Penetration or Loyalty

I’ve been discussing with dunnhumby’s Rob Turtle the question of penetration vs. loyalty. My interest was renewed after attending WARC’s recent Measuring Advertising Performance where Andrew Ehrenberg’s work on this was quoted in the first presentation and then referred to several times over in subsequent sessions.

This is a very simplistic synopsis (but it reflects how it was raised by speakers at the conference): Ehrenberg and his team at the Ehrenberg Bass Institute have concluded that in many situations ‘the main driver of [market share is] not customer loyalty but large differences in market penetration – the sheer number of customers each brand has.’ Consequently, the rallying call by the speakers was to not employ marketing strategies focussed on loyalty but instead to focus on driving increased penetration.

The theoretical model behind this is the Dirichlet.

As Rob Turtle eloquently explains:

Dirichlet is a choice model – it allows different customers to buy different stuff differently. To handle that mathematically, it tries to model the distribution of customers for a particular measure (for the technically minded, the model assumes zero-order Poisson choice process, Gamma purchase incidence/ rate of purchase, and multivariate Beta (=Dirichlet) brand loyalty).

Sticking to the coffee example {this is an example used by Ehrenburg] … the Beta distribution for loyalty might look like this (weight of purchase on the x-axis, number of customers on the y-axis):



There are two ways in which we can grow sales … either get more customers in (blue doted line), or shift the mode of the distribution to the right (red dotted line).


What Ehrenberg points out is that it is easier to get more customers to buy you, than it is to get customers to buy you more. Do one doubts that one can get new customers in … however, there is simply a limit to how much coffee someone can drink in a fixed period, and every distribution yet observed is positively skewed (i.e. has the mode on the left). It is this thinking that drives the conclusion.

This having been said, he also talks about “double jeopardy” which states that “big brands not only have more customers, but those customers tend to buy them more often as well” (the flip side being that small brands have less customers that buy less frequently). Which would seem to confuse the issue again!

So what might this mean for penetration vs. loyalty? As a matter of principle, I think that both Rob and I would disagree with any statement that simply says ‘penetration is more important that loyalty’ – trial that does not build repeat purchase will skew the distribution even more to the left. As Rob says, ‘reading between the lines, Ehrenberg would agree as there is a tacit admission that we want our trialists to then become typical buyers of the brand (in the example cited, buying 3 times – not just once!) In this way it does indeed drive the moments of truth … marketing gets the trial, the product gets the repeat. (Which incidentally explains the double-jeopardy: products have higher market shares because they are genuinely better products … which when I try one makes me more likely to repeat as it is better etc.)’

Graham Thomas