy
  • Deployed the solution on the cloud and created workflows to predict lead time daily

Optimizing inventory at store and warehouse

  • Optimized EOQ and reorder point using an integer programming model
  • Formulated and validated the approach under the Senior Director of Supply chain at Walmart
Back to top