The game is played on a board that portrays the production and distribution of beer (figures 1-2). Each team consists of four sectors: Retailer, Wholesaler, Distributor, and Factory (R, W, D, F) arranged in a linear distribution chain. One or two people manage each sector. Pennies stand for cases of beer. A deck of cards represents customer demand. Each simulated week, customers purchase from the retailer, who ships the beer requested out of inventory. The retailer in turn orders from the wholesaler, who ships the beer requested out of their own inventory. Likewise the wholesaler orders and receives beer from the distributor, who in turn orders and receives beer from the factory, where the beer is brewed. At each stage there are shipping delays and order processing delays. The players' objective is to minimize total team costs. Inventory holding costs are $.50/case/week. Backlog costs are $1.00/case/week, to capture both the lost revenue and the ill will a stockout causes among customers. Costs are assessed at each link of the distribution chain.

The game can be played with anywhere from four to hundreds of people. Each person is asked to bet $1, with the pot going to the team with the lowest total costs, winner take all. The game is initialized in equilibrium. Each inventory contains 12 cases and initial throughput is four cases per week. In the first few weeks of the game the players learn the mechanics of filling orders, recording inventory, etc. During this time customer demand remains constant at four cases per week, and each player is directed to order four cases, maintaining the equilibrium. Beginning with week four the players are allowed to order any quantity they wish, and are told that customer demand may vary; one of their jobs is to forecast demand. Players are told the game will run for 50 simulated weeks, but play is actually halted after 36 weeks to avoid horizon effects.

Each player has good local information but severely limited global information. Players keep records of their inventory, backlog and orders placed with their supplier each week. However, people are directed not to communicate with one another; information is passed through orders and shipments. Customer demand is not known to any of the players in advance. Only the retailers discover customer demand as the game proceeds. The others learn only what their own customer orders.

These information limitations imply that the players are unable to coordinate their decisions or jointly plan strategy, even though the objective of each team is to minimize total costs. As in many real life settings, the global optimization problem must be factored into subproblems distributed throughout the organization.

The game is deceptively simple compared to real life. All you have to do is meet customer demand and order enough from your own supplier to keep your inventory low while avoiding costly backlogs. There are no machine breakdowns or other random events, no labor problems, no capacity limits or financial constraints. Yet the results are shocking.

Typical Results: Boom and Bust

Figure 2 shows actual results from teams consisting of graduate students and business executives. Each column shows the results of a single team. The top four graphs show the orders placed by the players, from the retailer (bottom) to factory (top). The bottom four graphs show the players' inventories and backlogs (negative values), in the same order. Average team costs are about $2000, though it is not uncommon for costs to exceed $10,000; few ever go below $1000. Optimal performance (calculated using only the information actually available to players themselves) is about $200. Average costs are ten times greater than optimal!