198 The Marketing Book
In this way, most of the models and tech-
niques can be analysed. Their validity can be
judged from their usage, how accurately they
represent the problem environment, their pre-
dictive power, and the consistency and realism
of their assumptions.
In selecting an appropriate method of ana-
lysis, two major factors should be taken into con-
sideration: first, whether the variables analysed
are dependent or interdependent and, second,
whether the input data are of a metric or non-
metric form. Metric data are measured by inter-
val or ratio scales, while non-metric data are
only ordinal scaled. The dependent variables are
those which can be explained by other variables,
while interdependent variables are those which
cannot be explained solely by each other.
Marketing variables are usually interde-
pendent. For example, a firm’s objectives are
usually interdependent with marketing mix
variables; profits usually depend on sales;
market share depends on sales; firms’ growth
depends on profits and sales and vice versa, etc.
Also, firms’ marketing mix variables, such as
price, promotion, distribution and product, are
interdependent.
Since marketing research is very often a
multivariate analysis involving either depend-
ent or interdependent variables, the major
groups of techniques that can be used are as
shown in Figure 9.1.
1 Multivariable methods. So called because the
various techniques attempt to investigate the
relationships and patterns of marketing
decisions that emerge as a result of the
interaction and interdependence among main
variables at the same time.
2 Regression, correlation and forecasting
techniques.Regression and correlations
are methods that can be employed in
inferring the relationships among a set of
variables in marketing. Forecasting
methodsare mainly applied in forecasting
sales and market demand. Sales forecasting
methods are a function of an aggregation of
non-controllable environmental variables andmarketing effort factors, which have to be
taken into consideration.
3 Simulation methodsare a group of techniques
which are appropriate to use when the
variables affecting the marketing situation
(such as competition) require complex
modelling and are not amenable to analytical
solutions. The importance of the simulation
technique in marketing is that it offers a form
of laboratory experimentation by permitting
the researcher to change selected individual
variables in turn and holding all the others
constant.
4 Fuzzy setscould be used for modelling
consumer behaviour, marketing planning, new
product testing, etc., by determining the rank
and size of the possible outcomes.
5 Artificial intelligence (AI) techniquesare relatively
very recent tools for simulating human logic.
There are two main models in this set of
techniques: expert system – requiring user
intervention to accommodate changes within
the model; and neural network – less ‘rigid’
than expert system, facilitating ‘retraining’
(mainly via addition of new input and output
data).
6 Statistical decision theory or stochastic methods
represent stochastic or random responses of
consumers, which allow a multitude of factors
that might affect consumer behaviour to be
included in the analysis. This means that
market responses can be regarded as
outcomes of some probabilistic process.
Essentially, there are two main uses of these
methods: to test structural hypotheses and to
make conditional predictions.
7 Deterministic operational research methodsare
OR techniques looking for solutions in cases
where there are many interdependent
variables and the research is trying to
optimize the situation. A classical example of
such a situation in marketing is when a
company producing various products (or
parts) is selling them through two different
channels which vary with respect to selling
costs, typical order sizes, credit policies,
profit margins, etc. Usually in such cases the