It's widely acknowledged that among the best ways to enhance supply chain performance is to improve forecasting accuracy. But how to improve forecasting accuracy—that's a different matter entirely.

One approach with great promise is Collaborative Planning, Forecasting, and Replenishment (CPFR). CPFR is easy to explain but incredibly hard to execute. The idea: Trading partners throughout the supply chain work together to determine inventory and replenishment strategies and make efforts to align their sales forecasts accordingly. If successful, this collaborative initiative should lead to lower inventory throughout the supply chain, more inventory turns, fewer stock-outs, and higher profitability. CPFR is heralded by the Voluntary Interindustry Commerce Standards Association, a group that includes some of the most prominent businesses in the world.

The intent of CPFR—to enable information to replace inventory throughout the supply chain—is more than just laudable. It is quite arguably necessary if supply chain management strategies are to keep contributing to the productivity improvements that have been at the heart of the growth-without-inflation prosperity of the last decade.

CPFR is just one of a number of collaborative initiatives under way to improve supply chain performance and forecast accuracy. Supporting these initiatives is a wealth of new technology. Numerous supply chain software solutions available today promise to give trading partners instant connectivity and product visibility across the pipeline. And recently, a spate of Internet-enabled solutions have been introduced that claim to accomplish the same end (and at a lower cost, the ads say).

But there's a cautionary note that needs to be sounded here. Many companies—and the supply chain professionals within them—have become enamored of the promise of the new technology. But in all too many cases, technology is viewed as a panacea. If you can just get that latest software package up and running, the thinking goes, you can effectively collaborate up and down the supply chain.

But no initiative like CPFR and its supporting technology—no matter how well developed and no matter the computing power brought to bear—will ever get forecasts exactly right. Safety stocks, to some extent, are here to stay. The reason is that inventory systems are sufficiently complex that they are susceptible to large fluctuations brought on by small changes in those systems. They are, in other words, subject to chaos, as described by the modern scientific and mathematical chaos theory.

I had occasion to think about this lesson earlier this winter when the first snow storm of the season crossed southern New England.

When I heard that snow was imminent, I visited a local hardware store to pick up an additional snow shovel. Unfortunately, the store had sold out of shovels within hours after news of the impending storm reached the airwaves. The store's proprietor told me he had sold 500 shovels that day. Neither his nor his supplier's inventory systems could anticipate the sudden surge in demand in time to react to it. Undoubtedly, plenty of shovels to meet the demand were available elsewhere but could not be moved efficiently to where they were needed.

This was the chaos theory at work. The nationwide demand for snow shovels in any given winter is probably relatively easy to forecast. But the variations in demand fluctuate to a greater degree with each move deeper into the supply chain. Regions will vary from one another, depending on the progress of the winter, and the individual retail stores are likely to vary most of all. While my local store ran out of stock, for instance, a retailer 50 miles to the west, bypassed by the snow, may have been sitting on plenty of shovels. A slight change in the storm track could easily have reversed their status. No inventory system I can conceive of could enable both retailers to have all the shovels they needed, and no more, in stock at the right time.

Using snow shovels as an example is arguably too easy a way to make the case. After all, snow shovel sales are directly related to the weather. And weather, of course, is the best known chaotic system. In fact, the first effort to formulate chaos theory came in 1963 from a meteorologist, Edward Lorenz. He described what is now well known as the butterfly effect—the idea that a butterfly flapping its wings in China will eventually affect the weather in New England. Even so, I believe the point is still valid.

Other writers have applied the chaos theory to a number of business systems, most notably the stock market. There is no reason to expect that anything as complex as business supply chains would be exempt. My unscientific understanding of chaos theory is this: Even very small changes in the variables in a complex system lead to greater and greater variability as the system progresses, before long leading to effects that are unpredictable. So even a small variation in, say, data input for demand for a particular SKU can lead to large variability in how successful a supply chain executes the task of getting the right inventory to the right place at the right time. And the data may not even be wrong, just not precise enough. If my understanding of the origin of Lorenz's initial insight is correct, he found vast differences in a weather forecasting model when he used a number that went out to three decimal places instead of six.

This does not mean that supply chain forecasting will fail, only that it is not perfectible. One of the critical, though often misunderstood, elements of chaos theory is that chaos is not random. Systems perform according to rules. Cause and effect are observable and measurable. Although it may seem a contradiction in terms, chaos is ordered. A system can be managed.

Efforts to improve inventory forecasting and replenishment strategies are essential to reduce the number of unpredictable variables in supply chain systems. Accelerating the speed of information through the supply chain and compressing cycle times throughout the chain are crucial to reducing uncertainty. Each time forecasting processes are made more sensitive to each of the elements that can affect their accuracy, forecast accuracy will improve. But it will never be dead on, when even the wisp of a change in the supply chain, like two small stones tossed simultaneously into a pond, creates patterns that follow definitive rules but that are impossible to predict.