Abstract
This paper reports findings from a project comparing retail buying behaviour
in Poland and Germany. The study demonstrates several differences in the way
listing decisions are made in the two countries – differences which at the same
time raise problems and offer opportunities for small and medium-sized
suppliers.
The Research Question
Since the fall of the wall nearly ten years ago, western producers have
held an open eye on the opportunities of the new East European markets, at the
beginning perhaps a sceptical eye. Now, however, with some of the former
Comecon-countries such as Poland and Hungary on the threshold of EU, these
markets have become closer, and West European investments in the ‘new’
EU-markets are growing.
This is also true for several
western producers in the food industry, who are looking for sales opportunities
in Eastern Europe.
However, as is well known, with the
concentration in the retail trade, the retail chains have become the
gatekeepers to the consumers: The producer has to sell his product two times,
first to the trade buyer, and then to the consumer.
The large producers, who have had to adapt to
new trends – e.g. green products and the retail chains growing preference for
producers willing to engage in long-time relationships, have of course
recognized this.
However, the question is: Can a producer use
experience earned on West European markets when trying to enter East European
Markets?
The Project
A study concerning food retailers’ buying behaviour is presented. The
study compares the retail buyers’ buying behaviour towards food products in
Germany and Poland. These two countries have been selected for several
reasons:
1.
Germany is one of the largest (if not the largest) importer of food
products in western Europe.
2.
Poland is the market, which at present attracts most interest from western
food producers.
3.
Besides, the two countries are both members of the North European food
culture (as opposed to the Mediterranean food culture) which should make a
comparison more meaningful.
In studies of the impact of various factors on
trade buyers' listing of new products/suppliers a broad spectrum of data
collection
methods could be used. At one extreme we find
the realistic but time consuming mapping of all buying decisions made in a
retail chain in the course of a year (e.g. Rao & McLaughlin 1989). At the
other extreme we find the one-shot interview, where the buyer is asked to
weight the decision criteria on a five- or seven-point scale (e.g. Wall et al.
1994).
The first-mentioned method requires close
co-operation with the chain, and the number of chains will necessarily be very
limited. While the latter method is easy to perform it is of course unrealistic
and prone to after-rationalization and ‘acceptable’ answers.
Conjoint analysis (e.g. Wagner et al. 1989)
seems to be a sound compromise between the two extremes: it is much more
realistic than the use of self-explicated weights, and at the same time opens
up for using enough respondents to generalize the results. Besides, the use of
an orthogonal experimental design avoids co-linearity problems.
It is assumed that the trade buyers are
expressing the retail chain’s buying policy of the particular food product
category when they evaluate the conjoint cards. At the same time it is assumed
that the intentions they express when they mark how likely it is that they will
buy the product from the vendor described on the cards, are the immediate
determinant of their behaviour in connection with choice of products and
suppliers. It is to say that their behaviour is under volitional control and
thus predictable from intentions. According to Ajzen and Fishbein (1980) an
important factor when evaluating the strength of the relationship between
intention and behaviour is the degree of correspondence between the measure of
intention and the behavioral criterion. To strengthen this relationship the
trade buyers were asked to evaluate cards describing the type of products they
bought ‘most of’ or ‘second most’. In that way they were only asked to evaluate
product types with which they were very experienced. This should ensure that
the measured intention, i.e. the measure of the indicated likelihood of making
a purchase, will lead to a very accurate prediction of buying behaviour, at
least at the aggregate segment level where the idiosyncratic events on the
whole are assumed to balance out.
For each profile, the trade
buyers were asked to indicate the likelihood of making a purchase of the
product from the vendor described. In this connection it is important to realize
that the evaluations take place in a comparison context. The trade buyers
compare the product and supplier attributes with what they are buying at that
point in time and in comparison with what they see of available products and
suppliers. It is worth noting that the evaluation of the cards depends on the
trade buyers’ decision environment i.e. the task environment interpreted by
each trade buyer (Schommer 1995) and on his assessment of competing products
and suppliers. Therefore questions about the respondents’ current suppliers
are included in the line of background questions asked about the trade buyer
and his organization. The various variables influencing the utility function of
the buyer can be grouped as follows:
1.
Variables characterizing the
organization of the chain: chain retailer, co-ops etc. and number of outlets.
2.
Variables characterizing the
face of the chain as seen from the customer: department store, supermarket,
discount store? etc.
4.
Variables characterizing the
organization of the buying process: existence of a buying centre, number of
members? etc.
5.
Variables characterizing the
buyer: education, experience, sex? etc.
Choice of Variables
Based on literature and interviews, the following
product and supplier attributes were included in the analysis (for pork, not
ranked):
1.
Quality of product
2.
Product price
3.
Consistency of product quality from one delivery to the next
4.
Whether the supplier’s product is developed on the basis of market
information from consumers in the buyer’s country
5.
Whether the producer can guarantee acceptable breeding and feeding
conditions for the animals
6.
Whether the producer is able to supply sufficient quantities to meet
the whole chain demand for his products
7.
The level of the producer’s support for advertising, and in-store
promotion of the product
8.
The producer’s ability to supply a broad range of cheese products
9.
Whether the producer is interested in developing a long-term
relationship with the chain
10.
The producer’s reputation among retailers
11.
Whether the producer is national or foreign with or without
representation in the buyer’s country
In short,
the framework used as a basis for this study is shown in table 1.
Data Collection
There is an ongoing discussion on how many attributes a respondent is
able to judge in the same go with an acceptable degree of certainty.
While there is a general belief that on the consumer market the limit
is about 5-6-7, it is generally believed that in the case of professional
decision-makers the limit should be higher.
The German data was collected in connection with an earlier project
focusing on segmentation of European retail chains (Skytte & Blunch 1997).
At that time – to play it safe – we decided to use bridging.
Bridging means that the conjoint job was broken up into two separate
conjoint jobs consisting of seven and six attributes respectively. As the total
number of attributes was eleven, this means that two attributes appeared in
both jobs and so to speak served as a bridge, which made it possible afterwards
to combine the results of the two separate jobs.
We bridged the attributes
‘quality’ and ‘price’ because if one of these attributes were missing in a
conjoint job, the respondent would most certainly use the included attribute as
an indicator of the other one. If both were missing the respondent would base
his evaluations on assumptions about the levels of these two attributes.
Therefore, we had no other choice than to bridge on these two attributes.
However, it turned out that for several respondents these two attributes
played so small a role compared to some of the other attributes, that the
estimated utilities could be considered the result of a random process, and
therefore they did not correlate well between the two sets of cards.
This problem deserves special mention, as part
of the reason for doing that project was that we thought that other attributes
than the traditional four Ps (Product, Price, Promotion and Place (i.e. distribution)) were gaining
in importance. It can be said that the failure of the two factors to bridge
supported this hypothesis. However, from a technical point of view the
randomness connected with the evaluation of the bridging factors was a problem.
This randomness appeared as very small
differences in utility for the various levels of the attributes at the same
time as people preferred high prices to low ones (for the same quality) and/or
low quality to high quality (at the same price).
In order to reduce this
problem, the utility functions were re-estimated for this study by using a
non-linear least squares optimization algorithm, placing order restrictions on
the various coefficients.
As the levels for most of the attributes are
logically ordered we placed order restrictions on all, but the following:
1. Support for
promotion (perhaps this seems a little surprising, but the reason is given
later)
2. Willingness to
engage in long-time relationships
3. National/foreign
As a respondent’s breaking of
a natural order generally was connected with low weight on the attribute in
question, this procedure will most certainly increase the predictive validity
of the study.
Only 1 of the 353 returned (German)
questionnaires were not bridged due to lack of variation in the utilities of
the bridging factors. We decided, however, to use only bridged questionnaires,
where:
1. R squared (between
observed and predicted utility for a concept) was higher the 0.30 for both sets
of cards
2. The correlation
between the evaluations of the bridging attributes in both sets of cards was
higher than 0.30
3. The average ratio
between the utility for the for the same attribute levels for the bridging
attributes was in the interval from 0.3 to 3.0
It could be argued that these conditions are
too restrictive and in fact they reduced the number of ‘successful’ bridgings
to 177 compared with the 353 returned (German) questionnaires. It must,
however, be emphasized that an analysis of the connection between the
background variables and the evaluation of the crucial attributes does not
necessitate a bridging, but can be performed on the separate card sets. Such
an analysis can be based on 343 questionnaires fulfilling the first restriction
(a loss of less than 3 pct.). So we see that the problem is not with the
original conjoint jobs, but alone with bridging. And we will try console
ourselves by keeping in mind that the relative failure of the bridging supports
our point, that quality and price are indeed loosing relative importance as decision
criteria.
When we collected the Polish data in 1999 we decided not to bridge, as
recent research has shown that it should be well within professional decision
makers ability to judge eleven attributes in one conjoint job (Pullman et al.
1999). However, the same non-linear
least squares optimization algorithm, with the same order restrictions was
used.
We are well aware that the use of bridging in one country and not in
the other can be criticized. However, so could the decision to use the same
(less effective) data collection method in the case of Poland, given our
experience from the first project, and the insight received from Pullman,
Dodson & Moore (1999).
We received 50 questionnaires from Poland, of which four were discarded
because of obvious interviewer cheating and three because they did not fulfil
the conditions R squared (between
observed and predicted utility for a concept) greater than .30. (In fact only
two Polish R squared accepted was below .60).
Mapping the utilities for the two countries
In tables 2 and 3 we find the
average utility functions for the two
countries. We observe that the
part-utilities – or part-worths as they are usually called – are scaled in such
a way that they sum to zero for each attribute. These part-worths should be
interpreted as the additional utility of a product with the attribute in
question. If we look at table 2 we find that for example ‘premium quality’
(for pork) increases the utility of the product with 0.62 points as measured on
an eleven-point scale similar the one used in collecting the data, whereas
‘average quality’ reduces the utility by 0.71 points. The price vector
indicates that for each per cent the price is raised, the utility will be
reduces by 0.12 points on that same scale.
The relative importance (or
RI) of an attribute is defined as the numerical difference between the highest
utility and the lowest utility for an attribute, divided by the sum of these
differences across all attributes:
where the symbols 
should be self-explanatory.
The relative importance can be calculated for the utility function of an
individual respondent or for an average utility function covering several
respondents.
Even if the use of average
utility functions is quite common in research, they are usually misleading,
because the distribution of the various part-worths across the respondents
generally is very skew. This is also the case in our data, and consequently we
prefer using median values to describe utility functions for groups of
respondents.
As most of the attribute
levels are logically ordered, the RIs carries about the same information as the
utilities, and are simpler to grasp, so the following arguments will have table
4 as its basis instead of tables 2 and 3.
What first strikes the eye
when looking at the German data in
table 4 is the similarity of the utility functions for the three products: The
three attributes with the highest relative importance (RI) are – for all three
products – ‘traceability’, ‘sufficient quantities’ and ‘long-term
relationship’. These attributes in all three cases account for about 50% of the
weight. The RIs for other attributes also seem quite similar across the three
products.
We also observe the rather
modest roles played by three of the traditional ‘four P’s’: Product, Price and Promotion. This is just a reflection of
general trends in Western Europe (Skytte & Blunch 1997). As far as quality and price are concerned the reason is not that
these factors are unimportant. A more reasonable explanation is the existence
of certain standards which the supplier should live up to in order to be taken
into consideration – standards which are taken for granted by the buyer as well
as the seller. With regard to promotion the explanation is perhaps a different
one, as we know from our interviews that many retailers prefer to run this
activity on their own with as little interference from the supplier as
possible.
So the message for a
potential supplier is not that the traditional factors quality and price are
without importance, but rather that you cannot use these attributes to
differentiate yourself from your competitors. As a means for differentiation
and preference you have to use the ‘new’ attributes ‘traceability’, ‘sufficient
quantities’ and ‘long-term relationship’.
In the Polish data we also observe a marked similarity between the utility
functions for the various products, but to a lesser degree than in Germany.
The two attributes ‘long-term
relationship’ and ‘national/ foreign’ are among the ‘top-three’ attributes in
all three cases. ‘Sufficient quantities’ joins these attributes in the cases of
pork.
Comparing German and Polish trade buyers
As should be clear from the remarks above, the differences between the two countries seem to be
larger than the differences between the utility functions for the three
products within each country.
At first sight it seems as if two of the ‘new’ factors, which play a
large role in Germany (and indeed in Western Europe as a whole), namely
‘traceability’ and ‘sufficient quantities’ play a much smaller role in Poland,
whereas the third ‘new’ factor ‘long-term relationship’ is a crucial factor in
both countries.
On the other hand, the supplier’s reputation – which receives near zero
weight from the German trade buyers – plays a larger role for the Polish trade
buyers.
Also, it is apparent that the preference for doing business with a
supplier from one’s own country (in the case of pork and fish) or at least with
a foreign supplier who is represented in one’s country (in the case of cheese)
is crucial to the Polish trade buyers. This last factor is of less importance
to the German trade buyers.
There also appears to be some differences between the two countries in
the weights assigned to ‘wide range’ (for cheese), and you can perhaps also
find a few other differences. However, differences in factors which in both
countries receive relatively little weight are not very interesting from a
management point of view, and also the differences in average or median RI
should be evaluated in the light of the spread around the mean or median within
each country.
The traditional use of analysis of variance for this purpose is,
however, not feasible in this situation, because ANOVA is a test of averages,
and we are dealing with medians. Besides, the ‘P-values’ of an ANOVA (ore any other statistical test for that sake
– including the ordinal counterpart of ANOVA) has only limited value in this
study, firstly because the data is not two random samples, but rather two
incomplete censuses, and secondly because P-values
are not measures of the size of
influence as they depend on the amount of data.
Therefore the more ‘primitive’ descriptive methods cover our purpose
better.
Why are there differences between the two countries?
In order to explain the differences found in the last section, we must
use the background variables shown in table 1. Among these is the size of the
chain as measured by the variable ‘number of outlets’.
Now, the Polish chains are generally slightly smaller than the German
ones, and one could hypothesize that this is the reason for some of the
differences found between these two countries. However, an analysis shows that
the RIs for most of the attributes are relatively independent of the size of
the chain.
The ‘missing’ (or at least minimal) influence of the variable ‘number
of outlets’ is perhaps a surprise, because one should expect e.g. the
preference for ‘sufficient quantities’ to be a growing function of the size of
the chain, since the largest chains must have met this problem most often. The
same could be said with regard to the preference for suppliers who are
interested in establishing long-time relationships, since the larger chains
should profit most by co-operation on product development etc. – and should
also have more resources to put into a co-operation.
With regard to the last point, one could also argue that the limited
resources of a small chain would make it desirable to have a fixed supplier and
thus minimizing the cost of establishing new supplier relations. So, whereas
the preference for suppliers who are willing to engage in long-time
relationships is relatively independent of chain size, the motives for such preferences are likely to vary.
Indeed, if we plot the preference for suppliers interested in long-time
relationship against the size of the chain, we observe a weak increase in preference with size of
chain in the German data and a weak decrease
in the Polish data. Is it too bold to hypothesize that this difference is
caused by different motives in the two countries?
Now let us take a closer look at the attributes which receive different
weight in Poland and Germany.
Traceability
We found that when the product is bought by a German trade buyer fresh
(instead of e.g. chilled, frozen or canned), the weight placed on traceability
on average goes up with a factor 3.00, and the same is true if the buyer is a
female as opposed to a male. So perhaps differences with regard to these two
factors could explain the difference; but no: In Poland the same (low) interest
is shown by male and female trade buyers and independently of the form in which
the product is supplied. In fact, we were unable to connect any of our many
background variables with the different interest in ‘traceability’ in the two
countries.
Sufficient quantities
The preference for suppliers who are able to supply enough to fulfil
the demand of the whole chain is very weak in the Polish data (except perhaps
for pork) as opposed to the German data. Perhaps the chains in Poland are not
quite so ‘streamlined’ with regard to carrying the same selection in all their
outlets and therefore rely more on local suppliers. If this is true it could
also explain the lack of interest in ‘traceability’ and would go well with the
expressed preference for doing business with local firms.
Wide range (for cheese)
It is perhaps not very surprising that the preference for suppliers who
can supply a wide range of products is most prominent for delicatessen such as
cheese (in Poland), because this is only a reflection of the buying habits of
the consumer. I suppose that the typical ‘cheese lover’ is more inclined to
sample various cheese variants than to just stick with one variant only. But
why is this difference between the three products not found in German data? Is
it because the Germans are more experienced in handling business connections
and therefore do not find it problematic to handle a large number of
suppliers? Or is it because the Germans are looking for more ‘special’ and
higher priced cheeses and therefore have to go to suppliers who specialize in
one type of cheese only?
Reputation (for all three products)
An obvious hypothesis is that the smaller the chain, the less resources
it has for judging the ‘quality’ of the supplier, and the more probable it is
that it will resort to the supplier’s ‘general reputation’ as an indicator of
that ‘quality’. Consequently the impact of ‘reputation’ should be a decreasing function of the size of the
chain.
Another – just as plausible hypothesis – is that the more outlets, the
more vulnerable the chain is, if it should engage itself with a supplier whose
conduct or products are below what is acceptable. Consequently the impact of
‘reputation’ on the decision to engage with a supplier should be an increasing function of the size of the
outlet.
However, as mentioned above, no simple relationship between the two
variables exists. When we plot the weight placed on ‘reputation’ (for Polish
trade buyers) against the number of outlets, the curve decreases at first (as
it should according to the first hypotheses), but only up to a number of outlet
around twenty. Then it increases rather fast. Is there a ‘critical number’ (in
the vicinity of twenty) above which the second hypothesis takes over?
On the German data this function decreases weakly all the way, but this
observation is not of very great interest in itself, because the curve hardly
raises above the X-axis.
National/foreign (for pork and cheese)
One hypothesis which comes easily, is that the preference for doing
business with a native firm (or at least a firm which is represented in the
country), which dominate the Polish data, is that the Polish firms do not have
the experience in doing business with foreign companies as do e.g. the German
chains.
One should expect that the larger the chain, the more experience in
doing business with foreign firms and the more resources for ‘buying’
experience and skills, and consequently the less weight on this attribute.
However, a plot of these two factors against each other shows no co-variation
at all.
Conclusions, further questions and consequences
To sum up: Whereas German trade buyers have shown an increased
interest in some ‘new’ product and supplier attributes such as ‘traceability’,
sufficient quantities’ and ‘long-time relationship’ at the expense of more
‘traditional’ attributes such as ‘the four Ps’, this West European trend is not
observable in Poland.
Instead the Polish trade buyers place heavy weight on doing business
with suppliers who are at least represented in Poland and they also rate the
reputation of the supplier higher than their German counterparts. At the same
time they rate the supplier’s interest in long-time relationships as nearly as
important as the German trade buyers do – but perhaps for different reasons.
As we have not been able to explain the differences found in preference
functions by differences in background variables between the two countries,
the reason must be placed in ‘the country as such’.
The pattern we see in the utility functions of the Polish trade buyers
seems to us to show a retail culture where the chains are less ‘streamlined’,
have less experience in handling (many) foreign suppliers and (therefore)
perhaps are more locally focused in their search for suppliers. If Poland seems
to be ‘Germany minus 40 years’, how many years will it then take to catch up?
What does that imply to the foreign supplier seeking opportunities on
the Polish market?
It means that not much of his experience earned in Western Europe can
be used. It means that small or medium-sized suppliers are forced to
co-operate horizontally in order to establish sales offices in Poland. A
special problem for such a supplier is that he is generally less known than
larger firms, and as reputation means a lot to the Polish trade buyer, perhaps
getting the first few customers could demand some resources.
On the other hand it also opens up some avenues for this type of
suppliers – especially perhaps suppliers of more ‘traditional’ products, who
all too often have seen promising businesses with West European chains go down
the drain, because they could not supply ‘green products’ or quantities large
enough to cover the demands of the whole chain.
References
Ajzen, I & Fishbein, M.
(1980), Understanding Attitudes and Predicting
Social Behavior, Englewood Cliffs N J, Prentice Hall.
Pullman, M.
E., Dodson, K. L. & Moore, W. L. (1999), ‘A Comparison of Conjoint Methods
when there are Many Attributes’. Marketing
Letters 10(2): 125-138.
Rao, V. and
McLaughlin, E. (1989), ‘Modeling the Decision to Add New Products by Channel
Intermediaries.’ Journal of Marketing
53(1): 80-88.
Schommer, J. C. (1995),
‘Definition and measurement of environmental uncertainty in channels of
distribution research: A proposal and pilot test’. Journal of Marketing Channels, 4(3) 53-73.
Skytte, H.
& Blunch, N. J. (1997), ‘A Conjoint Analysis of Retailers Buying Behaviour
of Fish Products in 14 European Countries’. Scott M Smith (Ed.): Sixth
Symposium on Cross-Cultural Consumer and Business Studies, Honolulu, Hawaii
10-13 December, 1997.
Wagner, J., R.
Ettenson, et al. (1989), ‘Vendor Selection among Retail Buyers: An Analysis by
Merchandise Division.’ Journal of
Retailing, 65(1): 58-79.
Wall, M. et al.
(1994), ‘The Retail Buying of Fashion Goods: Underlying Themes of the Sourcing
Process,’ International Review of Retail
Distribution and Consumer Research 4(2): 177-193.