Why Modeling Averages Is not Good Enough: A Critique of the Law of Double Jeopardy

Martin Bongers and Jan Hofmeyr



According to the law of double jeopardy (DJ) in markets, big brands benefit doubly: They have more users, and their users use them more. More formally: “In all markets, all conventionally used brand performance measures (BPMs) can be predicted from market share using the Dirichlet distribution (once calibrated)” (Chatfield and Goodhardt, 1975).

The law challenges many marketing-research approaches. Consider brand equity. Most attitudinal systems are based on some sort of commitment or engagement ”ladder” (Keller and Lehman, 2004). These are used to identify strong and weak brands—typically by quantifying the percentage of people found at various “levels.” Ehrenberg and others have argued repeatedly, however, that all can be accounted for by brand size alone. To quote “… there are clearly big brands and smaller brands, (but) there is no evidence… that over and above this there are ‘strong’ and ‘weak’ ones” (Ehrenberg and Goodhardt, 2004).