The challenge of demand forecasting varies greatly by company and industry. Perhaps in no industry is the challenge as difficult as in the semiconductor industry, which has to deal with the double challenge of fluctuating customer demand as well as cycle times that are longer than they are in many other industries. To ensure adequate supply, many companies stock up on excess inventory and manufacturing capacity, which can prove lethal to profitability and productivity.

One company doing things a better way is ON Semiconductor of Phoenix, Ariz., which has seven manufacturing locations worldwide and a number of subcontracting partners. The strategy ON Semiconductor has implemented is called demand sensing and demand shaping for enterprise profitability. It is based on AMR Research's Demand-Driven Supply Network (DDSN) concept. The three core elements of DDSN are demand sensing, demand shaping and profitable demand response.

According to John Mallon, director of supply chain solutions for ON Semiconductor, demand sensing falls into two timeframes or "horizons." One is the execution horizon or what the company can actually build in its factories. "We have a greater challenge in forecasting than companies in other industries, because cycle times in our industry tend to be long," Mallon says. "They could be anywhere from five to thirteen weeks. As such, we have to do a much better job of understanding demand than companies in some other industries."

The timeframe beyond the execution horizon is defined as the forecasting period, which focuses on capacity planning and making sure the company has the right amount of capacity for the future.

"It is very difficult to add capacity in a current quarter," explains Mallon. "We continue to try to improve our forecasting methods." Like a lot of other companies in the industry, ON Semiconductor is working to improve the process of passing upstream information down through the supply base, with as few time lags as possible.

"We do this by trying to get more direct forecast information from our customers," he states.

This provides much-needed information that can be fed back into the execution horizon. "On the execution horizon again, if we get good information from customers, we can start to understand customer demand patterns, such as how variable their demand is and how much it normally deviates over time," Mallon tells Purchasing.

This is where inventory planning tools come into play by analyzing demand variability patterns, such as order leadtime and supply chain data such as cycle time and yields. This provides the information necessary to determine what needs to be held in safety stock and at what stage of manufacturing this stock needs to be held (eg: component level, subassembly, or finished item). "These tools have helped us drive up our service levels, and do so in an inventory-efficient way," says Mallon.

The next phase, demand shaping, focuses on selling what ON Semiconductor already has available. "Some of the advanced planning tools that have been introduced in recent years are very good at helping us understand the inventory and capacity that is available," continues Mallon. The company can then shape the demand to some degree, based on how inventory and capacity are allocated.

The third phase, profitable demand response, involves corporate dynamic pricing that includes supply constraints, corporate planning that allocates critical/constrained parts as a way to optimize service for some segments and maximize profits for others, and multilevel inventory analysis and disposition.