Quantitative methods in marketing 219
characterized by a high level of decision
urgency (e.g. buying and selling stocks) or
complexity (e.g. retail site selection). Expert
systems are particularly useful in these
contexts because of their flexible and ‘friendly’
user interaction facilities, coupled with their
ability to explain their reasoning (Rangaswamy
et al., 1989). A number of expert systems in
marketing have been developed over the years,
in particular focusing on the following domains:
marketing research, test marketing, pricing,
generation of advertising appeals, choice of
promotional technique, selection of effective
sales techniques, negotiation strategies, site
selection, allocation of marketing budget,
promotion evaluation, strategic positioning,
strategic marketing, assessment of sales
territories, brand management, marketing
planning, international marketing, bank
marketing, tourism marketing and industrial
marketing (see Curry and Moutinho, 1991).The greatest single problem with regard to the
effectiveness and applicability of expert system
models in the marketing context concerns the
construction and validation of the knowledge
base.
Neural networks
Neural networks are designed to offer the end-
user the capability to bypass the rigidity of
expert systems and to develop ‘fuzzy logic’
decision-making tools. Several authors claim
that neural networks provide the user with the
ability to design a decision-support tool in less
time and with less effort than can be accom-
plished with other decision-support system
tools. Neural networks have been successfully
applied in the following marketing areas: con-
sumer behaviour analysis (Curry and Mou-
tinho, 1993), market segmentation, pricing
modelling (Ellis et al., 1991), copy strategy and
media planning (Kennedy, 1991).
Neural networks use structured input and
output data to develop patterns that mimic
human decision making. Input data are com-
pared to relative output data for many data
points. The relationships between the input
data and output data are used to develop a
pattern that represents the decision-making
style of the user. The development of patterns
from data points eliminates the need to build
rules that support decision making. Unlike
expert systems, which require user intervention
to accommodate variable changes within the
model, the neural network is capable of retrain-
ing, which is accomplished through the addi-
tion of new input and output data.
An important strength of this method is its
ability to bring together psychometric and
econometric analyses, so that the best features
of both can be exploited. Whereas expert
systems are good at organizing masses of
information, neural networks may prove cap-
able of duplicating the kind of intuitive, trial-
and-error thinking marketing managers typi-
cally require. The accuracy of the neural
network is not as high as of other methods, yet
it has the ability to learn from increased input/
output facts and the ability to address data that
other decision-support systems cannot handle
logically.
Table 9.9 presents the main applications,
advantages and limitations of expert systems
and neural networks.Statistical decision theory or stochastic methods
In this category there are a number of methods,
all of which are useful in solving marketing
problems.Queuing
This method is of importance to large retailing
institutions such as supermarkets, petrol sta-
tions, airline ticket offices, seaports, airports
and other areas where services are available
through queuing. A notable problem in retail-
ing institutions is that of making salesforce