Building Scalable Supply Chain Through Inventory Optimization


  • CASE STUDIES
  • June 24th, 2021
  •   17280 Views

Overview


While pinpointing demand accurately is imperative for developing a scalable supply chain, forecasting demand has become even more complex for businesses in the last few years due to the increasing uncertainties and unpredictability in the demand trends.

The inaccurate prediction of demand often leads to a sharp rise in the inventory costs. Additionally, the capital tied up due to overstocking can lead to opportunity loss as the company could use this capital towards other verticals – like marketing, R&D, etc.

Businesses are constantly looking for ways to reduce excess inventory in the warehouse to better optimize the supply chain. Therefore, companies are leveraging advanced analytical techniques to bridge the gap between optimizing inventory space and reducing stockouts. Mu Sigma worked with the Innovation and Supply team of a leading energy company to explore ways to optimize their inventory management.

The Problem

The supply chain team of one of the leading Fortune 500 energy companies approached Mu Sigma to provide data-driven solution for the problem of overstocking materials due to uncertainty in demand in supply chain.

The client’s inventory levels were based on historic demand data and was dynamically changing throughout the year to meet fluctuating demand over time. As a result, the overstock drove increased storage and handling cost, excess purchasing cost, and wastage of warehouse space.

The Supply Chain team of the energy giant was looking to reduce stockouts of in-demand materials and optimize the inventory of rarely used materials that occupy space.

Current State

•   Buyers are responsible for purchasing materials and maintaining inventory
•   Volume of requests from Outage Management depend on the weather, human behavior, and are more volatile than planned maintenance
•   Material Requests (MR) are submitted for each project to make the Supply team aware of the requirements

Gap

•   Absence of a data-driven process to accurately anticipate non-emergent as well as emergent demand leading to overstocking in the inventory

Desired State

•   Reduction in excess inventory and associated strain on the warehouse by estimating demand more accurately and providing a dynamic min-value for materials
•   Improved min-value to reduce excess inventory in warehouse

The Solution

Mu Sigma started addressing the problem by understanding the pain points of material handlers and analysts and captured information from inventory, purchase order, material request, and material issuance data by using KPIs like excess inventory, the number of days the demand was satisfied, potential savings, etc.

The analysis of historical data on material utilization and demand shows that the min-value consistently overpredicts the actual demand leading to potential overstocking.

Redefining the Business Goal

While the business goal before defining the problem space was to accurately forecast future demand and thereby help the business avoid stock-outs, once the Mu Sigma team analyzed the data trends, it was deduced that the cost of overstocking incurs more loss than the stock-outs.

Thus, the business goal was redefined to predict the optimal inventory size and thereby reduce overstocking.

Solution Design

The clients wanted to prioritize targeting inefficiencies in their Materials Management process.
For developing the Forecasting Model, Mu Sigma and the business stakeholders aligned on 15 materials from a universe of over 10,000 materials. These were the materials that went out of stock very quickly.
•  A master dataset was created by including features from the varied data sources, based on which several models were implemented and the best performing model for each case was selected
•  An intelligent modelling algorithm capable of running an ensemble of 150+ combinations of time series techniques like VAR and prophet models was created
•  Instances where a forecast model is not recommended were identified
•  This algorithm then selected the best model to predict the optimal inventory levels for each material stored at each facility

The forecasting model learns from various demand and weather variables.

The Impact

Through our out-collaboration initiatives with the client, we were able to design a demand forecasting solution that helps reduce the stockouts and optimizes the excess inventory and is projected to save $15 – 20 MM annually for limited products projected to save over $100 MM when scaled to other materials.