Market shareReturn on investmentThe basics of marketing strategy 71
dependent variable and, second, the extent to
which even if it does this regularity actually
enables one to produce a clear prescription
for managerial action.
2 Case study-based research on selected firms,
often based on the notion of some form of
outliers, such as those that perform
particularly well. Here the problems are the
extent to which the story that is told about
the particular nature of the success concerned
can be used to guide action in other
organizations. In practice this often results in
managerial prescriptions that are rather
tautological and at the same time
non-discriminating.
We will now consider examples of both types of
this research.
Market share and ROI: the 10 per
cent rule in practice
One of the most famous results from the PIMS
database was that first reported by Bob Buzzell,
Brad Gale and Ralph Sultan in the Harvard
Business Reviewin 1975 under the title ‘Market
Share – A Key to Profitability’. They reported
on the relationship between ROI and market
share on a cross-sectional basis within the then
current PIMS database. Although, over the
years, estimates of the R^2 of this relationship
have varied, it generally shows a value of
around 10 per cent up to a maximum of 15 per
cent. We can start by simulating the original
data that were used (Figure 4.4).
Figure 4.4 is a scatter plot of 500 datapoints
(notional observations) where the relationship
between the two implied variables is actually
the equivalent of an R^2 of 0.12 or 12 per cent.
Because of the statistical nature of the data
distribution in the PIMS database, the fact that
it is not strictly normal, it is only possible to
simulate a dataset which has either the right
range or the right slope within the correct
proportion of variance explained. This simula-
tion is based on the right range of values, so
that the extreme points are estimated correctly.
As a result, however, the actual slope is
underestimated (see Wensley 1997a, b; Roberts,Figure 4.4 A scatter plot of 500 databases (notional observations)