TY - JOUR T1 - Why Modeling Averages Is Not Good Enough JF - Journal of Advertising Research JO - J Advert Res SP - 323 LP - 333 DO - 10.2501/S0021849910091488 VL - 50 IS - 3 AU - Martin Bongers AU - Jan Hofmeyr Y1 - 2010/09/01 UR - http://www.journalofadvertisingresearch.com/content/50/3/323.abstract N2 - For more than 40 years, Andrew Ehrenberg and others have demonstrated the value of the Dirichlet and related distributions as a method for modeling typical brand performance measures (BPMs) in markets. This work has led to trenchant critiques of many attitudinal measures related to branding—specifically, brand strength, differentiation, and persuasion. Using panel data, we show that though the Dirichlet leads to accurate BPMs for fixed time-slices, its assumptions about individual behavior are wrong. We therefore challenge the way the law-of-double-jeopardy theorists use patterns identified by the Dirichlet to generalize about brand performance, the psychology of consumer preference, and the role of advertising. Far from being problematic, attitudinal measures may be the only path to the successful understanding of the effects of marketing initiatives. ER -