P&G modeled its whole North American supply network and cut plants ~20%

P&G made and marketed over 300 brands in 70+ countries, and its North American sourcing and distribution had duplicated plants, warehouses and overhead. Managers in each market naturally argued for a plant of their own, so the true cost of the network was invisible.

In 1993 the SGE effort launched the North American product supply study: about 500 employees reexamined sourcing and distribution, and an OR team built models combining integer programming, network optimization, GIS and LINDO. The models made millions of options comparable, and the recommendation was to consolidate North American plants by almost 20 percent.

The network view worked because it exposed scale economies and freight tradeoffs no local decision could see. The restructuring was saving over $200 million a year in pretax costs when the 1997 Interfaces paper was published; LINDO's own write-up put the savings at well over $250 million a year.

Optimize the network, not each plant

Whole-network optimization exposed duplication no local view could see

GIS made geographic tradeoffs concrete for executives

The study planted P&G's permanent OR group, IT Global Analytics

Local intuition says keep a plant here; an end-to-end network model sees scale economies and freight that local views miss. Make the whole system the unit of analysis.

The 1997 Interfaces paper became a Franz Edelman Award finalist. P&G kept a 17-person OR group running about 100 projects a year on a target of returning 10 times its own salary cost.