The encyclopedia · Engineering & Operations · Operational decision · 1994–2000
IBM modeled its whole extended supply chain, saving $750M in 1998.
IBM built AMT, a tool that models the whole extended supply chain with optimisation and simulation, then cut $750M in costs in 1998.
IBM
the move
IBM sells through its own divisions and a web of channel partners, so responsiveness with minimal inventory is both valuable and hard to judge. In 1994 it began re-engineering the global supply chain.
The usual move is to improve each plant's plan. IBM instead built the Asset Management Tool, an extended-enterprise supply-chain analysis tool that models the whole chain, not just one node.
AMT wires graphical process modeling, analytical performance optimization, simulation, activity-based costing and enterprise database connectivity together, so IBM could test inventory budgets, turnover objectives, customer-service targets and new-product launches quantitatively.
Rolled out at a number of IBM business units and their channel partners, AMT saved over $750 million in material costs and price-protection expenses in 1998 alone.
why it works
- Decisions at one node ripple nationwide, so a whole-chain model is required
- Coupling optimisation to activity-based costing makes the dollars visible
- Simulation lets IBM test inventory and service trade-offs before committing
- Modelling channel partners, not just IBM plants, captures the real system
what transfers
When a chain spans companies, a shared model beats separate plans; linking optimisation to costing makes the trade-off visible.
what came after
AMT was implemented at several IBM business units and their channel partners, saving over $750 million in material costs and price-protection expenses in 1998. The work was published in Interfaces in 2000 and became a Franz Edelman Award finalist.
references
- Extended-Enterprise Supply-Chain Management at IBM Personal Systems Group and Other Divisions
- Extended-enterprise supply-chain management at IBM personal systems group and other divisions
spotted an error? The archive wants to know.