Why Marketing Needs Science
1John Scriven and Colin McDonald
Marketing courses and textbooks still encourage a number of myths and misconceptions,
which are attractive because they chime with `common sense‟ (or rather, wishful
thinking), but which we now know to be wrong.
The scientist, in contrast, is guided by the importance of empirical generalisation as the key process leading to knowledge. He/she looks for simple relationships that hold
under a wide range of different conditions. Then, and only then, does it makes sense to
describe these patterns in a `model‟, test this model against other examples, and observe
where deviations (if any) occur. One thus arrives at knowledge which can properly be
called scientific: knowledge based on thousands of disparate observations of what
normally happens in markets, under different conditions, and therefore what we may expect for our market.
This paper describes how the Ehrenberg Centre puts this principle into practice, illustrated by some of the important results produced.
Marketing courses and textbooks encourage a number of myths and misconceptions,
which are attractive because they chime with „common sense‟ (or rather, wishful
thinking), but which we now know to be wrong. The scientist, in contrast, is guided by
the importance of empirical generalisation as the key process leading to knowledge. S/he
looks for simple relationships that hold under a wide range of different conditions. Only
then does it makes sense to describe these patterns in a „model‟, test this model against
other examples, and observe where deviations (if any) occur. One thus arrives at
knowledge which can properly be called scientific: knowledge based on thousands of
disparate observations of what normally happens, under different conditions, and
therefore what we may predict to happen again. “Laws of nature are the generalisation of
natural occurrences with which the occurrences are always in accordance” (Frege, 1956).
At this point, many people are tempted to switch off. Practical marketing people tend not to give marketing science a high priority. Reactions range from the weakly
positive (“very interesting, but I really don‟t have the time”) to the bluntly dismissive
(“what‟s it got to do with my market?” Or “My market is different”?). Marketers are too
busy fighting off the intense pressures of the particular to care much about the general.
They want to change their market in their favour. They want their research to tell them
what to do, now; general descriptions of how markets work can seem at best remote, at
worst frustratingly irrelevant.
Faced with this, there is no point in the marketing scientists simply carrying on while complaining that people don‟t pay attention. They have to be proactive in
explaining the value of marketing science, because without it marketers are operating in a
1 Director and Associate, Ehrenberg Centre for Research in Marketing, London South Bank University.
The Ehrenberg Centre was established to develop and disseminate soundly based knowledge about how
vacuum. If targets are set or promotional programmes developed in line with these myths,
the results can be counter-productive and expensive. This is why the Centre‟s ongoing
work is so important. Marketing science is easily resented (especially, perhaps, in the
advertising world) because it seems to set rules and constraints which frustrate our
creative impulses. But this is an illogical position to take. An engineer or an architect
would not get very far designing a bridge or a building without prior knowledge of
architectural principles and the strengths and weaknesses of building materials. The
„constraints‟ in this knowledge are not an obstacle but a help, because they guide the
design in those directions where success can be achieved and wasteful failures avoided. It
is exactly the same in marketing.
There is a lot more to marketing science than finding and modelling statistical relationships in particular markets. Marketers often use multivariate statistical tools (such
as regression, factor or cluster analysis, etc.) to unravel complex situations, but they are
beset with well-known problems of interpretation: how to decide what variables to
include, whether a relationship is truly causal, which of several reasonably-fitting models
to believe. Trying to predict from a „best fit‟ model of a single set of historical data is not
science. The scientist, in contrast, looks for recurring patterns from a wide range of
different conditions. He likes to do deliberate interventions and experimentally observe
what happens. Only after observing a pattern that holds robustly, in all sorts of markets
and countries, will he model it and call the result (in Ehrenberg‟s phrase) a „law-like
relationship‟. Having established in this way what is „normal‟, he has some hope of then
understanding the (relatively rare) variations that may occur. The marketer, crucially, can
use this knowledge to interpret actual results: are they higher, lower or about the same as
the model tells him to expect?
The Centre‟s work has established such norms in several areas including:
? People‟s purchasing of the category
? Buyer concentration
? Brand buying
? Loyalty (and implications for communication strategy)
Here are some examples of the practical implications derived from a thorough
understanding of the patterns, and their underlying cause.
It is commonly argued that it is wasteful to try to communicate with people who are only
light or occasional users (let alone non-users) of your brand. The Pareto principle is
quoted, that 20% of buyers account for 80% of sales. As it happens, we now know that
the true figure for repertoire products is nearer 50% than 80%, but the principle still holds:
it is said it costs much less, and is more profitable, to retain a customer than to attract a
new one. Therefore, since these people already know the brand and presumably love it,
why not talk exclusively to them and get them to buy even more?
Preaching to the converted is attractive and appeals to our comfort zone. But our prior knowledge of how markets work tells us that a strategy based exclusively on this
idea must fail: if talking to our most loyal customers is all we do, not only will the brand
not grow, it will eventually decline. The reasoning is simple. Important as it is to keep
good relations with our heavy buyers, it is unrealistic to expect them to buy more, and we
risk cutting ourselves off from the 80% of our existing customer base who account for
half our sales, let alone the majority of category buyers who do not buy us at all.
The reason is that most consumers are polygamous in their buying habits, which
results in all brands having more light than heavy buyers. This manifests itself in the
phenomenon that has come to be known as Double Jeopardy (DJ). In any market where
there are differing market shares, those shares are accounted for mainly by penetration
(the number of customers the brand has). Measures of loyalty, such as average purchase
frequency or share of category, also differ, but by very much less, often hardly at all. The
brand leaders do show more loyalty, but only a little more: they have more frequent
buyers, but to balance that they also have more occasional buyers in absolute numbers.
Table 1 shows a typical example.
Toilet Soaps, UK Market Penetration Purchase % Buying Table 1 Key performance metrics for UK Toilet Soap Top 10 Brands Share % per buyer Once 5+ 52 w/e Jan 2005 %
Dove 24 16 2.9 49 15
Imperial Leather 20 18 2.0 58 7
Tesco 6 5 2.2 64 6
Palmolive 5 4 2.0 65 7
Cussons 4 4 1.8 68 7
Asda 4 4 1.9 64 6
Simple 3 3 2.1 60 10
Pearl (Cussons) 3 3 1.7 71 6
Fairy 3 3 1.7 68 3
Pears 2 2 2.2 67 11
Average 8 6 2.1 64 8
Source: TNS Superpanel
Here, the biggest brand is bought by nearly ten times more customers than the smallest
brand, but the loyalty of its buyers (as measured by the average purchase frequency) is
only 30% higher. This same general pattern (though the numbers vary) holds for virtually
all markets studied across the world, including „high involvement‟ purchases like cars
and durables: it has been found in commodities like petrol, doctors‟ prescribing habits,
channels viewed on TV. It has been found not only between brands but also between
functional varieties within a market. It also applies when brands grow or decline. In a
study of 72 brands which gained and 79 which lost share between successive years, the
gain or loss in penetration was 2-3 times greater, on average, than the gain or loss in
purchase frequency. In typical FMCG markets where the brand leader‟s share is no more
than, say, 20%, the DJ difference in loyalty measures can be very small indeed.
Underlying DJ is a pattern of consumer behaviour which was established by Ehrenberg and his colleagues more than 30 years ago. People buy a product as often as
they need to, some very often (say, once a week), many only occasionally: many grocery
products are bought only a few times a year on average. When they buy, they choose
from a limited repertoire of brands – again, some more often, some less. We say they
have propensities to buy (have formed habits of buying) the brands in their consideration
set. We could call these propensities loyalty: everyone is „loyal‟ to the brands they buy,
and almost everybody switches between them. In most markets, few people are loyal
100% to one brand only, and they are mostly infrequent buyers of the category; the
biggest repertoires and the most switching belong to the frequent, regular buyers.
The constraint that DJ puts on market planning will be clear. To increase sales, the main thing you have to do is to sell to more people: to get the brand into more
people‟s repertoires. For this, reach is essential. It is not enough simply to work on your
existing loyal customers. This does not of course mean that loyalty is unimportant. A
growing brand will grow in both user numbers and user loyalty. But the smallness of the
DJ difference means that the loyalty balance, between lighter and heavier users, cannot
change much. Good relationships with the heavy buyers are important because we don‟t
want them to defect, but if we talk to them alone, the lighter buyers will lose interest and
drift away. Since there is always movement, some heavier buyers will shift towards the
competition, and with no new input the brand will „rebalance‟ downwards and lose share.
So, communication only with one‟s „most valuable‟ customers will make it hard even to
hold a position, much less gain.
DJ is a vital benchmarking device for planning and evaluation of new brands, for monitoring established ones, and for understanding why so often loyalty schemes, CRM
etc. fail to deliver dramatic shifts in sales. It is no good basing a launch plan on achieving
a purchase frequency of 8 per year if the average in the market is 2 and even the leader
has no more than 4. If Brand X has a low repeat level, it is not necessarily a danger signal;
Brand X may merely be behaving normally for its size.
Brand differentiation, niche appeal and salience
Another widespread belief is that your brand is different from the others and therefore
appeals to different people, so that the best approach is to segment the market and target
But again, the knowledge that we now have shows otherwise. The Duplication Law (based on generalisation from hundreds of categories across the world, including
patterns of TV viewing and the choice of the next car) shows that the extent to which
buyers of any brand will choose any other brand is mainly proportional to, and
predictable from, the other brand‟s share alone. This means that brand choices are
virtually independent: buyers of Brand A are no more, or less, likely to also buy Brand B
than are buyers of Brands C, D or E (to also buy B). And Brand A buyers are not
different sorts of people from Brand C or D buyers: there may simply be more or fewer of
What the Duplication Law means is that, in brand choices (it doesn‟t always apply to product variants), markets are mainly unsegmented. The rule is general but not
immutable: sometimes a duplication table will reveal a little cluster of above or below
average duplications (see the example in Table 2). What this usually tells us is that we
have a sub-category: e.g. those who prefer pre-sweetened products within RTE cereals, or
fruit-flavoured carbonated drinks. Allowing for this, it seems that people do not behave as
if brands were different, but as if they were substitutable alternatives: the brand
combinations that occur in people‟s repertoires are quasi-random.
But, it may be argued, this „quasi-randomness‟ in people‟s choices may conceal the
possibility that they respond to their brands for different reasons. After all, most people
like variety: they may be fulfilling a basic need, but they want to ring the changes. Their
choices may reflect different emotional states or moods, which are satisfied by different
brand „personalities‟. Therefore, it pays marketers to segment potential consumers
according to psychological or „lifestyle‟ factors, to find out which brands appeal to which
emotional states, and if necessary use advertising to create the required brand personality.
This could be true at the individual level: everyone has their own reasons for their
repertoire choice. The trouble is, it does not seem to generalise. A study of some 400
brands in 40 product fields (from fmcgs to airlines) covered on the TGI showed no brand-
related segmentation in terms of psychological or lifestyle profiling. There is thus no
evidence that what a brand may mean to one person, in emotional terms, will be the same
as what it means to another. Emotional attachment does not seem to define a market for a
brand (other than that users of a brand tend to like it!). Furthermore, the Duplication Law
and the patterns of repertoire buying strongly suggest that in general people do not have
strong emotional bonds with brands: rather, they use branding to help them form habits
which satisfy them with minimal effort. Brands are simply a way to short-cut the
inconvenience of having to make essentially trivial choices. This is equally the case with
high-ticket items: around half of repeat-buyers of cars buy the same make again, and the
others switch in line with the brands‟ penetrations, as the Duplication Law predicts.
Table 2 % Buyers of a brand who also buy each other brand
Toilet Soaps, UK Who also buy
Dove Leather Tesco olive Cusson Asda Simple Pearl Fairy Pears Top 10 Brands
52 w/e Jan 2005 Horizontal %
% Buyers of Dove - 18 5 5 5 5 3 4 3 3 Imperial Leather 16 - 6 6 8 4 3 6 5 3 Tesco 16 23 - 7 6 6 6 5 3 5 Palmolive 18 28 8 - 11 4 4 11 9 3 Cussons 18 33 7 11 - 7 4 16 7 2 Asda 20 20 7 5 9 - 4 6 5 2 Simple 16 19 9 6 6 5 - 3 3 4 Pearl (Cussons) 17 31 7 14 21 6 3 - 7 1 Fairy 17 29 4 13 11 6 3 8 - 2 Pears 25 25 13 8 4 5 7 3 3 -
18 25 7 8 9 5 4 7 5 3 Average
Source: TNS Superpanel
Marketers will obviously exploit functional superiorities in their brands where they exist
(and have not yet been copied). But many brands with different shares would not be
distinguishable in blind testing. What drives market share and therefore sales is getting
the brand more widely known: simply getting as many people as possible to think about it,
whether it is „different‟ or not.. This has important implications for brand-building and advertising. Most media advertising does not present a brand as functionally different
(and, another study shows, most advertisements are not perceived as attempting to do
that). What it does do, when it works, is to make the brand widely visible, „salient‟, and
therefore more likely to be thought about and trusted. It has long been known that
positive evaluations of brands correlate with their shares. As McPhee originally
recognised in his „two restaurants‟ example, behavioural and attitudinal DJ go together. This is why brand leaders are hard to dislodge, and why small brands have to work harder
and spend relatively more on advertising.
These are just two examples of how the past work of the Centre has established
frameworks of knowledge within which marketers can work more effectively. There are
many others, past and ongoing. For example,
? Pricing: how to set prices optimally is always a major concern, and experimental
work at London South Bank University has produced valuable evidence about
elasticities, reference prices and when price changes are effective.
? Media: current concerns about fragmentation have been shown to be only partly
justified: a recent study of US and UK viewing data has shown that patterns of TV
channel use have changed little from 20 years ago, though spread over more
channels, and that there is little programme loyalty – the DJ and Duplication
patterns apply to TV viewing.
Work continues in all these areas, always with the aim of establishing meaningful metrics
and benchmarks for marketers. Besides the above, the Centre‟s staff and associate
researchers are producing original work with relevance to branding, salience and
forecasting, media planning, and advertising effectiveness.
This important work could not happen without the continuing financial support
from corporate members across Europe, Australasia, USA, South Africa and Canada.
Members receive prior reports of all the dramatic discoveries made by the Centre‟s teams, and regular tailor-made seminars and in-house briefings explaining their significance and
Frege, G. (1956) „The thought: a logical enquiry‟, tr. A.M. and M. Quinton, Mind, 65,
reprinted in Philosophical Logic, ed. P.F. Strawson, OUP, 1967
50 Ehrenberg Centre reports are available to sponsoring companies.