[From iSource Business, June/July 2002] June and July may seem like a strange time to think about winter colds, but the flu offers a singular example of why supply chain optimization is necessary. Even as the summer sun prods temperatures upward and pushes thoughts of last season's colds out of mind, pharmaceutical companies are already hard at work on vaccines for the coming winter's season of sniffles. To ensure they have the right vaccines in sufficient quantities, the drug companies must anticipate which strains of the flu virus will hit the public next winter and gear up their supplies and their suppliers accordingly. Of course, if the virus mutates, the new flu that emerges could render an old vaccine useless and put a pharmaceutical firm's bottom line at risk, unless the company can readjust its supply network rapidly to fulfill the new demand.
This scenario likely will be familiar to any number of manufacturers that have tried to use supply chain optimization to cope with demand fluctuations and the frequent mismatches between forecasts and actual orders. Traditional optimization has relied on a centralized linear programming approach to model supply and demand planning and optimize the throughput of a factory floor. Using such systems, companies have generally allocated individual inventory buffers to each of their customers to absorb the variability of a particular customer's demand
The problem with the centralized approach to optimization, however, is that most companies participate in multiple supply chains and operate under manifold business models, with different customers of different sizes dictating different rules for the ways in which they want suppliers to meet their demand. As a result, suppliers usually wind up striving to maintain sufficient inventory for each customer in isolation, based on that customer's business requirements. The result is high inventory exposure.
The Virtual Inventory Buffer
To address this shortcoming, the next generation of supply chain solutions is moving beyond the centralized linear model to an adaptive, distributed and, therefore, decentralized and flexible framework. This allows a company to aggregate demand and variability across all of its customers and manage a virtual inventory buffer that spans the customer base.
For example, a semiconductor company might have very little flexibility in managing its capacity because building a new fab is such an expensive proposition. Meanwhile, the company might want to hedge the financial risks of demand fluctuations by targeting multiple verticals for its products. By employing the adaptive and distributed model, which constantly monitors incoming demand from all the company's customers, reallocating capacity on the fly, the company can satisfy its current customers' demand while reducing its overall inventory levels. This is good for the company itself, since it is better meeting its customers' requirements. However, it's also good for the company's own suppliers, who will be less subject to radical fluctuations in the manufacturer's demand for components and therefore able to reduce their own inventory position and liabilities.
This type of dynamic optimization has the capacity to move industries away from the zero-sum game of just-in-time (JIT) arrangements, where a buyer's JIT system depends on inventory maintained by suppliers. They shift to a win-win, where inventory exposure is reduced across the supply chain and there is greater flexibility in meeting volatile demand. Such optimization is not for everyone, of course. In industries where demand forecasts are stable and products have long lifecycles, such dynamic optimization would be overkill. Moreover, only a handful of companies currently are looking at this new model. But where forecasts are volatile, the next generation of optimization solutions will provide a competitive edge to those companies willing to invest in the systems and solutions necessary to optimize their supply chains on the fly.
Dr. Cosimo Spera is an expert in the fields of operations research and advanced mathematical modeling, having taught at University of California at Berkeley, Columbia University and the University of Michigan. He is co-founder and CEO of Saltare, a provider of adaptive supply chain management applications.