When forecasts are made in corporate settings, their accuracy will be affected by biases derived from individual and organizational incentives. Seeing little research concerning how to manage the organizational and political dimensions of generating and improving forecasts in corporate settings, Rogelio Oliva of the Mays Business School of Texas A&M University, and Noel Watson of Harvard Business School, examined forecasting at an anonymous California-based consumer electronics firm. They studied the implementation of a supply chain planning process at the company by interviewing 25 people involved in the sales forecasting process. Their focus was on the forecasting process and how it was affected by individual, group, and organizational biases that detract from forecast accuracy.

The authors introduce a framework of functional biases whose sources are intentional, driven by misalignment of incentives, and unintentional resulting from informational and procedural blind spots. The sub-optimal system at the company resulted in an inventory write-off equivalent to about 10 percent of revenues and the appointment, the fall of 2001, of a new CEO and five new vice-presidents. In April 2002, the newly hired director of planning and fulfillment took on the task of improving the forecasting process by making planning information move more rapidly and accurately through the supply chain. The new method did not allow functional groups to make their own forecasts. Instead, they had to agree on an overall consensus forecast—a process that was now led by an independent group called the Demand Management Organization.

The creation of an independent group for managing the forecasting process (rather than managing forecasts directly) succeeded in moderating the multiple sources of bias. Between the summer of 2002 and the fall of 2003, the accuracy of sell-through forecasts rose from 58 percent to 88 percent, and inventory turns increased, on-hand inventory decreased, and obsolescence costs were cut. The authors found that the deployment of the new system introduced new dynamics that affected biases. They note that “The recognition that the system both needs to account, and is in part responsible, for partners’ biases introduces a level of design complexity not currently acknowledged in the literature or by practitioners.”