Improving Supply and Inventory Management through Demand Planning Tools and Processes
The first step in our project was conducting stakeholder interviews. These interviews provided insight as to the demand planning methods currently used within Stanford Healthcare and helped us to identify areas of opportunity. We conducted over 10 interviews and met with individuals from each department within the Supply Chain: Distribution & Logistics, Data & Analytics, Procurement Strategy & Ops, and Business Transformation. We found that all teams faced two primary issues: (a) there is currently little insight into expected future demand or supply of medical supplies, and (b) the processes in place to address fluctuations in demand or supply are still being improved.
Our project also aimed at understanding supply chain best practices that currently exist within healthcare and analogous industries. We met with Brent Johnson, the creator of Intermountain Healthcare’s award winning supply chain, and discussed what makes a supply chain successful and ways to implement demand planning.
We also were able to tour Stanford Healthcare’s supply chain both in the Stanford Adult Hospital and the Stanford Children’s Hospital. This gave us valuable insight into the day-to-day operations as well as a better understanding of the path a product takes from entering Stanford’s supply chain to reaching the end user.
We explored the data through Excel visualizations focusing on the areas that experienced shortages the most: masks, gloves, and surgical drains. Through creating several time series graphs, we saw periodicity in demand, demand increasing before shortages, and numerous negative orders to account for ordering errors.
Using three years of demand data, we evaluated several potential forecasting models against a baseline rolling average forecast. We found that the Prophet package—an open source forecasting tool developed by researchers at Facebook—provided both reasonably accurate forecasts, since it accounts for seasonality factors that heavily influence demand, and useful tools such as customizable confidence intervals and robust graphing packages.
We then created an R script that can be easily run by the Stanford Supply Chain’s Analytics Team so that they can maintain our forecasting models in the future. The current version of our tool includes options to forecast by UNSPSC commodity or by item, and to forecast daily or weekly demand.