Title: Placing the Right Product at the Right Location: Studies on Inventory Placement

Date: September 26th, 2025

Time: 1:00 pm – 3:00 pm (EST)

Location: GTMI 114 and Zoom:  

GTMI 114 is located on the first floor in the MARC/Callaway Building at 813 Ferst Drive, N.W., Atlanta, GA 30332-0560. 

Upon entering the building on the main entrance off Ferst St., turn right, 114 is the first door to the right past the large TV monitor.

Katja Meuche

Industrial Engineering PhD Candidate

H. Milton Stewart School of Industrial and Systems Engineering

Georgia Institute of Technology

Committee

1 Dr. Benoit Montreuil (ISYE, Georgia Tech) (Advisor)

2 Dr. Mathieu Dahan (ISYE, Georgia Tech) (Advisor)

3 Dr. David Goldsman (ISYE, Georgia Tech)

4 Dr. Xin Chen (ISYE, Georgia Tech)

5 Dr. Cristiana Lara (Amazon)

6 Dr. Kevin Dalmeijer (ISYE, Georgia Tech)

Abstract

This thesis addresses challenges in omnichannel retail inventory placement. As customers increasingly expect faster delivery and retailers want to offer a broad product portfolio, placing the right product at the right location at the right time proves to be operationally complex and costly.

To control complexity and achieve efficiency in the fulfillment network, retailers, like Amazon, regionalized their operations in the US. This process involved defining areas, assigning fulfillment centers accordingly, and aiming to meet customer demands using the inventory of the fulfillment centers within each area.

The benefits of the approach are shorter delivery times and reduced truck usage due to shorter delivery distances as long as the right inventory is placed in each area. 

In chapter 1, we develop a framework for profit-maximizing inventory decentralization when delivery time expectations impact sales conversion by extending Newsvendor theory and analyzing 42,000 products. The large-scale empirical study results in guidelines for the profit-maximizing decentralization level based on demand and gross profit margin to support practitioners in network design.

Chapters 2 and 3 explore how dynamic customer substitution behavior can be incorporated into inventory placement models for storage-constrained regionalized fulfillment networks and what benefits substitution-aware inventory placement yields for the retailer. First, we propose six alternative inventory placement models accounting for demand uncertainty. We design a simulation-based model selection methodology to identify the inventory placement model which most accurately captures substitution behavior across products and delivery times. By running 5,566 experiments, we identify that the inventory placement model prioritizing fulfillment of the originally demanded product and using integer fulfillment variables best approximates dynamic substitution decisions with superior consistency and 1.91% average cost prediction accuracy. Building on the findings from the previous chapter, our last chapter shows that while average savings of substitution-aware inventory placement are modest (0.73%), product groups with medium substitutability, moderate demand variability, and low-to-medium demand correlation achieve exceptional savings of 10-24%.