ISE 754 (Fall 2026) 2.6 Logistics Network Design

Overview

Michael G. Kay frames logistics network design as a strategic problem and compares bottom-up analysis, which starts with known transportation rates, with top-down analysis, which infers a nominal rate from an existing network’s costs. Using Popco Bottling Company’s 42 western U.S. plants, aggregate demand, production and procurement costs, and distribution costs, he demonstrates a Julia and LogJam analysis that solves an uncapacitated facility location problem and helps identify candidate plant closures or openings. The model is a rapid first assessment, not a complete design: capacity, transportation, road networks, routing, and inventory detail matter when they differ meaningfully across locations, customers, or products.

Key takeaways

Top-down network design can estimate a nominal transportation rate from an existing network’s total distribution cost, avoiding the need to reconstruct every freight rate when the first goal is a comparative cost analysis.
Popco’s 42-plant example shows how plant locations, aggregate demand, production and procurement costs, and annual distribution costs can support an uncapacitated facility location analysis in Julia and LogJam.
A rapid initial model helps determine whether additional data is worth requesting: if plant capacity is not binding, collecting detailed capacity figures may be unnecessary; if proposed assignments exceed capacity, such as reaching 120 to 130 percent, the model needs capacity constraints.
Aggregate transportation costs and great-circle distances are useful only when service conditions are sufficiently uniform; supplying Hawaii or navigating barriers such as Manhattan’s waterways calls for more detailed transport or road-network modeling.
Point-to-point delivery assumptions can misstate costs when small orders require multi-stop routes but large customers receive full truckloads, so those distinct service patterns should be modeled separately.
The Twinkies example illustrates that product shelf life can make a logistics network unsuitable: a network designed for products lasting six months to a year may cause three-month-shelf-life bakery goods to be returned after sitting too long.

Chapters

0:00 Strategic Network Design and Bottom-Up versus Top-Down Costing

Network design is presented as a strategic logistics problem, with operational processes often systematized and network changes reconsidered when those processes fail.
Bottom-up analysis begins with a known transportation rate and combines it with other costs; top-down analysis starts from an existing network’s cost data and estimates a nominal transportation rate.
Top-down analysis requires an existing network, so it is not suited to a greenfield design; comparing cost changes also assumes network choices do not affect revenue.
The earlier three-facility example illustrates both approaches: working from known rates or working backward from annual costs can produce comparable results.

5:00 Popco’s 42 Plants and the Six-Step Top-Down Analysis

Popco Bottling Company has 42 existing plants across the western United States, many accumulated through acquisitions rather than deliberate network optimization.
The top-down model uses plant locations, aggregate demand in tons, plant-level production and procurement costs, and each plant’s annual distribution cost.
The analysis estimates fixed costs from production data, uses ZIP-code populations to weight market differences, and solves an uncapacitated facility location problem.
Julia and LogJam implement the six-step workflow using previously introduced functions; the results compare recommended locations with the existing network and suggest possible closures and a new plant.

10:00 Why a Fast Initial Model Helps Consultants Prioritize Data

Aggregate annual data can support an initial Popco analysis without requiring detailed plant-floor information, making a first response to a client feasible in less than a week.
A quick model provides a concrete basis for requesting further data and can keep a client engaged before delays or shifting priorities stall the project.
Top-down analysis can also estimate a competitor’s network or evaluate geographic expansion when publicly available information is sufficient to approximate costs.
The model’s uncapacitated assumption needs checking: if proposed facilities would handle 120 to 130 percent of available capacity, plant capacities and possible second or third shifts must be examined.

15:00 When Freight, Road Networks, Routing, and Inventory Need More Detail

For Popco, total distribution cost supports a nominal transportation rate that aggregates transport methods; a more detailed freight model is warranted when costs differ substantially, such as road service versus supplying Hawaii.
Great-circle distances are a reasonable first approximation across much of the western United States, but lakes, rivers, or Manhattan’s island geography make actual road-network distances important.
Point-to-point delivery to ZIP-code centroids can approximate average service distance, but omitting multi-stop routes can distort costs when small deliveries share routes while large customers require truckloads.
Aggregated production and procurement costs can absorb inventory costs when customers receive similar product mixes; high-value or perishable products require more detail when customer mixes differ.

20:00 Twinkies Show Why Product Shelf Life Can Require a Different Network

Kay describes Twinkies as having an approximately three-month shelf life, compared with many bakery goods that lose value within days.
The example contrasts a distribution system serving convenience stores and managing Twinkie spoilage with a broader network designed for products lasting six months to a year.
Sending Twinkies through a network not designed around their shorter shelf life can lead to product sitting on shelves for five or six months and being returned.
The case reinforces the modeling rule: aggregate assumptions can work when conditions are uniform, but meaningful differences in shelf life, customer mix, or delivery requirements must be represented.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Michael G. Kay.