EN
Back to the archive

The encyclopedia · Strategy & Leadership · Operational decision · 2001–2006

Caterpillar's BCPD cut order-to-delivery time 20% with one multiechelon stock model

A multiechelon inventory model set dynamic stock policies across plant and suppliers, cutting delivery time 20%, its variance 50%, and inventory 16%.

Caterpillar Inc.

the move

In 2001 Caterpillar's Building Construction Products Division (BCPD) faced order-to-delivery times that were both long and erratic. Dealers could not promise dates, so customers drifted to competitors with steadier availability.

Rather than adding inventory everywhere, the division ran a six-sigma project and built a multiechelon inventory optimization model. It set dynamic stocking policies for assembled machines at the plant and for subassemblies and components at the plant and at suppliers.

The results were unusually clean: mean delivery time for orders with identified customers fell 20%, its standard deviation fell 50%, total chain inventory fell 16%, and BCPD won more orders, raising revenue about 2%.

why it works

  • Modeling the whole echelon chain shows where variance enters, so buffer stock is placed at the right level instead of everywhere.
  • Dynamic policies are recomputed as demand and supply shift, so the system keeps absorbing shocks rather than freezing into stale targets.
  • Improving variance as well as mean lets the division promise reliable dates, which is what wins orders.
the payoffOptimize inventory as one network, not per plantclever

what transfers

Availability problems are usually inventory-position problems: set stock levels jointly across echelons against real variance, and both speed and inventory improve instead of trading off.

what came after

The initial implementation paid off in faster, steadier delivery and a 16% inventory cut, and other Caterpillar manufacturing and assembly facilities replicated the approach across their own supply chains. The work became a reference case for multiechelon inventory optimization in industry practice.

references

spotted an error? The archive wants to know.

same kind of clever