The Second Meeting of the
Value Chain Academic-Industry Consortium

Thursday, May 22, 2003. Lucent Technologies, Bell Labs, Murray Hill, NJ.



 Invited Talk:
    The Martingale Model of Forecast Evolution
                                 - Peter L. Jackson, Ph.D.
                  Associate Professor
School of Operations Research and Industrial Engineering

Cornell University, Ithaca, NY 14853

(607) 255-9122       pj16@cornell.edu
www.orie.cornell.edu/~jackson


Abstract:
Sales forecasts drive long run planning in supply chain
production-distribution systems. Studying how these forecasts evolve
over time can lead to significant insights into the impact of forecast
accuracy on supply chain costs. In one study, we were able to suggest
that an alternative forecasting technique could generate savings for a
company in the food industry of several million dollars a year.
The savings came not from the improved accuracy of the final forecasts
but from the fact that greater accuracy was obtained at an earlier point
in time. This meant that the supply chain had time to react to forecast
changes and not have to rely on costly reactions to late-changing
forecasts. The methodology we have developed to conduct these studies
is referred to as the Martingale Model of Forecast Evolution.
In this talk, we will review the model, discuss its properties, and
illustrate its application.