Here's a more subtle example. Suppose you're comparing two suppliers on their ability to deliver materials quickly. Supplier A requires an average of 17 days to fill its orders, whereas Supplier B requires 19 days on average, so Supplier A is obviously the better choice. But the actual measurements underlying these averages tell a different story (see Figure 4): A is much less consistent in its delivery times than B, with lead times as little as nine days and as long as 25 days. Early deliveries cause you to hold inventory longer than necessary, and late deliveries require you to increase inventory levels to avoid stockouts. So, any deviation from the requested delivery date requires you to hold more inventory. When you take these holding costs into account, you find that consistency is more important than absolute speed, and that makes Supplier B the better choice.
FIGURE 4 Two lead time distributions.
These examples show the importance of examining the entire distribution of measurements to understand what's actually going on in your business. Of course, poring over graphs of data is a rather tedious process, and you don't want to do that for every set of results. But statisticians solved this problem long ago. Just as statistics exist to represent the average value of a set of measures, other statistics can represent variation from that average.
The best choice for most business purposes is the standard deviation, because it's easy to interpret. When the results follow the normal distribution of data shown in Figures 3 and 4, 99.7 percent of the data falls within three standard deviations above or below the mean. In Figure 3, for example, the mean is 100 and the standard deviation is 10, so you know without even looking at the distribution that the likelihood of getting a value of 130 or more is pretty close to zero.
These examples illustrate the importance of analyzing the variability in your measurements, but they also illustrate a deeper insight: Variability is bad for business. Despite the long-standing focus on average values, variation around the average is often the real killer in a supply chain. Variation almost always increases total cost, and even minor deviations in the normal flow of goods cascade down the chain in a self-amplifying pattern know as the bullwhip effect, a phenomenon that inflicts a great deal of needless pain.5
In many situations where time is of the essence, it's actually better to increase the average value of an interval, if necessary, in order to reduce its variability. Consider, for example, the supply chains that serve JIT production facilities, in which small shipments of materials arrive on a frequent, periodic basis. In these environments, suppliers are often given a 15-minute window in which to deliver their goods, and they're penalized for being too early as well as too late. When a producer enjoys this level of reliability in its suppliers, the conventional measure of fulfillment time becomes much less important.
In short, don't focus too tightly on your average performance. Consider the variability in that performance as well, and seek to improve consistency along with your averages. You already have all the data you need to do that. Just by looking at two numbers rather than one, you may be able to take your company well beyond what the competition is doing.
David A. Taylor, Ph.D is a writer and consultant in the area of supply chain technology and performance. He can be reached through his Web site, www.supplychainguide.com.
- "What Is the Right Supply Chain for Your Product?" M.L. Fisher, Harvard Business Review, March-April, 1997.
- Supply Chains: A Manager's Guide, D.A. Taylor, Addison-Wesley, 2004. See especially Chapter 9.
- "Forging the Chain," G. Taninecz, Industry Week, May 15, 2000.
- "Lessons from the Leaders," M. Cook and R. Tyndall, Supply Chain Management Review, November-December 2001.
- "The Paralyzing Curse of the Bullwhip Effect in a Supply Chain," H. Lee, P. Padmanabhan, and S. Whang, Sloan Management Review, Spring 1997.