Whang and Lee: Eliminating the Bullwhip Effect in Supply Chains

If you ask Seungjin Whang about the research and colleague Hau Lee do in supply chain management, he is likely to answer with a story about diapers. Babies consume diapers at a relatively steady rate, he will tell you. Month to month, the number of babies and the number of diapers that their parents purchase from Kmart, Safeway, or the local convenience store remain roughly the same. But several years ago, Procter & Gamble, which manufactures Pampers, studied its diaper sales to retailers and, despite the steady rate of demand among babies, found puzzling, dramatic fluctuations in retailers’ orders to wholesalers.

Even wilder were the fluctuations in orders that P&G was receiving from the wholesalers. “If the variability in demand among babies was small, why was there such a marked variability in demand from retailers?” asks Whang, Jagdeep and Roshni Singh Professor of Operations, Information, and Technology at the Stanford Business School. “And why was it even more extreme among wholesalers placing orders to manufacturers?”

Whang and Lee, the Thoma Professor of Operations, Information, and Technology, are describing the “bullwhip effect,” named for the variations in reaction down the length of a whip after it is cracked. They blame it for a host of expensive manufacturing problems. The bullwhip effect happens in manufacturing when information about consumer demand—for diapers or for any other product—becomes increasingly distorted as it moves upstream in the manufacturing process. This distortion leads to excessive inventory throughout the system, poor product forecasts, insufficient or excessive capacities, product unavailability, and higher costs generally.

In 1995 Lee, and V. “Paddy” Padmanabhan, now of INSEAD University, Singapore, and Whang wrote a paper that attempts to identify the causes of the bullwhip effect and explore ways to begin eliminating it. Today Lee and Whang are co-directors of the Stanford Global Supply Chain Management Forum.

They believe the problem can be fixed. But first, companies have to start cooperating and sharing data that has traditionally been considered proprietary. Sales estimates and forecasting are usually done separately by retailers, manufacturers, and suppliers. When retailers notice a slight increase in demand for diapers, say, besides putting in an order with the wholesaler to replace the diapers sold, they may order extra in case the small upturn in sales indicates a trend. The wholesaler gets the order, sees an uptick in diaper orders, and makes its own forecasts — which are blurrier than the retailers’ because they aren’t based on any real sales figures. Then, when a manufacturer tries to interpret orders coming from the wholesaler, the perceived increase in demand can become further exaggerated: the bullwhip effect.

To get around this, information about demand at the site farthest downstream must be made available to the upstream sites, say the researchers. In other words, retailers must tell manufacturers exactly how various items are selling. This gives the manufacturers necessary data for making sound plans for the future. “If I know more about real demand patterns, I can reduce my inventory,” says Whang. “But I have to have my retailers’ information. Instead of having each company myopically optimize its own inventory, companies need to look at the whole supply chain.”

Another traditional business practice that leads to the bullwhip effect is “batching” orders at certain times of the month or year. Companies make a practice of batching because it saves time and money. It is cheaper, for example, to transport infrequent large orders than frequent small ones. Likewise, the administrative costs of generating one large order of a certain item are smaller than the cost of generating many small orders of the same item.

But while sporadic ordering is cheaper, it leads to unsettling surges in demand on the manufacturing side. Plants may scramble to accommodate beginning-of-the-month orders, then lie idle for weeks while the products they rushed to make sit in a retailer’s warehouse. A healthier scenario, say the researchers, would be to have manufacturers receive a steady stream of orders, reflecting actual day-by-day consumer demand. The higher administrative and transportation costs of this system could be reduced by consolidating loads from multiple suppliers to achieve transportation economies of scale and switching from a paper-based to a computer-based ordering system.

A third cause of the bullwhip effect is a practice known as “gaming” — intentionally presenting manufacturers with a false picture of consumer demand. For instance, an electronics chain might know that the maker of a popular new laser printer currently has very limited capacity to produce the item. Meanwhile, the printer is selling rapidly and generating healthy profits for the chain, which wants to order 100 more. The chain knows the manufacturer is going to be rationing printers: retailers will be allocated half of what they request. So in order to get the 100 printers it wants, the retailer asks for 200. When other retailers behave the same way and each orders 200, the chain may not receive 100 printers. But it also leaves the manufacturer with a very distorted picture of consumer demand for the product. To discourage gaming, they suggest manufacturers allocate scarce products to retailers based on past sales records rather than on the amount requested in a particular order.

A final strike against the bullwhip effect would be to minimize price incentives. When manufacturers offer bargains, retailers stockpile inventory and don’t order again for months. This is not the way to keep the supply chain running smoothly. The logic behind an “everyday low price” is that it promotes steady, regular purchases at all levels, from the retailer to the wholesaler to the manufacturer, rather than sporadic shopping binges, which are at the root of the bullwhip effect.