December 31, 2023

Understanding Discrepancy Between Design and Implementation of Inventory Policies

This project investigated disparities between designed multi-echelon inventory policies and their implementation and developed ways to mitigate them by improving policy design input or providing tools for better execution.

Key Outcomes
Completion Date
December 31, 2023
Project Awarded:

May 1, 2022

Project Number:

21-10-06

Project Lead

Rochester Institute of Technology
The DOW Chemical Company

Problem

Designed inventory policies are not always closely followed, and the reason for discrepancies is typically not obvious. For example, what information gaps or other factors will cause an inventory manager to deviate from the policy and lead to over- or under-ordering? Identifying those issues and ensuring that needed information is spelled out can improve policy design and implementation.

Proposed Solution

Starting with case studies and real (anonymized) data, the team from the Rochester Institute of Technology and DOW developed a framework for evidence-based answers about the causes of observed discrepancies.

Impact

A more streamlined relationship between designed inventory policies and their implementation will unlock the potential for better, more realistic optimization.

Outcome

The project team developed a framework for analyzing inventory data that can detect irregularities and changes in inventory behavior as well as less-than-optimal inventory management, such as over- and under-ordering. The framework was implemented in Python and delivered as a single class within a JupyterLab notebook. The notebook illustrated some typical uses of the framework, including detecting sub-optimal inventory management.

The data used was anonymized and delivered along with the notebook so researchers could reproduce and extend the results.



Fig.1 Screenshot of the main interface of the inventory analyses tool.


Fig.2 Inventory balance (top) and computed service-level metric, with over- and under-ordering limits (bottom).


Fig.3 Classifying inventory types based on short-term observations (30-day segments).