Like so many manufacturers, ON Semiconductor Corp. has to make big bets on capacity and then wait to see if demand matches up. To mitigate the risks, ON Semiconductor implemented Vivecon Corp.'s predictive-analytics software to help it more accurately decide when to increase capacity and when to pull back.

This month, ON Semiconductor put the new software to the test, making the first of eight decisions in the next two months on whether to add more capacity internally so it can produce and package products or outsource to another manufacturer. The analysis led it to choose to do both.

Making semiconductors is a complex process. Pieces of silicon chips will fly across the ocean two or three times before they become finished products. "So we need sophisticated software to drive cost competitiveness in our market," says Sal Barlett, director of operations at ON Semiconductor's integrated power group.

Vivecon's Launch Manager is able to recommend the best possible amount of production capacity for wafers. It tracks production that's outsourced and keeps tabs on the supply chain. Ultimately, ON Semiconductor will use its monitoring capabilities to get an early warning of future capacity shortfalls, Barlett says. Currently the company looks at capital investments in capacity on a quarterly basis. But with the software in place, it's trying to monitor this more frequently so it can quickly pick up changes in ordering patterns.

In 2004, the $1.3 billion-a-year ON Semiconductor worked to cut capital costs by decreasing inventory levels and refinancing high-interest bonds. Vivecon's software is expected to further cut capital costs 9% to 15%, Barlett says.

Predictive analytics can do more than save money. Says Noha Tohamy, an analyst with Forrester Research: "Using predictive analytics to simulate possible events--such as a port shutdown or a supplier's failure to meet its contractual commitments--and understand their impact on supply chains is an essential tool."