An examination of differential weighting of successive waves of product usage estimates

In a 1999 paper by the same authors, it was concluded that giving more weight to the more recent data in combining successive waves of magazine readership data led to better predictions of subsequent estimates. Specifically, 20/80 weighting produced better predictions than the traditional 50/50 weighting used in double-bases, especially in cases where there was a particularly significant trend in a magazine’s readership.
The purpose of this paper is to report on a similar examination of product usage data. The underlying thought is that if similar findings resulted, then it would make sense to apply 20/80 weighting to both readership and product usage data when employing double-bases, allowing media planners to perform their customary cross-tabulations with appropriate weights. We believe that the differences observed in two successive report periods of product and brand usage data may be partially due to sampling variation, but may be partly due to actual trends in usage for at least some of the product categories and brands. In order to test that hypothesis, successive periods of product usage estimates were averaged based on various weighting schemes, such as 40/60 and 20/80 (in comparison to the 50/50 of “doublebase�) so as to determine the ideal weight to place on the more recent data with the purpose of best predicting the subsequent report.

Download Paper (PDF)