The encyclopedia · Engineering & Operations · Operational decision · 2022-2024
Samsung cut bottleneck WIP 35% with data-driven targets
Samsung set optimal work-in-process levels at its fab bottlenecks; average WIP fell 35% with no loss of productivity.
Samsung Electronics
the move
In a semiconductor fab, wafers spend one to three months in process, and the bottleneck tool determines the throughput of the entire line. Too much work-in-process (WIP) at that tool ties up capital and lengthens cycle times; too little risks starving the bottleneck and losing output.
Samsung Electronics used optimisation and statistical techniques on real-world production data to find the optimal WIP level and threshold bounds for managing its bottleneck process, a target that balances utilisation against inventory.
The result was a 35% reduction in average WIP at the bottleneck process without compromising productivity, demonstrating that the right WIP target is a lever a fab can pull deliberately rather than a number that just happens.
why it works
- It targeted the one process that constrains the whole line.
- Data-driven bounds replaced trial-and-error WIP management.
- Lower WIP means shorter cycle times and less tied-up capital.
- Productivity was protected, so the gain was pure.
what transfers
For a process with long cycle times, the bottleneck's queue size is a policy choice. Compute the right target from data instead of guessing lower, and you can cut inventory without hurting throughput.
what came after
The approach was published in the INFORMS Journal on Applied Analytics (2024) by Samsung Electronics researchers and presented at the 2024 INFORMS Analytics conference.
references
- Samsung Uses Data-Driven Approach to Manage Work-in-Process of Bottlenecks in Semiconductor Manufacturing Operations
- Samsung Uses Data-Driven Approach to Manage Work-in-Process of Bottlenecks in Semiconductor Manufacturing Operations
spotted an error? The archive wants to know.