Marketing research 189
scores may then be compared with other
individuals with respect to the same object, or
two or more objects may be compared with
respect to the same individual or group of
individuals.
Profile analysis– involves calculating the
arithmetic median or mean value for each pair
of adjectives for an object for each respondent
or respondent group. The profile so derived
can then be compared with the profile of the
object.
The principle disadvantage of semantic differ-
ential scales lies in their construction. For valid
results, scales need to be made of truly bipolar
pairs of adjectives/phrases; the problem arises
when some of the pairs/phrases chosen may
not be true opposites in the respondents’
minds.
The Stapel scale
This is a modified semantic differential scale,
and uses a unipolar 10-point verbal rating scale
with values from +5 to –5 which measures both
the strength and direction of the attitude
simultaneously.
For example, respondents are asked to
evaluate how well each of the adjectives
describes the object under test, e.g. Sheila’s
apple pie tastes:
+5 +5 +5
+4 +4 +4
+3 +3 +3
+2 +2 +2
+1 +1 +1
Rich Bitter Expensive
–1 –1 –1
–2 –2 –2
–3 –3 –3
–4 –4 –4
–5 –5 –5Stapel scales are easy to administer and
require no test that the adjectives are truly in
polarity.
Sampling
Without the ability to extract a sample of a
population, as opposed to conducting a census,
the majority of marketing research projects
could not take place. There are four main
reasons for this:1 Cost– except where the populations are
very small, it is usually cheaper to take a
sample rather than conduct a census in the
same population.
2 Time– a census, compared with a sample for
a given population, is always going to be
larger (hence it will take longer to collect
the results); thus, by the time the results
have been collected and analysed, the
situation under investigation might have
changed. Samples may be extracted and
analysed much quicker than a census, for a
given population.
3 Accuracy– defined as the degree of precision
with which a measure of a characteristic in a
sample compares with the measure of the
same characteristic in the population from
which the sample was drawn. In sampling,
accuracy is affected by: (1) Sampling error
(caused by selecting a probability sample
from a population which is not representative
of that population); such error can be
reduced by increasing the size of the sample.
(2) Non-sampling error (all other errors in a
marketing research project whose origin is
not based in sampling error).
4 The destructive nature of measurement– one
cannot carry out a census on a bottle of
whisky (for quality control) and still have any
product for sale. Thus, sampling is the only
alternative if quality assurance is required.
Some forms of measurement destroy; for
example, one can only measure a
population’s initial reaction to an
advertisement once. But, by extracting
non-overlapping samples, such an evaluation
may be repeated.