Loyalty/Engagement scores do not predict sales growth
11/05/11 10:22
I took the full list of Brand Keys 2011 Engagement Award ranks and correlated them again sales gains. What a surprise! A correlation of 0.4 but not in the direction that the people selling this claim, the data says the better the Band Keys rank the less the sales gain !
I took the full list of Brand Keys 2011 Engagement Award ranks and correlated them again sales gains. What a surprise! A correlation of 0.4 but it's round the other way.....
i.e. the better the Band Keys rank the less the sales gain !
Brand Keys, who sell these surveys, claim that these award scores will predict a brand's future. Seems like you've absolutely got to know which numbers to use and which to ignore, which can only be done after the event. That's not prediction, it's weaving a story afterwards (to sell a product).
Pearson's correlation 0.39 (or 0.11 for just the 5 top ranked brands)
Data sources: Brand Keys Loyalty Awards 2011 - http://www.brandkeys.com/awards/index.cfm Business Week Car Sales - http://online.wsj.com/mdc/public/page/2_3022-autosales.html
Professor Byron Sharp
PS it isn't quite technically correct to use Pearson's correlation for a continuous variable against a rank order variable, but here converting both to ranks and using a Spearman's correlation produces essentially the same result.
i.e. the better the Band Keys rank the less the sales gain !
Brand Keys, who sell these surveys, claim that these award scores will predict a brand's future. Seems like you've absolutely got to know which numbers to use and which to ignore, which can only be done after the event. That's not prediction, it's weaving a story afterwards (to sell a product).
Pearson's correlation 0.39 (or 0.11 for just the 5 top ranked brands)
Data sources: Brand Keys Loyalty Awards 2011 - http://www.brandkeys.com/awards/index.cfm Business Week Car Sales - http://online.wsj.com/mdc/public/page/2_3022-autosales.html
Professor Byron Sharp
PS it isn't quite technically correct to use Pearson's correlation for a continuous variable against a rank order variable, but here converting both to ranks and using a Spearman's correlation produces essentially the same result.