Content uploaded by Koen Vanhoof
Author content
All content in this area was uploaded by Koen Vanhoof
Content may be subject to copyright.
Identifying latently dissatisfied customers and
measures for dissatisfaction management
JoseÂe Bloemer
Department of Applied Economics, Limburg University Centre, Diepenbeek, Belgium
Tom Brijs
Department of Applied Economics, Limburg University Centre, Diepenbeek, Belgium
Gilbert Swinnen
Department of Applied Economics, Limburg University Centre, Diepenbeek, Belgium
Koen Vanhoof
Department of Applied Economics, Limburg University Centre, Diepenbeek, Belgium
1. Introduction
This paper deals with the problem of
identifying latently dissatisfied customers.
We define latently dissatisfied customers as
customers who when asked, report overall
satisfaction, but who possess other
characteristics that are strongly associated
with dissatisfaction. Therefore, these
customers have a high probability to defect.
We demonstrate the effectiveness of a data
mining technique, called characteristic rules,
for identifying these customers by using
secondary data from a large-scale customer
satisfaction survey carried out by a leading
Belgian bank. Moreover, we show how
actions upon these customers can be taken in
order to prevent them from defecting.
This article is organized as follows. First of
all, section 1 continues with a concise
overview of the relevant literature of
dissatisfaction in a banking context. In
section 2, we will introduce the concept of
latent dissatisfaction and provide a visual
representation of latently dissatisfied
customers. Subsequently, in section 3 we will
elaborate on the three-step methodology of
partial classification and focus on the data
mining technique of characteristic rules to
identify latently dissatisfied customers. In
section 4, the proposed data mining
technique will be applied to the empirical
data after which results will be shown and
validated. Section 5 presents some concrete
suggestions for dissatisfaction management.
Finally, section 6 presents the conclusions
and limitations of this study.
1.1 Customer loyalty and defection in the
financial services sector
During the past decade, the financial services
sector has undergone drastic changes. This
has resulted in a market in which intense
competition, little growth in primary
demand and increased deregulation have
become important characteristics. In this
new market, the occurrence of committed
and often inherited relationships between a
customer and his or her bank is becoming
increasingly scarce (Levesque and
McDougall, 1996). Therefore, several
strategies have been followed to retain
customers, in which satisfaction plays a
pivotal role (Meidan, 1996). Satisfied
customers provide recommendations for the
bank, especially those who state that they are
very satisfied. Satisfied customers will pay a
premium for services and reduce a bank’s
cost of providing services because there are
fewer complaints to deal with. Moreover,
satisfied customers are more likely to
concentrate their business with one bank and
to respond to cross-selling efforts. Winstanley
(1997) clearly shows that customer
satisfaction is not only linked to loyalty as
such, but is also linked to bank revenue
generation by means of the above mentioned
ways.
For instance, Reichheld and Sasser (1990)
and Reichheld (1993) also argue that higher
levels of customer satisfaction lead to higher
levels of customer retention. In turn,
customer satisfaction and retention drive
customer revenue and the cost of doing
business and, ultimately, are key factors in
the profitability of a business (Federal
Express, 1992). Waterhouse and Morgan
(1994) describe how the Lloyds Bank (UK)
tackled the issue of customer retention. They
used qualitative (in-depth interviews and
group discussions) as well as quantitative
research methods (telephone and mail
surveys) among both staff and customers
(defectors, dormants and dissatisfieds) to
investigate the extent of defection due to
dissatisfaction with the bank and to explore
the process and factors leading from
h t t p : / / w w w . e m e r a ld i n s i g h t . c o m / r e s e a r c h r e g i s t e r s h t t p : / /w w w . e m e r a l d i n s i g h t . c o m / 0 2 6 5 - 2 3 2 3 .h t m
[ 27 ]
International Journal of Bank
Marketing
20/1 [2002] 27±37
#MCB UP Limited
[ISSN 0265-23 23]
[DOI 10.1108/0265232 02104 15962]
Keywords
Customer satisfaction,
Customer orientation,
Data mining, Banking
Abstract
Customer satisfaction continues
to be an important topic in the
financial services industry.
However, there is an increasing
awareness that customer
satisfaction as such is not enough.
Distinguishes between overall
satisfied customers and latently
dissatisfied customers; the latter
being those customers who,
although reporting satisfaction in a
survey, have other characteristics
(i.e. satisfaction with specific
service items and/or socio-
demographic characteristics) that
resemble dissatisfied cus tomers.
The identification of these latently
dissatisfied customers may
function as an early warning
signal. Indeed, their probability to
defect is relatively high and can be
compared to that of dissatisfied
customers. Proposes a data
mining technique called
``characteristic rules’’ to identify
latently dissatisfied customers of a
Belgian bank. Appropriate
marketing actions (dissatisfaction
management) may help to avoid
these customers leaving.
Therefore, the objective of this
study is to provide scholars and
business managers with
theoretical, methodological and
managerial insights into
identifying latently dissatisfied
customers.
dissatisfaction to account closure. As a result
of this research, the bank was able to design
and implement a new customer retention
process. An interesting finding of the study
was that dissatisfaction and defection were
rarely attributable to just one factor. It was
more the interplay of several elements ±
including bank policy, staff handling and
customer characteristics ± that eventually
provoked customers to become dissatisfied
and defect.
In accordance with this, it has also been
shown that the relationship between
satisfaction and loyalty is neither linear nor
simple (Jones and Sasser, 1995). High levels
of measured satisfaction sometimes go hand
in hand with a continuous decline in
turnover (Heskett et al., 1994) or an increase
in customer defection (Reichheld, 1996). In
consumer markets, where changing styles
and impulse buying are observed, variety-
seeking behaviour is often put forward as a
cause for such a discrepancy (Van Trijp et al.,
1996). Within a banking context, the reasons
are not so clear. Traditionally, the retail
banking market has been characterized by
very strong customer inertia. However, this
situation is gradually changing (Waterhouse
and Morgan, 1994). Retail bank customers are
``shopping around’’ more than they used to.
Also, they may be customers at several
banks, depending on the products or services
that are offered by the bank (Yavas and
Shemwell, 1996; Worcester, 1997).
1.2 Determinants of satisfaction or
dissatisfaction
The issue in this article of distinguishing
latently dissatisfied customers from overall
satisfied customers requires the
determinants of satisfaction and
dissatisfaction to differ to some extent.
However, research on the determinants of
customer satisfaction often assumes that
there is no difference between the causes of
satisfaction or dissatisfaction. Several
authors, in contrast, suggest that there are
some determinants that tend to be primarily
a source of satisfaction and others that tend
to be primarily a source of dissatisfaction
(Cadotte and Turgeon, 1988; Hausknecht,
1988; Herzberg et al., 1959; Maddox, 1981;
Swan and Combs, 1976). For instance,
Vanhoof and Swinnen (1996) introduced a
method that enables the drawing of a
distinction between the impact of criteria on
satisfaction and dissatisfaction. Their results
indicated that there indeed exist differences
in the direction and magnitude of the impact.
In a study of the bank sector, Johnston
(1995) used the critical-incident technique
(see also Bitner et al. (1990) for the use of this
method in a service setting) to classify
customer perceptions (anecdotes) into
satisfying and dissatisfying factors. While
most determinants were found to be a source
of both satisfaction and dissatisfaction (but
ranking differently with respect to impact)
there were a few (four out of 18) that
exclusively determined either satisfaction or
dissatisfaction with the bank. Other studies
confirm these findings (Chakravarty et al.,
1996, 1997).
Identifying latently dissatisfied
customers
When satisfied and dissatisfied customers can
be clearly identified in terms of a number of
characteristics, like their satisfaction with
particular service items or socio-demographic
characteristics, this provides the ability to
identify overall satisfied customers who
possess typical characteristics of overall
dissatisfied customers. These customers can
be defined as latently dissatisfied, even
though they report overall satisfaction.
Figure 1 illustrates the conceptual idea of
identifying latently dissatisfied customers in
the realm of a data mining technique.
Suppose we are able to identify a set of
characteristics that is strongly (but not
necessarily exclusively) associated with
dissatisfied customers and that we can find
another set of characteristics that is strongly
(but not necessarily exclusively) associated
with satisfied customers. Moreover, suppose
we would be able to define an ordering on
these characteristics such that the most
discriminative (exclusive) characteristics of
overall dissatisfied customers would be
located on the left of Figure 1 and that the
most discriminative (exclusive)
characteristics of overall satisfied customers
would be located on the right. Then, curve 1
represents the increase of the cumulative
proportion of overall dissatisfied customers
in a survey that is identified by the
characteristics of dissatisfied customers
when moving to the right in Figure 1. The
curve shows a diminishing marginal rate of
increase when we move to the right, i.e. the
area that is more characteristic for
satisfaction. In contrast, curve 2 represents
the cumulative proportion of overall satisfied
customers being described by the
characteristics of dissatisfaction when
moving to the area that is more
characteristic for satisfaction. Curve 2, in
contrast to curve 1, shows a growing
marginal rate of increase when moving to the
right in Figure 1.
[ 28 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
Now we are able to conceptually define the
group of latently dissatisfied customers.
Indeed, from Figure 1, it can be observed
that, for some customers, their overall
service satisfaction evaluation does not
correspond with the characteristics. This
means that some customers, like for instance
those situated in the hatched area, report
satisfaction although they are typified by
characteristics for dissatisfaction. Therefore,
the customers in this area are defined as
latently dissatisfied because they may be
highly vulnerable to become overall
dissatisfied customers in the near future.
Although at this point in the paper, we
have not yet precisely defined the concepts
``characteristics’’ and ``ordering’’, we will
give concrete form to these concepts when we
operationalize them in the next section,
discussing the methodology.
3. Methodology
3.1 A three-step methodology based on
characteristic rules
The methodology to identify latently
dissatisfied customers is based on the data
mining technique of characteristic rules.
Characteristic rules is a well-known
descriptive data mining technique to find a
concise description or summary of the
instances in a data class, or to find general
properties of the instances of that data class.
The properties of these instances are
expressed in a rule-based (IF ± THEN) format,
which is easy to interpret. For instance, the
characteristic rule IF object ˆswan THEN
colour ˆwhite AND species ˆbird
expresses that if an instance belongs to the
class of swans, then its colour is white and it
is a bird.
In this paper, we will use the technique of
characteristic rules to automatically infer
strong characteristic rules associated with
the group of dissatisfied customers.
Subsequently, these characteristic rules will
then be used to identify latently dissatisfied
customers.
In brief, the proposed methodology in this
paper involves three steps:
1 Discover characteristic rules of overall
dissatisfied customers (set 1) and of
overall satisfied customers (set 2) (see
section 3.1.1).
2 Rank individual rules in both rulesets
according to how exclusive these
characteristic rules are to each of the
classes of overall satisfied and overall
dissatisfied instances. Namely, some rules
will exclusively describe instances of one
class, without describing any instances of
the other class, whereas other rules may
equally describe instances of both classes
(see section 3.1.2).
3 Use the ranked characteristic rules in set
1 to identify overall satisfied customers
that are described by the rules in set 1,
and call them latently dissatisfied
customers (see section 3.1.3).
Discovering characteristic rules
The data mining technique to discover
characteristic rules is based on an algorithm
to find frequent sets of attributes (Agrawal
and Srikant, 1994) in a class of instances.
Given the satisfaction opinions of the
respondents on different banking services in
the bank’s customer satisfaction survey, the
discovery of characteristic rules in this study
produces a set of rules for each group of
customers (i.e. overall dissatisfied (set 1) and
overall satisfied (set 2)) that expresses typical
characteristics of the instances of each
group in terms of the opinions of the
respondents on the different banking
services. For instance, IF overall
dissatisfaction THEN I receive sufficient
information from the staff ˆvery
dissatisfied AND the staff is friendly ˆ
dissatisfied (sˆ60 per cent). The metric ``s’’
in this example is the support of a
Figure 1
Identifying latently dissatisfied customers
[ 29 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
characteristic rule within the class of overall
dissatisfied customers. It expresses that 60
per cent of the overall dissatisfied customers
have responded to be very dissatisfied with
the amount of information given by the staff,
and to be dissatisfied with the friendliness of
the staff. Two sets of rules of this kind are
generated, one set of rules for the group of
overall dissatisfied customers (set 1) and one
for the group of overall satisfied customers
(set 2). The support, however, does not reveal
anything about the probability to belong to a
certain class, given these characteristics. In
other words, given the characteristics in the
consequent of the rule above, it is not a priori
known what is the probability to be overall
dissatisfied [1]. Therefore, we need to rank the
rules.
Ranking characteristic rules according to
their exclusiveness
The second phase in the methodology
involves the ordering of characteristic rules
obtained from step 1, by means of a ranking
procedure. This is a necessary step because
one could say that characteristic rules
describe the instances of a data class but do
not (necessarily) discriminate between the
target groups. To put it in the context of this
study: the discovered rules may not only be
characteristic for overall dissatisfied
customers but they may be characteristic for
the entire group of customers, including
overall satisfied customers. To solve this
problem, a measure of exclusiveness can be
used to rank the rules from uniquely
characteristic for a particular target group
(high exclusiveness), to characteristic for
both target groups (low exclusiveness).
Identify latently dissatisfied customers
In the final step of the methodology, the rank-
ordered characteristic rules for overall
dissatisfied customers, obtained from step 2,
are matched with overall satisfied customers
in the database. Indeed, the main idea of
identifying latently dissatisfied customers
consists of discovering characteristics in the
data that are strongly associated with overall
dissatisfied customers (see step 1 and 2) and
to look for so-called overall satisfied
customers that match with these rules for
dissatisfaction.
4. An empirical study
4.1. Data collection and coding
The secondary data being used for this study
concerns a satisfaction survey that was
conducted among customers of a major bank
in Belgium in 1996. Nationwide, 7,264
customers of the bank filled out a
questionnaire. This questionnaire includes
questions probing for the satisfaction with
specific service aspects of the bank (see
Appendix 1), questions on socio-demographic
characteristics of the customers and a
question probing for the overall level of
satisfaction.
Customers were asked to indicate to what
extent they could agree with the statements
presented in the questionnaire. All
statements related to the bank’s service
aspects were measured on a five-level ordinal
scale with responses ranging from always (5),
most often (4), sometimes (3), rarely (2), to
never (1) and no opinion, the latter indicating
a missing value. Unfortunately, the use of
this five-level response scale represents a
potential danger for the case of characteristic
rule discovery, which involves looking for
frequent sets of attributes values. Indeed,
some opinions, such as never, rarely and
sometimes, which do not occur frequently
enough in this dataset, will not show up as
frequent sets in the rule discovery phase and,
as a result, they will not show up in the
characteristic rules. Although the frequency
of these opinions is of course specific for each
dataset, this is not an uncommon situation in
satisfaction surveys where most people
report high satisfaction values. Therefore, in
consultation with bank management,
response values on specific service items (the
independent variables in this study) were
recoded, as shown in Figure 2.
In contrast, for the target attribute (overall
level of satisfaction), response values were
recoded into 2 groups, combining always and
most often into satisfied and sometimes,
rarely and never into dissatisfied.
Eventually, a total of 7,264 instances were
obtained of which only 445 (6.1 per cent) were
classified in the group of dissatisfied
customers, again illustrating the skewness of
the class frequency distribution.
4.2. Data analyses
The identification of latently dissatisfied
customers follows the three-step
methodology introduced in section 3.
Step 1
For both groups of overall satisfied and
overall dissatisfied customers, all
characteristic rules with a minimum support
of 20 per cent were discovered. The support
threshold indicates that at least 20 per cent of
the instances in the target group should
possess the characteristics contained in a
rule. In fact, setting the correct value for the
minimum support parameter is important,
since setting it too low results in overfitted
patterns (describing only few instances),
whereas setting it too high results in missing
[ 30 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
some important segments within the target
class of overall dissatisfied customers.
Therefore, different support thresholds
(ranging from 5 per cent to 30 per cent) were
first tested. The correct setting of the
minimum support threshold is largely
dependent on the given dataset. However,
lowering the support threshold too much
produces overfitting, whereas setting it too
high results in important information not
being discovered, both causing a bad
performance on unseen data. Somewhere in
between lies the optimal support value, and
finding it is basically a process of trial-and-
error.
The following sample provides an
illustration of some discovered rules of
overall dissatisfaction:
R1: IF overall dissatisfied THEN Question
1= dissatisfied ^Question 15 ˆ
dissatisfied [sˆ21 35 per cent]
Ri: IF overall dissatisfied THEN Question
4 = dissatisfied ^Question 7 ˆ
dissatisfied [sˆ37 53 per cent]
For instance, rule 1 should be interpreted as
follows: 21.35 per cent of the respondents who
indicated to be overall dissatisfied reported
dissatisfaction with question 1 (Q1 ˆIn my
bank office, leaflets are available with all
necessary information) and question 15 (Q15
ˆI get enough information from my bank by
means of correspondence).
This information is very useful because it
describes some typical characteristics of an
overall dissatisfied customer. However, as
already indicated in the methodological
section, one must be careful with the
interpretation of these results. The
discovered rules for dissatisfaction may be
characteristic for the whole dataset too.
Step 2
Therefore, after ranking the characteristic
rules according to their exclusivity to a
particular target class, they have to be
ordered from highly unique for overall
dissatisfied customers, to highly unique for
overall satisfied customers on the x-axis of
Figure 3. As a result, for the discovered
characteristic rules under the given support
settings, the cumulative proportion of
satisfied and dissatisfied customers can be
plotted as illustrated in Figure 3.
One particular point of interest is shown in
Figure 3 on the dashed curve (2) representing
the cumulative proportion of overall satisfied
customers. Visual inspection of this curve
reveals a clear breakpoint at the 29th rule of
dissatisfaction (see white arrow) where
suddenly the additional number of overall
satisfied instances covered, by introducing
additional (less exclusive) characteristic rules
for overall dissatisfaction, increases rapidly.
This is a clear indication that the optimal set
of characteristic rules for overall dissatisfied
customers contains just 29 rules. Since
starting from characteristic rule number 30
we tend to cover a rapidly increasing number
of overall satisfied customers.
Step 3
The third and final step in the methodology
involves the matching of (highly unique)
characteristic rules for overall
dissatisfaction with overall satisfied
instances in the database. Thus, taking the 29
most exclusive characteristic rules for
overall d issatisfaction, this set of rules covers
73.9 per cent of the dissatisfied customers and
10.7 per cent of the customers reporting
overall satisfaction but who are, in fact,
latently dissatisfied. These latently
dissatisfied customers are being identified by
the hatched area (3) in Figure 3. The
characteristics of these customers will be
described in section 5.
4.3. Validity of the results
4.3.2. Internal validity
To evaluate the stability of the selected rules,
the results of a second, but identical
questionnaire (consisting of 31,970 customer
surveys) carried out in 1997, have been used.
It was observed that under the same support
setting (20 per cent), 27 of the 29 interesting
rules discovered in 1996 were still valid in
1997, indicating a high stability of the
selected ruleset. For the two rules that were
negatively validated, support was slightly
insufficient. Therefore, one can conclude that
27 of the 29 rules for dissatisfaction
discovered in the 1996 analysis are highly
consistent over time.
4.3.3. External validity
To assess the external validity of the model,
two tests are carried out. From the literature
Figure 2
Recoding of response values
[ 31 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
it is known that the number of complaints
formulated by the customer is a valid
indicator for the level of dissatisfaction of that
customer (Day, 1984; Technical Assistance
Research Programs (TARP), 1986; Fornell and
Wernerfelt, 1987; Heskett et al., 1997).
Therefore, as a first test of validity, we use the
number of complaints as a measure of
criterion validity of the discovered model. The
second validity test concerns the analysis of
the defection rate of each group of customers
(dissatisfied, satisfied and latently satisfied).
Test 1: analysis of complaints. Specifically,
for each customer in the survey, the number
and type of complaints he or she submitted in
1997 were obtained from his/her survey (see
Appendix 2). Then, for each group of
customers, i.e. dissatisfied, satisfied and
latently dissatisfied, the number of
complaints (in percentages) is plotted, as can
be seen in Figure 4.
The first bar of Figure 4 represents
dissatisfied customers who are covered by
our model, i.e. they are considered as
prototypical examples of dissatisfied
customers. The second bar represents the
group of latently dissatisfied customers, i.e.
customers reporting overall satisfaction but
covered by the proposed model for
dissatisfaction. Finally, the last bar
represents the group of satisfied customers,
who are not covered by our model, i.e. they
are considered as prototypical examples of
satisfied customers.
Two important observations can be made
with regard to this figure. First, the
percentage of complaints for the groups of
dissatisfied and latently dissatisfied
customers is significantly higher than the
percentage of complaints that was observed
for the group of satisfied customers. This
observation can be considered as a proof for
the effectiveness and validity of our model.
Second, through not directly related to
validity, it can be seen that the number of
complaints is different with regard to the
type of complaint that was formulated per
group. In general, all customers seem to have
the largest number of complaints with the
opening hours of the bank, while this differs
for the other complaints per group. For
instance, the satisfied customers have the
fewest complaints with the ``reception’’ while
the latently dissatisfied customers have the
fewest complaints with the ``flexibility of the
staff’’ etc. In the light of dissatisfaction
management, these results can be used to set
priorities for corrective actions.
Test 2: analysis of defection rate.
Unfortunately, because the survey was
carried out anonymously, we have no exact
data concerning the defection rate of the
customers in our study. Instead, a proxy
variable in the satisfaction survey (see
Appendix 2) can be used which assessed
whether the customer has the intention to
have more of his activities concentrated with
other banks instead of with the current bank.
Figure 3
Identification of latently dissatisfied customers based on empirical data
[ 32 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
Figure 5 illustrates that, for the different
groups of customers, different proportions of
customers have the intention to concentrate
their activities more with other banking
institutions in the near future. For instance, it
can be seen that the relative number of
customers in the group of dissatisfied and
latently dissatisfied customers that have the
intention to go elsewhere (first and second bar
with answer ``yes’’), is much higher than in the
group of satisfied customers (third bar with
answer ``yes’’). Also, the relative number of
customers in the group of satisfied customers
that does not have the intention to go
elsewhere (third bar with answer ``no’’) is
much higher than in the other two groups
(first and second bar with answer ``no’’). To
conclude, both observations indicate that the
probability that a customer will go elsewhere
is significantly higher in the group of
dissatisfied and latently dissatisfied customers
than in the group of satisfied customers.
Finally, it is remarkable that within the
group of customers that is undecided, the
bigger proportion consists of latently
dissatisfied customers. In other words,
Figure 5
Does the customer have the intention to go elsewhere?
Figure 4
Complaints behaviour for different customer groups
[ 33 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
customers that tend to be somewhat
ambiguous, i.e. expressing overall
satisfaction but possessing characteristics of
dissatisfied customers, also tend to be
undecided with regard to their future
intentions with the bank.
5. Measures for dissatisfaction
management
5.1. Profiling (latently) dissatisfied
customers
Socio-demographic characteristics constitute
a rich source of information to describe
customer segments in more detail so that
they can be more easily targeted for the
purposes of dissatisfaction management. We
used a set of six socio-demographic variables
(for a full reference see Appendix 2) to profile
latently dissatisfied customers as well as
overall dissatisfied customers. Note that
latently dissatisfied customers have the same
characteristics as dissatisfied customers
although they report overall satisfaction
instead of overall dissatisfaction. So both
groups are identified by the same
characteristics here. Moreover, from the
perspective of dissatisfaction management
both groups can be targeted in the same
manner.
The central question is: are there any
socio-demographic characteristics or profiles
that tend to be more associated with
satisfaction or dissatisfaction? To answer
this question a two-tailed chi-squared
analysis was carried out.
Table I summarizes the results of the chi-
squared analysis for each of the socio-
demographic variables from low to high
p-values. The labels of the category numbers
are different for every attribute and can be
found in Appendix 2. For instance, the cell
(job,3) shows the contributions to the 2
statistic for the dissatisfied (first entry: 5.52)
and the satisfied (second entry: 0.35) workmen.
Table I shows that the first three socio-
demographic variables (job, age and level of
education) are highly significant above the
0.01 level and that the marital status of the
customer is just slightly insignificant at the
0.05 level (see last column in Table I). From
observing the contributions to the 2statistic
in Table I, one can isolate the most important
differences between the observed and the
expected frequencies. These contributions are
calculated as (O¡E)2E, i.e. the square of the
observed frequency minus the expected
frequency divided by the expected frequency
for that cell in the contingency table. Since the
difference between Oand Eis squared, this
number is a positive number, but we have
added the sign to the contributions to be able
to interpret them. A positive sign means that
the observed frequency is higher than the
expected frequency. For instance, within the
category of executives (job,4), we observe a lot
more dissatisfied customers (contribution:
‡19 8) than expected, whereas in the category
of pensioners (job,8) we observe significantly
less dissatisfied customers than expected
(contribution: ¡14 6).
A careful analysis of Table I consequently
reveals that the actual number of dissatisfied
customers is higher than expected in the
following socio-demographic categories:
employee or executive, aged between 18 and
39, having a relatively high level of
education. Segments where the number of
dissatisfied customers is lower than expected
Table I
Contribution to
2
statistic of socio-demographic groups for satisfied and dissatisfied customers
C a t e g o r y
A t t r i b u t e 1 2 3 4 5 6 7 8 9
2
d f
J o b +0.82 +0.36 ±5 . 5 2 +1 9 . 8 +1 5 . 8 ± 0 . 7 9 + 1 . 0 6 + 1 4 . 6 + 0 . 5 7 6 3 . 6 8 1 8 0 . 0 0 0
± 0 . 0 5 ± 0 . 0 2 + 0 . 3 5 ± 1 . 2 5 ± 1 . 0 1 + 0 . 0 5 ± 0 . 0 7 + 0 . 9 2 + 0 . 0 4
A g e +1.0 +2 . 3 +1 . 3 3 + 0 . 4 8 ± 6 . 4 6 ±9 . 5 2 3 . 9 8 7 5 0 . 0 0 0
± 0 . 1 7 ± 0 . 1 5 ± 0 . 0 9 ± 0 . 0 3 + 0 . 4 2 + 0 . 6 2
L e v e l o f ±3 . 1 ±1 1 . 7 +3 . 2 2 ±4 . 8 ±0.84 +0.84 +2 2 . 9 5 0 . 5 4 3 6 0 . 0 0 0
education + 0 . 2 0 + 0 . 7 8 ± 0 . 2 1 + 0 . 3 2 + 0 . 0 6 ± 0 . 0 6 ± 1 . 5 2
M a r i t a l s t a t u s ±2 . 7 5 + 0 . 6 3 3 . 6 2 4 1 0 . 0 5 7
+ 0 . 1 8 ± 0 . 0 4
A g e o f y o u n g e s t +2 . 1 7 ± 0 . 8 8 ± 1 . 9 5 5 . 3 7 0 2 0 . 0 6 8
c h i l d ± 0 . 1 6 + 0 . 0 6 + 0 . 1 4
W o r k 0 . 0 0 ± 0 . 0 4 0 . 0 4 9 1 0 . 8 2 5
s i t u a t i o n 0 . 0 0 0 . 0 0
N o t e : I m p o r t a n t c o n t r i b u t i o n s a r e i t a l i c i s e d a n d s i g n s a r e a d d e d f o r r e a s o n s o f i n t e r p r e t a b i l i t y
[ 34 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
can be profiled as not married or living alone,
workman or pensioner, aged above 55 and
having a relatively low level of education. It
becomes clear that relatively young to
middle-aged clients, with higher levels of
education and white-collar jobs appear to be
more critical in their evaluation because the
number of dissatisfied customers in these
groups tends to be higher than can be
expected.
5.2. Drivers for dissatisfaction
management
However, it is not enough to profile
dissatisfied customers in terms of socio-
demographic variables. In order to make
dissatisfaction management work, the most
important drivers of dissatisfaction related to
the service aspects of the bank must be known
so that they can be acted upon. We decide to
focus on one particular customer segment
here: the executives, since they turned out to
be more dissatisfied than other customer
segments (560 instances). However, the same
analysis can be done for other segments.
For each of the 29 most interesting rules of
dissatisfaction, the hit ratio in this customer
segment was calculated. In other words, we
counted the number of times each of the 29
rules was true for the group of executives.
Two patterns of dissatisfaction were striking.
First, dissatisfaction with question 4 (I obtain
sufficient explanation from the staff) together
with question 7 (Staff spontaneously inform
me about new possibilities concerning bank
services). Second, dissatisfaction with
question 7 together with dissatisfaction with
question 9 (I think I get an impersonal
treatment in my bank office). These patterns
implicitly suggest specific topics for
dissatisfaction management. Indeed, these
indicators can be characterized as the
communication and empathy components of
the service delivery process of the bank.
These findings imply that for this bank
studied, additional attention should be given
to these aspects of the services provided. For
example, the staff managing satisfaction
should make sure that the executive
customer segment gets prompt and complete
information concerning new bank services.
Moreover, sufficient explanation should be
given together with a personal treatment.
6. Conclusions and directions for
future research
6.1. Theoretical and managerial
conclusions
From a theoretical point of view, we tackled
the problem of identifying latently
dissatisfied customers, i.e. identifying
customers who report overall satisfaction but
who possess characteristics of dissatisfied
customers. The descriptive data mining
technique of characteristic rules helps to
discover typical characteristics or properties
of the target group. The model was validated
internally by using test data and externally
by using data on the number of complaints
and data concerning a proxy variable for
defection rate. Results indicated that
remarkable differences with regard to the
type of complaints (e.g. opening hours,
advice, reception, . . . ) and defection rate
could be identified when comparing the
numbers of overall dissatisfied, latently
dissatisfied and overall satisfied customers.
From a managerial point of view, this study
shows that the identification of latently
dissatisfied customers can indeed be
considered as an early warning signal,
providing the opportunity to correct a problem
before real damage is done. It has been shown
that after having identified latently
dissatisfied customers, standard profiling and
classification techniques could be used to
target the right customer segments with the
right corrective actions in order to deal with
latent dissatisfaction most effectively.
6.2. Limitations of the study
The results of this study indicate that
characteristic rules provide an efficient and
effective instrument to identify latently
dissatisfied customers. This has important
implications for (bank) marketing theory and
practice. However, acknowledgement of some
limitations of our study should be
considered, which also suggest new
directions for future research.
An important limitation of our study is
related to the response tendency, i.e. the
survey reflects the intentions of customers
instead of their actual behaviour. Indeed,
because the questionnaire was carried out
anonymously, actual behaviour (for example
defection rate) could not be observed.
Instead, customers were asked about their
intentions to defect (proxy variable).
Our study also has its limitations with
regard to the dataset that has been used: the
current results are based on a single
(analysis) data set from a single bank. A
longitudinal instead of an ad hoc research
project and cross-validation will increase the
reliability of the results of this study.
In the present study, no distinction is made
between customer segments. The impact of
attributes may indeed be different from one
segment to the other. For instance, if we use
product usage as a segmentation criterion,
we may expect that the impact of an attribute
[ 35 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
like personal treatment of the customer may
be different from one product usage group
to another.
Finally, it is possible that for some
customers their classification as latently
dissatisfied is undeserved. For example,
some customers may prefer an impersonal
treatment, and thus have characteristics of
overall dissatisfied customers, but still be
overall satisfied. Being able to separate this
group of customers from the latently
dissatisfied customers would however
require additional information about these
customers, such as their product usage.
Indeed, information of this kind could help in
determining whether a particular customer
prefers impersonal treatment or not, for
instance by observing that this customer
mostly uses self-banking products.
Note
1 Take for inst ance the hypothetical rule ``IF
overall dissatisfied THEN moustache colour =
black’’ (s= 100 per cent). This means that all
overall dissatisfied customers wear a black
moustache. However, the reverse does not
hold necessarily, i.e. given t hat we observe a
person with a black moustache, this is not
necessarily an overall dissatisfied customer,
since there may be overall satisfied customers
wearing a black moustache too!
References
Agrawal, R. and Srikant, R. (1994), ``Fast algorithms
for mining association rules’’, Advances in
Knowledge Discovery and Data Mining, AAAI
Press, Menlo Park, CA, pp. 307-28.
Bitner, M.J., Booms, B.H. and Tet reault, M.S.
(1990), ``The service encounter: diagnosing
favourable and unfavourable incidents’’,
Journal of Marketing, Vol. 54, pp. 71- 84.
Cadotte, E.R. and Tur geon, N. (1988),
``Dissatisfiers and satisfiers: suggestions for
consumer complaints and compliment s’’,
Journal of Consumer Satisfaction,
Dissatisfaction and Complaining Behavior,
Vol. 1, pp. 74-9.
Chakravarty, S., Feinberg, R. and Widdows, R.
(1997), ``Reasons of the ir discontent’’, Bank
Marketing, Vol. 29 No. 1 1, pp. 49-52.
Chakravarty, S., Widdows, R. and Feinberg, R.
(1996), ``How `moments of truth’ define bank-
customer relationships’’, Journal of Retail
Banking Services, Vol. 18 No. 1, pp. 29-34.
Day, R.L. (1984), ``Modelling choices among
alternative responses to dissatisfaction’’,
Advances in Consumer Research , Vol. 11,
pp. 496-9.
Federal Express (1992), ``Federal Express
customer satisfaction and service quality
measurements’’, company presentation.
Fornell, C. and Wernerfelt, B. (19 87), ``Defensive
marketing strategy by complaint
management: a theoretical analysis’’, Journal
of Marketing Research, Vol. 2 4, pp. 337-46.
Hausknecht, D. (1988), ``Emotional me asures of
satisfaction/dissatisfaction’’, Journal of
Consumer Satisfaction, Dissatisfaction and
Complaining Behavior, Vol. 1, pp. 25-33.
Heskett, J.L., Sasser, W.E. and Sch lesinger, L.A.
(1997), The Service Profit Chain: How Leading
Companies Link Profit and Growth to Loyalty,
Satisfaction and Value, The Free Press,
New York, NY.
Heskett, J.L., Jones, T., Loverner, W., Sasser,
W.E. and Schlesinger, L.A. (1994), ``Putting
the service-profit chain to work’’, Harvard
Business Review, March-A pril, pp. 164-71.
Herzberg, F., Mausner, B. and Snyderman, B.B.
(1959), Motivation of Work, John Wiley and
Sons, New York, NY.
Johnston, R. (1995), ``Determinants of service
quality: satisfiers and dissatisfiers’’,
International Journal of Service Industry
Management, Vol. 6 No. 5, pp. 53-71.
Jones, T. and Sasser, W. (1995), ``Why satisfied
customers defect’’, Harvard Business Review,
November-December, pp. 88-99.
Levesque, T. and McDougall, G.H.C. (1996),
``Determinants of customer satisfaction in
retail banking’’, International Journal of
Marketing, Vol. 1 4 No. 7, pp. 12-20.
Maddox, R. (198 1) ``Two factor theory and
consumer satisfaction: replication and
extension’’, Journal of Consumer Research,
Vol. 8, pp. 97-102.
Meidan, A. (1996), Marketing Financial Services,
MacMillan Press L td., Houndmills.
Reichheld, F. (1993), ``Loyalty-based
management’’, Harvard Business Review,
March-April, pp. 64-73.
Reichheld, F. (1996), The Loyalty E ffect. The
Hidden Force Behind Growth, Profits and
Lasting Value, Bain Company, Inc., Harvard
Business School Press, Boston, M A.
Reichheld, F. and Sasser, E. Jr (19 90), ``Zero
defections: quality comes to services’’,
Harvard Business Review, September-
October, pp. 106-11.
Swan, J.E. and Combs, L.J. (1976), ``Product
performance and consumer satisfaction: a
new concept’’, Journal of Marketing, Vol. 40,
pp. 25-33.
TARP (198 6), Consumer Complaint Handli ng in
America: An Update Study, United States
Office of Consumer A ffairs, Washington DC.
Vanhoof, K. and Swinnen, G. (1996) ``Attribute
importance. assessing non- linear patterns of
factors contributing to customer
satisfaction’’, in Research Methodologies for
the New Marketing, ESOMAR Pu blication
Series Vol. 204, November, pp. 16 0-71.
Van Trijp, H., Hoyer, W. and Inman, J. (1996),
``Why switch? Product-category level
explanations for true variety-seeking
behavior’’, Journal of Marketing Research,
pp. 281-92.
[ 36 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
Waterhouse, K. and Morgan, A. (1994), ``Using
research to help to keep good customers:
understanding the process of customer
defection and developing a strategy for
customer reten tion’’, Marketing & Research
Today, Vol. 22 No. 3 , pp. 181-94.
Winstanley, M. (1997) ``What drives customer
satisfaction in commercial banking’’,
Commercial Lending Review, Vol. 12 No. 3,
pp. 36-42.
Worcester, R.M. (1997 ), ``Managing the image of
your bank: the glue t hat binds’’, International
Journal of Bank Marketing, Vol. 1 5 No. 5,
pp. 146-52.
Yavas, U. and Shemwell, D.J. (1996), ``Banking
image: exposition and illu stration of
correspondence analysis’’, International
Journal of Bank Marketing, Vol. 1 4 No. 1,
pp. 15-21.
Appendix 1. Questions on specific service
aspects of the bank
(Translated into English.)
1 In my bank office, leaflets are available
with all necessary information.
2 When I have got important financial
matters to discuss, this is done in a
separate room.
3 Staff treat me in a friendly manner.
4 I obtain sufficient explanation from the
staff.
5 When I enter the bank, the staff take
enough time for me.
6 In my bank I have to queue for a long
time.
7 Staff spontaneously inform me about new
possibilities concerning bank services.
8 The investment advice is solid.
9 I think I get impersonal treatment at my
bank.
10 Opening hours of my bank are
convenient.
11 Explanations of the staff are clear.
12 My bank looks untidy.
13 Staff is discreet enough.
14 Staff is sufficiently flexible to answer my
questions.
15 I get enough information from my bank by
means of correspondence.
16 I can reach my bank easily by phone.
Appendix 2. Questions on socio-demo-
graphic aspects, complaints and defection
A. Socio-demographic aspects
(Numbers between brackets refer to columns
in Table I.)
1 Age: (1)0 ± 17, (2)18 ± 24, (3)25 ± 39, (4)40 ± 54,
(5)55 ± 64, (6)65 +
2 Job: (1)employee, (2)independent,
(3)workman, (4)executive, (5)professional,
(6)working at home, (7)unemployed,
(8)pensioner, (9)student, other
3 Level of education: (1)primary school,
(2)lower secondary, (3)higher secondary,
(4)lower technical, (5)higher technical,
(6)further education (not university),
(7)university
4 Marital status: (1)not married/living alone,
(2)married/living together
5 Age of youngest child: (1)<12, (2)between 12
and 18, (3)>18
6 Work situation: (1)full-time, (2)part-time
B. Complaints
How many and what type of complaints did
you place with the bank during the past year
(advice, discretion, waiting times, opening
hours, expertise, reception, flexibility,
others)?
C. Defection
Do you have the intention of concentrating
more of your banking activities with other
banking institutions in the future? (yes/no)
[ 37 ]
JoseÂe Bloemer, Tom Brijs,
Gilbert Swinnen and
Koen Vanhoof
Identifying latently
dissatisfied customers and
measures for dissatisfaction
management
International Journal of Bank
Marketing
20/1 [2002] 27±37
