Packaging optimization at Atlas Copco

Packaging is one of those costs that never gets measured properly. It sits between manufacturing and logistics, it is decided in a dozen different places, and nobody owns the total. Atlas Copco’s young talent programme gave me a two-month window to find out what the packaging policy actually was, and then a consulting stint to follow through.

Documenting. The first job was to establish the reality: which packaging configurations were used for which items, and what the existing policy was in practice rather than in theory. That baseline is the only thing that makes an optimization result mean anything, and it was more work than the modeling.

Optimizing. With the baseline in hand, I built a mixed-integer model that assigns a packaging option to each item, subject to the physical and commercial constraints each option implies. The interesting part was not the formulation but the constraints themselves: they came from weekly meetings with people in different departments, each of whom knew about a restriction that did not appear in any specification. Packaging design candidates were then generated with metaheuristics, such as simulated annealing, rather than enumerated by hand.

The objective was to cut annual distribution and storage costs, currently dominated by the volume and the number of handling steps, by roughly half, and the engagement ended with an implementation plan rather than a slide deck.