skip to content

Using the right incentives to improve forecast accuracy

New insights from the Department of Supply Chain Management & Management Science

Lisa Scheele

Many companies act in markets with uncertain demand. Usually, it is the task of employees in marketing and sales departments to collect market information and to create demand forecasts. These forecasts are communicated to the operative units within the company and they serve as input to a long planning process – ranging from inventory levels to production plans and purchasing quantities.

In practice, however, it can be observed that demand forecasts are often systematically too high. That is, the forecasted quantities are greater than actual demand. One reason behind this phenomenon can be incentive systems that motivate sales people to maximize their sales volume. This objective can best be achieved if inventory levels are high. This way, there is little risk to be out of stock on a certain product. Hence, it is not surprising that demand forecasts are systematically inflated.

Pilot project with a global pharmaceutical company
Lisa Scheele and Ulrich Thonemann of the Department of Supply Chain Management & Management Science investigated, if new incentive systems for sales people can improve the accuracy of their forecasts. In a pilot project with a global pharmaceutical company, they found out that incentivizing forecast accuracy significantly increases the awareness of sales people for the importance of the topic. They also observed, that such incentives can lead to improvements in forecast accuracy.

Game-theoretic model and laboratory experiments
In a second step, the research team built a game-theoretic model that captures the forecast information exchange between a sales and a production department. The model not only includes different incentive systems for sales people, but it also allows for non-rational decision making. This means, for example, that human beings have a natural aversion to lying. That is, people do not like to communicate a distorted forecast. Also, payoff streams resulting from different objectives of an incentive system might be unequally weighted. In particular, penalizing a bad forecast accuracy weighs more heavily than rewarding a good sales performance. The researchers tested their model in laboratory experiments and developed incentive systems that penalize forecast errors and lead to truthful and unbiased demand forecasts. Interestingly, incentive systems that penalize overforecasting more heavily than underforecasting deliver the best results.

Implications for practice
The outcomes of this research project provide guidelines for the design of incentive systems in practice. To improve their forecast quality, companies should measure forecast accuracy and include corresponding targets in the incentive system of their sales people.

Scheele, L.M., Thonemann, U.W., Slikker, M., Designing Incentive Systems for Truthful Forecast Information Sharing Within a Firm, Management Science, forthcoming.