The Marketing Book 5th Edition

COHORT MEANS

72 The Marketing Book

1997). In their original article, Buzzell et al.
‘removed’ much of the variation by calculating
cohort means. We can do the same and also use
more typical modern computer-generated
graphics to represent the results (Figure 4.5).
The cohort mean approach, although now
not commonly used in strategy research of this
sort, will show, as above, some deviations from
the straight line trend at sample sizes such as
500, but as samples get even larger the devia-
tions become, on average, even smaller: indeed,
some textbook representations of the results go
as far as merely illustrating the trend with no
deviations at all. Hence, in the process of
producing a clearer message from the data, we
have nearly eliminated nine tenths of the
variability in our performance variable. This is
much like the ‘trick’ used by many speakers
(including, of course, University Professors) of
allocating the last five minutes of a one-hour
talk to ‘other factors’, or ‘limitations with this
approach’, or some such heading. All very well
provided that the issues covered in the last five
minutes do not really dominate those which
were explored in much greater detail in the first
55 minutes!

How does one explain the

‘unexplained’ 90 per cent?

If we return to the scatter diagram and treat it

as if it represented the current performance of

500 business units within a single corporate

portfolio in terms of the relationship between

return on investment (ROI) and market share,

then we can see some of the problems that arise

when we try and make managerial evaluations.

The first set of problems relates to the nature of

the data and the way in the which the axes are

measured. In most analyses of this sort, and in

the PIMS data as we discussed above, the data

are essentially cross-sectional, i.e. either annual

or averaged out over a longer fixed period. Any

lead or lag effects are therefore excluded and

any particular one-off effects are compensated

for only to the extent that they are already

discounted from the input data, which are

normally based on management accounts. The

nature of the axes in a standard market share/

ROI analysis is a problem in that they are both

ratios. There are very considerable advantages

that accrue from using ratios in this situation –

most obviously the fact that it is possible to plot

Figure 4.5 Cohort means