You step into an unfamiliar shower and the water is cold. You turn the tap toward hot. Nothing seems to happen, so you turn it further. A few seconds later scalding water arrives, and you wrench the tap back toward cold, overshooting again. The shower is a tiny lesson in one of the most important ideas in systems thinking: delays between action and result can make sensible people behave in ways that look foolish.
Delays are everywhere
Almost every feedback loop contains a delay. It takes time to notice a change, time to decide what to do, and time for the action to work. It helps to separate three kinds. A perception delay is the time before a change is noticed and measured, such as the weeks before economic statistics are published. A response delay is the time spent deciding, such as a budget debate. A delivery delay is the time for the action to take effect, such as the years needed to build a power station or train a nurse.
Overshoot and oscillation
A balancing loop tries to close the gap between where a stock is and where someone wants it. If the information arrives late, or the action works slowly, the person in charge keeps acting after enough has already been done. A delay in a balancing loop tends to produce overshoot and oscillation, with the stock swinging above and below its goal. The shower shows the cause: adjustments keep being made before the earlier adjustments have taken effect.
The beer game
In the 1960s Jay Forrester's system dynamics group at MIT developed the beer game, a classroom simulation still played by business students around the world. Four players manage a retailer, a wholesaler, a distributor and a brewery. Each sees only the orders arriving from the next link down and decides how much to order from the next link up. Shipments take a few weeks to arrive. Partway through, customer demand at the retailer rises once, modestly, and then stays flat.
What follows is almost always chaos. Shelves empty, players order heavily to catch up, the orders take weeks to fill, and then a flood of beer arrives that nobody wants. Swings grow larger the further the order travels from the customer. After the game, players typically blame each other or guess that demand must have gone wild. Yet the swings come from the game's structure of delays and ordering rules, not from foolish players.
The bullwhip effect
Real supply chains behave like the game. In 1997 the researchers Hau Lee, V. Padmanabhan and Seungjin Whang described what they called the bullwhip effect: small changes in consumer demand become ever larger swings in orders farther up a supply chain. They noted that Procter and Gamble had found exactly this with disposable diapers, a product whose use by babies is about as steady as demand can be, yet whose orders from distributors swung widely. A flick of the wrist at the handle of a whip becomes a crack at the tip.
The pork cycle
Markets with long production times show the same rhythm. When pork prices are high, farmers breed more pigs. But a pig takes many months to reach market, so for a while prices stay high and more farmers join in. Eventually the new pigs all arrive together, prices collapse, farmers cut back, and the cycle starts again. The German economist Arthur Hanau documented such cycles in 1928. The pork cycle arises from the delay between the decision to breed and the moment the pigs reach market.
Living with delays
The instinctive response to a delayed system is to react faster and harder. That usually makes things worse. When a delay is long, acting more slowly and less forcefully often stabilizes a system better than acting faster. The shower is best handled with a small turn and a moment's patience. Shortening the delay helps too: sharing sales data directly with every link in a supply chain cuts perception delays. So does keeping track of what is already on its way, the supply line, which beer game players tend to forget.
Policies on schools, crime and the economy work through long delays, and a government judged on this year's numbers will be tempted to keep pushing until it has pushed too far. Patience is not always weakness; sometimes it is the only way to steer.
Worked example: toilet paper, spring 2020
In March 2020 shelves of toilet paper emptied across North America and Europe. Total use barely changed: people needed no more than before. What changed was where they used it. With offices and schools closed, use shifted from workplaces, supplied with large commercial rolls through one set of factories and distributors, to homes, supplied with small retail rolls through another. Retail demand jumped while commercial demand collapsed.
Apply the ideas. Shoppers saw empty shelves, a perception signal, and bought extra to be safe. Stores, seeing sales soar, ordered more from distributors, who ordered more from mills. Mills could not quickly switch lines from commercial to retail rolls, a delivery delay. Each link amplified the signal, a bullwhip. When supply caught up a few months later, many households had a closet full and bought little for weeks. A structure built for steady demand, with separate channels and thin inventories, turned a modest shift into a national scramble.