Access the full text.
Sign up today, get DeepDyve free for 14 days.
Segmentation in banking for the business client market is traditionally based on size measured in terms of income and the number of employees, and on statistical clustering methods (e.g. hierarchical clustering, k-means). The goal of the paper is to demonstrate that self-organizing maps (SOM) effectively extend the pool of possible criteria for seg- mentation of the business client market with more relevant criteria, including behavioral, demographic, personal, operational, situational, and cross-selling products. In order to attain the goal of the paper, the dataset on business clients of several banks in Croatia, which, besides size, incorporates a number of different criteria, is analyzed using the SOM-Ward clustering algorithm of Viscovery SOMine software. The SOM-Ward algorithm extracted three segments that differ with respect to the attributes of foreign trade operations (import/export), annual income, origin of capital, important bank selection criteria, views on the loan selection and the industry. The analyzed segments can be used by banks for deciding on the direction of further marketing activities. Keywords: self-organizing maps, segmentation, banking, neural networks, data mining JEL classification: M31, G21 InTr OduCTIOn Client segmentation is the process of dividing markets into regression methods, neural networks, self-organizing maps, homogenous groups of consumers. The idea is to create and others (e.g. Wedel and Kamakura, 2003; Chan, Kwong customized marketing strategies for selected segments in and Hu, 2012; Hanafizadeh and Mirzazadeh, 2011). Classical order to satisfy clients’ needs better. Banks can offer cus- Mirjana Pejić Bach tomized products to those market segments in order to in- Full professor crease their profitability (Anderson, Cox and Fulcher, 1976; Faculty of Economics & Business, University of Zagreb Laroche, Rosenblatt and Manning, 1986; Garland, 2005). In E-mail: email@example.com applying marketing strategy in financial institutions, the provider of financial services makes a distinction among Sandro Juković various market segments. Services, the marketing mix and IT Solutions Manager the communication mix are tailored to one or more selected Erste&Steiermärkische Bank segments (Denton and Chan, 1991). Market segmentation E-mail: firstname.lastname@example.org can be done according to various criteria when it is ap- Ksenija Dumičić plied to an individual client market, e.g. geographic, demo- Full professor graphic, psychographic and behavioural (Kotler et al., 2001). Faculty of Economics & Business, University of Zagreb However, the criterion most commonly used by financial E-mail: email@example.com institutions in the business client market is size measured in terms of income and the number of employees (Piercy, Nataša Šarlija 1992; Meadows and Dibb, 1998). Full professor Various methods have been used for market segmenta- Faculty of Economics in Osijek, University of Osijek tion, such as different clustering methods, fuzzy methods, E-mail: firstname.lastname@example.org 32 Copyright © 2013 by the School of Economics and Business Sarajevo Business Client Segmentation in Banking Using Self-Organizing Maps LITera Ture re VIeW cluster analysis methods (e.g. k-means, hierarchical cluster analysis) give a remarkable level of precision (Mingoti and Lima, 2006). However, the problem is that they lack the pos- Segmentation in banking is one of the most important busi- sibility to automatically determine the number of resulting ness decisions, since the practice of designing special groups clusters, and the number of clusters depends on prior un- or baskets of products for special groups of clients is at the derstanding of the dataset. Consequently, analyzing com- root of the modern approach to banking. The two main plex databases would not be easy. Therefore, the need for groups of clients in banking are individual clients (Ekinci, automatic characterizing is obvious. The self-organizing Uray and Ulengin, 2014) and business clients (Turnbull and maps (SOM) method is considered to be a successful sup- Gibbs, 1987). Most of the research in segmentation in bank- plemental method for classical clustering methods. Kuo, Ho ing is oriented towards the retail market. Mäenpää (2006) and Hu (2002) compared three clustering methods for seg- provides the framework for developing market segments for mentation in the 3C (computer, communication, consumer consumers based on their perceptions of the Internet bank- electronic) market: the conventional two-stage method, the ing service, with the goal of detecting possible improve- SOM and the two-stage method as a combination of SOM ments to the banking application. Machauer and Morgner and k-means. They showed that, compared to the conven- (2001) present segmentation of individual banking clients tional two-stage method, the two-stage method combin- using a mixture of customer attitudes and perceptions of ing SOM and k-means gives better results on the basis of bank service benefits. On the other hand, our research is ori- theoretical and practical evaluations. Hung and Tsai (2008) ented towards the business client segmentation in banking. developed a hierarchical SOM model for market segmenta- The traditional approach to business segmentation in tion of multimedia demand in Taiwan. They showed that the banking for the business client market is mainly based on hierarchical SOM model provided better interpretation of the criterion of size, measured in terms of income and the the results than the traditional statistical clustering analysis number of employees (Piercy, 1992; Meadows and Dibb, and the growing hierarchical self-organizing map. Schmitt 1998). Additionally, other financial criteria include a com- and Deboeck (1998) conducted a study using the results of pany’s loan exposure and the ownership of a company a consumer survey in Beijing and Shanghai that illustrated (Sponer, 2012). In addition, Edris (1997) used multiple discri- the efficacy of self-organizing maps. Mangiameli, Chen, and minant analysis and revealed that in Kuwait the ownership West (1996) compared the performance of the SOM and the of a corporation (Kuwaiti, non-Kuwaiti and joint business hierarchical clustering method using 252 datasets with vari- corporations) was related to the selection of a bank. The ous levels of imperfection, and showed that the SOM meth- business intelligence approach regards data as a company’s od outperformed the hierarchical clustering method. assets (Watson and Wixom, 2007). Hence, our research is The aim of this paper is to create and explain business based on the proposition that business client data that en- client segmentation in the Croatian banking industry using compass business client behavior, such as decision maker the SOM method. The motivation for this research is three- characteristics, business client characteristics, an operating fold. First, to our knowledge, there are no papers describ- criterion, a supply management criterion and a situational ing business client segmentation in the banking industry criterion should also be used as a basis for business deci- in Croatia. Second, banks in Croatia use traditional segmen- sion-making, in this case segmentation. The rationale for tation of the corporate sector, which can sometimes blur selecting the abovementioned criteria will be presented. the actual situation. Therefore, data mining, such as cluster In all business situations, including commercial and con- analysis, can find segments that have previously been dis- sumer markets, it will ultimately be individuals who make regarded. Since the dataset for this research is complex and decisions, but a company’s policy limits them in decision- has never been analyzed before, we decided to implement making, which stresses the importance of decision maker the SOM method. Third, the criterion for banking segmenta- preferences (Shocker et al., 1991). Business client characteris- tion is traditionally size measured in terms of income and tics refer to the company size (Piercy, 1992; Meadows and the number of employees (Piercy, 1992; Meadows and Dibb, Dibb, 1998), industry (Athanassopoulos, 2000), location 1998), and in this research we have used many other criteria, (Venkatesh, 2011), and the international orientation of the such as decision makers’ characteristics, an operating crite- company (Agarwal, Malhotra and Bolton, 2010). The operat- rion, and a supply management criterion. In this research ing criterion facilitates segmentation based on the specific we had a complex database consisting of many variables, behaviour of the client, which is reflected in specific trans- which is the reason why the SOM method is chosen for the actions conducted by the client (e.g. credit card usage) or analysis. planned transactions (e.g. planned credit card usage), or The paper’s composition is the following. First, an over- attitude towards the most important operating service view of previous research is given. The second part of the (Patsiotis, Hughes and Webber, 2012). Supply chain manage- paper describes the methodology of the paper, encompass- ment determines the procurement organization of the cli- ing research methods and the data used in the research. The ent and its rules when deciding whether to take a familiar, a third part of the paper presents the cluster analysis resulting national or the cheapest vendor, etc. The supply chain man- in the business customer segmentation. The discussion and agement criterion describes a client’s affinity towards price the concluding part of the paper present the profile of each and service quality (Chen and Bell, 2012). Finally, situational cluster and summarize the results of the analysis. criterion elements do not have a permanent character and push companies to understand a client more deeply (Kim South East European Journal of Economics and Business 33 Business Client Segmentation in Banking Using Self-Organizing Maps and Grunig, 2011). In our research, they include the major while grouping similar data together enables data cluster- bank where most of the services are used, and the level of ing. Thus, it is possible to apply this method in the field of satisfaction with the current major bank. market segmentation. In fact, any connection between neu- The cluster analysis is the basic method of knowledge rons of the input-output layer has its corresponding weight. discovery in databases that can be used for market segmen- In practice, neural networks, among other things, are used tation (Mäenpää, 2006). Traditional clustering methods are in the financial services industry (Kumar and Ravi, 2007). hierarchical methods and k-means. The aim of these meth- In humans, various sensory stimuli are neurologically ods is the detection of global data structures. In this method mapped in the brain so that the spatial relationship be- the target attribute is not predefined, so there is no differ - tween stimuli of certain parts of the brain corresponds to ence among the attributes’ importance. Clustering methods the spatial relationship between neurons organized in a fall into a group of „unattended“ methods. Self-organizing two-dimensional map (Kohonen, 1995). Consequently, the maps (SOM) are the two-layered artificial neural networks location of points on the map reflects the relative similar - which were initially designed to conduct the tasks of clus- ity of points in a multidimensional space, i.e. the more two tering, data visualization and abstraction (Han and Kamber, datasets are similar to each other, the closer they will be on 2006, pp. 434). SOM has been used for market segmentation the resulting two-dimensional map. Accordingly, the SOM’s in tourism (Kun et al., 2002) as well as in various other fields primary function is to display multidimensional data from of market segmentation, such as telecommunication (Kiang, the inputs of the network in the two-dimensional map, et al., 2006), on-line game markets (Lee et al., 2004), on-line while retaining the relationships among data (Kiang, 2001). shopping (Vellido et al., 1999), tourism (Bloom, 2004), com- The Ward Clustering Method is a part of the Viscovery puter, communication and consumer electronic markets SOMine data mining software tool, which encompasses the (Kuo, Ho and Hu, 2002), and multimedia products (Hung SOM-Ward algorithm that is used for the data analysis in this and Tsai, 2008). Our research explores the possible usage of paper. The Classical Ward method belongs to hierarchical SOM methods for business client segmentation in banking. agglomeration algorithms which have the following charac- To summarize, further research is needed in the field teristics: they start with the clustering process where every of business client segmentation in banking. In order to ex- cluster represents one cluster, and in each of the cluster- pand knowledge on business client segmentation in bank- ing steps the two clusters that have a minimal distance are ing, new research is needed that would, in addition to size joined. The minimal distance is called the Distance Niveau. measured in terms of turnover and the number of employ- The SOM-Ward algorithm is a so-called hybrid algorithm ees, use a more extensive set of criteria, including decision which was developed on the grounds of the soft comput- maker characteristics, business client characteristics, an op- ing paradigm, which is primarily based on the usage of vari- erating criterion, a supply management criterion and a situ- ous intelligent methods and algorithms, as opposed to the ational criterion. In addition, by using the SOM method in hard computing paradigm. In SOMine, the distance matrix is our research, we test the applicability of this novel approach initialized with respect to the number of data records which to the field of business client segmentation in banking. are represented by nodes on the output map. The nodes with many corresponding data records have a higher im- pact in comparison with the nodes with fewer connections. MeThOdOLOG y Sample description Self-organizing maps (SOM) Self-organizing maps (SOM) are based on the concept of In order to conduct segmentation using the SOM-Ward neural networks. The neural network is a model of a biologi- algorithm, a questionnaire survey approach was utilized cal neural network of the human nervous system, a set of to collect data from Croatian firms. Questionnaires were parallel interconnected neurons that are organized in layers distributed among Croatian micro (<10 employees), small (Jagric and Jagric, 2011). The main feature of artificial neural (11-49 employees), medium (50-249 employees) and large networks as well as SOM is the property of learning from (250+ employees) firms. Business clients that participated the data that are entered in the network input layer. Thus a in the research were selected randomly from the Croatian complex interaction of the analyzed data can be modelled Company Directory, available at the website of the Croatian without having to know the nature of the phenomena the Chamber of Commerce. The stratified sample approach was data describes. In other words, it is not necessary to create used in order to capture more large and medium sized firms, an algorithm that will analyze the data, but through train- which are important for banks because, in comparison to ing the neural network learns how to analyze data on its small and micro firms, they conduct larger financial transac - own. Besides the learning capacity of the SOM method, two tions. Data for this research were collected through a ques- other important features of this method have to be men- tionnaire answered by 850 Croatian firms. tioned (Bigus, 1996). The first is the dimensionality reduc - Table 1 visibly specifies that, according to the sample tion of the data and the second is the placement of simi- size, the structure of firms that participated in the survey dif- lar data to neighbouring nodes on the resulting map. The fers to a large extent from the structure of firms in Croatia as dimensionality reduction facilitates easier understanding a whole. Our sample contained 38.83% of micro firms, while of the attribute interrelationship between the input data, in Croatia micro firms comprise 92% of the total population. 34 South East European Journal of Economics and Business Business Client Segmentation in Banking Using Self-Organizing Maps Table 1: Comparison of firms that participated in the survey and population characteristics Sample Croatia Chi-square Firm size (p-value) Number of firms Structure in % Number of firms Structure in % Micro firms (<10 employees) 304 35.8% 152251 92.0% 63.656 (0.000**) Small firms (10- 49 employees) 263 30.9% 10757 6.5% Medium firms (50-249 employees) 182 21.4% 1986 1.2% Large firms (250+ employees) 101 11.9% 496 0.3% Total 850 100.0% 165490 100.0% ** statistically significant at 1% Source: Authors’ calculation Table 2: Segmentation criteria used in the research Segmentation Variables describing the criterion (type of variable) criterion Decision maker (1) the importance of selecting a bank providing services for clients (Likert 1-5) preferences (2) client rating of the best bank in Croatia (Likert 1-5) Business client (1) total number of employees (numeric) characteristics (2) increase / decrease in revenues compared to the previous year (numeric) (3) total income in the previous year (numeric) (4) plans related to the number of employees (nominal) (5) main business activity (nominal) (6) legal form of the company (nominal) (7) percentage of import and export in annual sales (numeric) (8) ownership form of the company and the capital origin (nominal) (9) the financial position of the company (nominal) (10) method of wage payment (nominal) (11) headquarters of the company and the year of establishment (nominal) Operating criterion (1) the most important reasons which affect the selection of a bank for a loan request (Likert 1-5) (2) credit cards used by a client (binomial 0-1) (3) credit cards which a company is planning to use (binomial 0-1) Supply management (1) the three most important characteristics of a bank (e.g. security and stability, commissions approach criterion and fees, keeping up with promises) (Likert 1-5) (2) the three most important characteristics of bank services (e.g. a wide spectrum of services, quality and price ratio and promptness in solving requests) (Likert 1-5) Situational criterion (1) the major bank where most of the services are used (binomial 0-1) (2) level of satisfaction with the current major bank (Likert 1-5) Source: Authors On the other hand, our sample contained 20.58% medi- the other hand, medium and large firms, although small in um-sized firms, and the total Croatian population of firms number, generate a larger number and amount of financial contains only 1.20% medium-sized firms. Larger firms are transactions. also overrepresented in our sample. The chi-square test proved that these differences are statistically significant at 1% (χ =63.656, p-value=0.000). Hence, the overpresence of research instrument large, medium and small firms, and underpresence of mi- cro firms in our sample should be taken into account when The literature review presented in the paper contains the considering the implications of the results of the research. rationale for selecting the segmentation criteria used for However, such an approach was chosen taking into ac- this research: decision maker characteristics, business client count that micro firms, although great in number, generate characteristics, an operating criterion, a supply manage- a smaller number and amount of financial transactions. On ment criterion and a situational criterion. The questionnaire South East European Journal of Economics and Business 35 Business Client Segmentation in Banking Using Self-Organizing Maps Figure 1: Identified clusters of business clients in banking for this research is developed based on the criteria de- scribed in the literature review section (Table 2). There were three types of questions: (1) Likert scales from 1 to 5, (2) questions with predefined answers (nominal vari- ables), and (3) numeric variables. Statistical analysis The SOM-Ward algorithm implemented in Viscovery SOMine software was used in order to segment busi- ness customers. The SOM-Ward algorithm extracted three clusters, presented on a resulting map with only the main criteria used for a short description. In analyzing and interpreting clusters, we first de - scribed each cluster according to all segmentation criteria. Second, we compared clusters descriptively. Finally, we applied the F-test for testing differences in mean values between clusters and chi-square for testing the association between clusters and different criteria. Source: Authors’ calculation strategies for each cluster. In Table 3 clusters are compared reSuLTS Of B u SIneSS Cu STOMer according to decision maker preferences. The percentages SeGMenT aTIOn in the table present percentages of the companies which Figure 1 visually represents the identified clusters of busi- belong to a specific cluster and which have chosen certain ness clients in banking, as well as presents the structure of criteria. For example 53.1% of companies in Cluster 1 think the total sample according to clusters. Cluster 1 contains that selecting a bank is of very high importance. It can be 50%, Cluster 2 contains 36% and Cluster 3 only 14% of the noticed that in all of the three clusters the importance of total sample. selection on the basis of services provided by a bank is very In analyzing and interpreting clusters it is very useful high, although in the third cluster that percentage is the to compare clusters according to the segmentation cri- highest and in the second the lowest. As for the ranking teria. This analysis serves as a basis for creating marketing of the best banks, bank A is most common in the second Table 3: Clusters according to the importance of selecting a bank providing services to clients, and to the clients’ rating of the best bank in Croatia (% of firms in the cluster) Chi-square Cluster 1 Cluster 2 Cluster 3 Total (p-value) The importance of selecting a bank providing services for clients Very high importance 51.3% 45.1% 53.4% 49.4% 23.610 (0.009**) High importance 31.7% 34.0% 28.8% 32.1% Medium importance 7% 6% 7% 6.6% Low importance 0% 4% 2% 1.5% Not important 0% 2% 0% .7% No answer 9.9% 9.5% 9.3% 9.7% Clients’ rating of the best bank in Croatia Bank A 23.2% 32.7% 25.4% 26.9% 116.301 (0.000**) Bank B 16.8% 19.6% 24.6% 18.9% Bank C 19.9% 12.7% 10.2% 15.9% Bank D 5.7% 4.9% 6.8% 5.5% Other banks 15.30% 11.40% 15.20% 14.00% There is no best bank 5.40% 3.30% 5.10% 4.60% No answer 13.70% 15.40% 12.70% 14.20% ** statistically significant at 1%; * statistically significant at 5% Source: Authors’ calculation 36 South East European Journal of Economics and Business Business Client Segmentation in Banking Using Self-Organizing Maps cluster, banks B and C are also highly represented in all clus- sales and an average % of export in annual sales that are ters. Chi-squares show significant associations between statistically significant at the 1% probability level. clusters and the importance of selecting a bank providing services for customers (χ =23.610, p=0.009) as well as asso- Table 5 presents the business customer characteristics of ciations with customer ratings of the best bank (χ =116.301, the clusters. The percentages in the table present percent- p<0.000). This means that companies in clusters differ ac - ages of the companies which belong to a specific cluster. cording to decision makers’ characteristics. It can be noticed that clusters differ significantly according In Table 4 clusters are compared according to the aver- to the industry, total revenue, change of revenue compared age number of employees, and the average % of import and to previous year, origin of capital and ownership, but not export in annual sales. The F-test revealed that there is no according to plans for employment (χ =11.591, p=0.072). statistically significant difference in clusters’ average num- Furthermore, in each cluster most business customers are bers of employees (F=1.711, p-value=0.181). As for import from trade. Domestic capital is well represented in all clus- and export, Cluster 1 includes companies with the highest ters, and the most in the third cluster, while most foreign average import and export, and Cluster 2 with the lowest. In capital can be found in the first cluster. SMEs are the most addition, the clusters have an average % of import in annual represented in Cluster 2. Table 4: Clusters according to the average numbers of employees, and average % of import and export in annual sales F-testC Cluster 1 Cluster 2 Cluster 3 Total (p-value) 1.711 Average number of employees 281.1 135.2 145.2 209.5 (0.181) 36.107 Average % of import in annual sales 29.6% 10.9% 22.1% 21.8 (0.000**) 15.255 Average % of export in annual sales 16.1% 6.9% 9.9% 11.9 (0.000**) ** statistically significant at 1% Source: Authors’ calculation Table 5: Clusters according to the structure of firms based on business client characteristics (% of firms in the cluster) Chi-square Cluster 1 Cluster 2 Cluster 3 Total (p-value) Industry Trade 39.0% 33.7% 35.6% 36.6% 79.641 (0.000**) Production or mining 24.8% 12.7% 11.0% 18.5% Construction 5.4% 8.2% 12.7% 7.4% Tourism: hotels & restaurants 1.9% 6.5% 2.5% 3.7% Financial and other services 2.1% 6.5% 1.7% 3.7% Transport and communications 2.6% 2.6% 3.4% 2.7% Community services 1.9% 1.3% 3.4% 1.9% Agriculture or fishing 0.9% 2.3% 3.4% 1.8% Other 21.4% 26.2% 26.3% 23.7% Total revenue in previous year Up to 0.5 million EUR 11.1% 54.9% 22.0% 28.5% 215.449 (0.000**) From 0.5 to 1 million EUR 6.6% 7.5% 11.0% 7.6% From 1 to 1.5 million EUR 4.7% 3.6% 3.4% 4.1% From 1.5 to 2 million EUR 7.1% 3.9% 6.8% 5.9% From 2 to 5 million EUR 19.9% 8.5% 17.8% 15.5% From 5 to 8 million EUR 10.6% 3.9% 15.3% 8.9% From 8 to 10 million EUR 5.4% 2.0% 3.4% 3.9% From 10 to 50 million EUR 14.4% 2.3% 11.0% 9.6% 50 and more million EUR 5.9% 0.7% 3.2% Don’t know 5.2% 2.9% 2.5% 4.0% Refuse to answer 9.0% 9.8% 6.8% 9.0% South East European Journal of Economics and Business 37 Business Client Segmentation in Banking Using Self-Organizing Maps Table 5: con’t Change of revenue compared to previous year Increase of revenue compared to 69.7% 54.2% 72.0% 64.5% 24.263 previous year (0.000**) Revenue on the same level 18.7% 31.0% 19.5% 23.3% Decrease 10.4% 13.1% 8.5% 11.1% No answer 1.2% 1.6% 1.2% Plans for employment in the next year Retain all the employees 41.6% 48.7% 44.9% 44.6% 11.591 (0.072) Employ new workers 48.9% 39.5% 39.8% 44.3% No answer 5.9% 8.2% 7.6% 7.0% Lay off employees 3.5% 3.6% 7.6% 4.1% Origin of the capital Domestic 80.9% 94.1% 97.5% 88.0% 57.361 (0.000**) Foreign 10.2% .3% 1.7% 5.4% Mixed (domestic & foreign) 9.0% 4.6% .8% 6.3% No answer 1.0% .4% Belonging to the group Independent company 61.2% 55.6% 70.3% 60.4% 103.383 (0.000**) SME 9.9% 35.3% 13.6% 19.6% Owned by another company 14.4% 6.5% 9.3% 10.9% Company owns other companies 13.7% 2.3% 5.9% 8.5% No answer 0.8% 0.3% 0.9% 0.6% ** statistically significant at 1%; * statistically significant at 5% Source: Authors’ calculation In Table 6 clusters are compared according to operating cluster. Chi-square results show a significant association be - criteria. The interest rate is the most important criterion in tween clusters and criteria for a loan request (χ =231.032, the first and the second cluster while the speed and simplic - p<0.000), which means that companies in clusters differ ac - ity of procedures are the most important criteria in the third cording to important criteria for a loan request. Table 6: Clusters according to operating criteria (% of firms in the cluster) Chi-square Cluster 1 Cluster 2 Cluster 3 Total (p-value) Important criteria which affect the selection of a bank for a loan request Interest rate 65.5% 51.3% 3.4% 51.7% 231.032 (0.000**) Speed of loan approval 13.0% 12.1% 28.0% 14.8% Knowing people at the bank 0.2% 0.1% Repayment 3.1% 4.9% 3.4% 3.8% Fees 0.5% 1.7% 0.5% Simplicity of procedures 9.9% 10.1% 24.6% 12.0% Size of the bank 1.7% 0.7% 10.2% 2.5% Documentation required to obtain a loan 2.4% 4.9% 7.6% 4.0% Staff 1.3% 0.8% 0.6% Bank not requiring unrealistic guarantees 1.2% 11.8% 6.8% 5.8% Something else 3.4% 0.5% Don’t know/No answer 2.6% 2.9% 10.2% 3.8% ** statistically significant at 1%; * statistically significant at 5% Source: Authors’ calculation 38 South East European Journal of Economics and Business Business Client Segmentation in Banking Using Self-Organizing Maps Table 7: Clusters according to the supply management approach criteria (% of firms in the cluster) Chi-square Cluster 1 Cluster 2 Cluster 3 Total (p-value) The most important characteristic of a bank Security and stability 72.3% 81.7% 60.2% 74.0% 102.624 (0.000**) Low interest rates 4.7% 2.0% 0.8% 3.2% Low commissions and fees 6.9% 2.3% 4.2% 4.8% Approves the loan 0.7% 0.3% 1.7% 0.7% Familiar person 0.7% 0.7% 2.5% 0.9% Support for business abroad 6.6% 2.9% 0.8% 4.5% Specialization for the company’s industry 3.8% 2.6% 16.9% 5.2% Familiar bank 0.9% 2.9% 8.5% 2.7% Other 3.3% 4.6% 4.2% 3.9% Major bank (where most of the services are used) Bank A 24.8% 33.3% 28.8% 28.5% 78.146 (0.058) Bank B 11.8% 14.7% 13.6% 13.1% Bank C 9.9% 6.2% 5.1% 7.9% Bank D 9.5% 9.2% 12.7% 9.8% Other banks 44.0% 36.6% 39.8% 40.7% No answer 22.2% 18.7% 20.1% 19.3% Level of satisfaction with the current major bank Highly satisfied 30.3% 26.8% 28.0% 28.7% 12.877 (0.231) Mostly satisfied 44.4% 47.4% 47.5% 45.9% Undecided 18.4% 20.9% 16.1% 19.0% Mostly not satisfied 2.6% 3.6% 3.4% 3.1% Not satisfied at all 1.7% 3.4% 1.3% No answer 2.6% 1.3% 1.7% 2.0% ** statistically significant at 1%; * statistically significant at 5% Source: Authors’ calculation In Table 7 clusters are compared according to the supply 16.1%) and they are also the largest in terms of the annual management approach criteria. Chi-squares show signifi- income in the last year (10-50 million EUR). Therefore, this cant association between clusters and important character- cluster was labelled as Largest-growing. This cluster has istic of a bank (χ =102.626, p<0.000) but not in the case of the highest proportion of state-owned companies, but it the most important characteristics of the bank (χ =78.146, also consists of foreign companies (with higher turnovers). p=0.058) nor in the case of frequency of usage of services Companies in this cluster plan new employment more often (χ =12.877, p=0.231). It can be concluded that the level than companies in other clusters, which could indicate that of satisfaction and frequency of usage are the same in all they have a clear growth strategy. When selecting a bank for clusters. a loan, important factors include interest rates, quickness of loan approval and a simplified method of funds with- drawal. In addition, they also demand greater security and affordable interest rates. In banking services, they want high dISCu SSIOn transaction accuracy, and promptness in solving problems Each cluster can be described by combining all the variables or requests. These companies want to ensure undisturbed included in the analysis. In order to clearly indicate the dif- functioning of business processes, especially with interna- ference among clusters, appropriate labels were given to tional partners. Therefore, banks should offer specialized the clusters. services. Cluster 1 – Largest-growing Cluster 2 – Smallest-stagnating Companies in Cluster 1 have the largest average export and Companies in Cluster 2 have the lowest turnovers that are import ratio in the annual turnover (import 29.6%, export stagnating. They have a minimum average trade ratio with South East European Journal of Economics and Business 39 Business Client Segmentation in Banking Using Self-Organizing Maps foreign countries (import 11%, export 6.9%) and the mini- easier understanding of the attribute interrelationship be- mal annual revenue in the last year (under 0.5 million EUR). tween the input data as a basis for decision-making can be Therefore, this cluster was labelled as Smallest-stagnating. A improved by using dimensionality reduction and visualiza- great proportion of these companies are privately owned tion of the multi-dimensional data on a two-dimensional and established by domestic capital. They expect employ- map. The combination of self-organizing maps with the ment stagnation, but do not plan any layoffs. When select - classical cluster analysis in the Ward technique of clustering ing a bank for a loan, the most important factors for them done using Viscovery SOMLine proved to be a useful tool are interest rates, quickness of loan approval and simple for cluster analysis. To our knowledge, there is no research loan approval procedures. They demand that banks main- paper that investigates segmentation of the corporate sec- tain a high degree of security, affordable interest rates, and tor in the banking industry in a transition country. Thus, by warranties on investments. Also, they expect banking ser- examining Croatia this paper provides required finidings vices to provide a wide spectrum of products and services, and cognitions. In addition, banks often use the tradition- transaction accuracy and quickness in solving problems or al segmentation of the corporate sector, and with this re- requests. The data in this cluster have shown that most of search we showed the advantages of using the SOM-Ward those companies are privately owned and are mainly estab- method. The combination of different segmentation crite - lished by domestic investors. Thus, this cluster has the lowest ria with the SOM-Ward method resulted in extracting three number of foreign investors. The minimal annual turnover in clusters which give a detailed explanation of the corporate this cluster indicates that these companies are mostly small- sector in Croatia. sized enterprises or crafts. This is also supported by the fact Our research showed that important segmentation char- that they have an underdeveloped employment strategy acteristics are based on the characteristics of the corporate and their plans on the increase of the number of employ- sector itself (industry sector, import, export, total revenue, ees are stagnating. It is important to notice that they want a origin of the capital, ownership of the companies) and the wide spectrum of banking services, which makes them dis- given bank’s characteristics (the importance of selecting a tinctive from the other two clusters, while the main concern bank, rating of a bank, criteria for selection, the most im- for large companies (Cluster 1) is transaction accuracy. portant characteristics of a bank). Based on the description of each cluster, banks could create a business strategy cus- tomized to each cluster. This means that different strategies Cluster 3 – Medium-growing should be tailored not just according to what customers Companies in Cluster 3 are mostly medium sized in terms of want, such as criteria for selection and the important char- annual turnover. Their turnovers show the highest growth acteristics of a bank, but also according to their characteris- rate of all three clusters. Therefore, this cluster was labelled tics, such as revenue and the industry sector they belong to. as Medium-growing. They have the largest import to annual It is important to emphasize that the following segmenta- turnover ratio (22.1%), but they also have a small export to tion criteria are shown to not be significant: the number of annual turnover ratio (9.9%). They have a yearly income be- employees in companies, plans for employment, the level tween 0.5 and 10 million EUR, are mostly privately owned of satisfaction with the bank and the current bank selection. and are established with domestic capital. They do not plan Market segmentation by SOM could also provide market- to lay off many employees but plan to maintain the current ing experts with the ability of making a variety of different number of employees, and a certain number of companies activities tailored to each segment, which enables banks to also plan new employments. When selecting a bank for increase their profitability. loans, the most important factors are quick loan approval, However, when using the results of our research several simple procedures or warranty instruments and reasonable limitations should be taken into account, which consequent- warranties in exchange for the loan. An ideal bank should ly opens possibilities for future research. First, we selected be secure and stable, have a dedicated person for prob- our criteria based on a broad selection of criteria. However, lem solving and the ability to quickly answer requests, and we did not include other criteria that could be also impor- should guarantee the stability of savings. Desirable bank- tant for the decision on the selection of a bank by business ing services include transaction accuracy, quick request re- clients, such as the amount of banking provisions and the sponse and problem solving (similar to Cluster 1). The main ease of use of the Internet banking sector (Assunção, 2013). segment characteristics are domestic capital and an effort Since the banking industry is experiencing the constant to maintain the current number of employees. When apply- threat of intruders’ attacks on Internet banking software ing for loans, they prefer the quickness of loan approval and (Nasri and Charfeddine, 2012), the issue of Internet security a simple procedure rather than lower interest rates. They should also be taken into account. Second, in our research also demand transaction accuracy, like the companies in we purposefully included a higher percentage of large firms Cluster 1. in our sample because of their importance for the Croatian economy. Therefore, further research in the field of market segmentation could be conducted for large, small and me- COnCL u SIOn dium firms separately. The aim of the paper was to create and explain business customer segmentation in the Croatian banking indus- try using self-organizing maps. It has been shown that an 40 South East European Journal of Economics and Business Business Client Segmentation in Banking Using Self-Organizing Maps referenCeS 18. Kiang, M.Y., Hu, M.Y., Fisher, D.M. 2006. An extended self-organ- izing map network for market segmentation—a telecommuni- cation example. Decision Support Systems 42(1): 36-47. 1. Agarwal, J., Malhotra, N. K., Bolton, R. N. 2010. A cross-national 19. Kim, J. N., Grunig, J. E. 2011. Problem solving and communica- and cross-cultural approach to global market segmentation: tive action: A situational theory of problem solving. Journal of An application using consumers’ perceived service quality. Communication 61(1): 120-149. Journal of International Marketing 18(3): 18-40. 20. Kohonen, T. 1995. Self-organizing maps. Berlin: Springer-Verlag. 2. Anderson, W.T., Cox, E.P., Fulcher, D.G. 1976. Bank Selection 21. Kotler, P., Armstrong, G., Saunders, J. & Wong, V. 2001. Principles Decisions and Market Segmentation. Journal of Marketing of Marketing, Prentice Hall, Harlow. 40(1): 40-45. 22. Kumar, P. R., Ravi, V. 2007. Bankruptcy prediction in banks 3. Assunção, J. 2013. Eliminating entry barriers for the provision and firms via statistical and intelligent techniques - A review. of banking services: Evidence from ‘banking correspondents’ in European Journal of Operational Research 180: 1-28. Brazil. Journal of Banking & Finance 37(8): 2806-2811. 23. Kuo, R.J., Ho, L.M., Hu, C.M. 2002. Integration of self-organizing 4. Athanassopoulos, A. D. 2000. Customer satisfaction cues to feature map and K-means algorithm for market segmentation. support market segmentation and explain switching behavior. Computers & Operations Research 29(11): 1475-1493. Journal of Business Research 47(3): 191-207. 24. Laroche, M., Rosenblatt, J.A., Manning, T. 1986. Services 5. Bloom, J.Z. 2004. Tourist market segmentation with linear and Used and Factors Considered Important in Selecting a Bank: non-linear techniques. Tourism Management 25(6): 723-733. An Investigation across Diverse Demographic Segments, 6. Chan, K.Y., Kwong, C.K., Hu, B.Q. 2012. Market segmentation International Journal of Bank Marketing 4(1): 35 – 55. and ideal point identification for new product design using 25. Lee, K., Booth, D., Pervaiz, A. 2005. A comparison of supervised fuzzy data compression and fuzzy clustering methods. Applied and unsupervised neural networks in predicting bankruptcy of Soft Computing 12(4): 1371-1378. Korean firms. Expert Systems with Applications, 2: 1-16. 7. Chen, J., Bell, P. C. 2012. Implementing market segmentation 26. Machauer, A., Morgner, S. 2001. Segmentation of bank clients using full-refund and no-refund customer returns policies in a by expected benefits and attitudes, International Journal of dual-channel supply chain structure. International Journal of Bank Marketing, 19(1): 6-18. Production Economics 136(1): 56-66. 27. Mäenpää, K. 2006. Clustering the consumers on the basis of 8. Deboeck, G., Schmitt, B. 1998. Differential Patterns in their perceptions of the Internet banking services. Internet Consumer Purchase Preferences using Self-Organizing Maps: A Research, 16(3): 304–322. Case Study of China, Visual Explorations in Finance: with self- 28. Mangiameli, P., Chen, S. K., West, D. 1996. A comparison of SOM organizing maps, edited by G. Deboeck, T. Kohonen, 141-156, neural network and hierarchical clustering methods. European London: Springer. Journal of Operational Research 93(2): 402-417. 9. Denton, L., Chan, A.K.K. 1991. Bank Selection Criteria of 29. Nasri, W., Charfeddine, L. 2012. Factors affecting the adoption Multiple Bank Users in Hong Kong, International Journal of of Internet banking in Tunisia: An integration theory of accept- Bank Marketing 9(5): 23–34. ance model and theory of planned behavior. The Journal of 10. Edris, T. A. 1997. Services considered important to business High Technology Management Research 23(1):1-14. customers and determinants of bank selection in Kuwait: a seg- 30. Patsiotis, A. G., Hughes, T., Webber, D. J. 2012. Adopters and mentation analysis. International Journal of Bank Marketing non-adopters of internet banking: a segmentation study. 15(4): 126-133. International Journal of Bank Marketing 30(1): 20-42. 11. Ekinci, Y., Uray, N., Ulengin, F. (2014). A customer lifetime val- 31. Shocker, A. D., Ben-Akiva, M., Boccara, B., Nedungadi, P. 1991. ue model for the banking industry: a guide to marketing ac- tions. European Journal of Marketing 48(3/4), Accepted for Consideration set influences on consumer decision-making publication. and choice: Issues, models, and suggestions. Marketing letters 12. Garland, R. 2005. Segmenting retail banking clients. Journal of 2(3): 181-197. Financial Services Marketing 10: 179–191. 32. Sponer, M. 2012. Segmentation of corporate clients in a bank, 13. Han, J., Kamber, M. 2006. Data Mining: Concepts and Masaryk University, Brno, Czeck Republik. Techniques. Tokyo: Morgan Kaufmann Publishers. 33. Turnbull, P. W., Gibbs, M. L. 1987. Marketing bank services 14. Hanafizadeh, P., Mirzazadeh, M. 2011. Visualizing market seg- to corporate customers: the importance of relationships. mentation using self-organizing maps and Fuzzy Delphi meth- International Journal of Bank Marketing 5(1): 19-26. od ADSL market of a telecommunication company. Expert 34. Vellido, A., Lisboa, P.J.G., Meehan, K. 1999. Segmentation of Systems with Applications 38(1): 198-205. the on-line shopping market using neural networks. Expert 15. Hung, C. Tsai, C.F. 2008. Market segmentation based on hier- Systems with Applications, 17(4): 303-314. archical self-organizing map for markets of multimedia on de- 35. Venkatesh, R. 2011. New Parameters in Market Segmentation- mand, Expert Systems with Applications 34(2): 780–787. Ethnic Marketing is the Key. Advances In Management 4(11): 16. Jagric, T., Jagric, V. 2011. A Comparison of Growing Cell 15-19. Structures Neural Networks and Linear Scoring Models in the 36. Watson, H.J., Wixom, B.H. 2007. The Current State of Business Retail Credit Environment. Eastern European Economics 49(6): Intelligence, Computer, 40(9): 96-99. 74-96. 37. Wedel, M., Kamakura, W. 2003. Market Segmentation: 17. Kiang, M. Y. 2001. Extending the Kohonen self-organizing map Conceptual and Methodological Foundations, Kluwer networks for clustering analysis. Computational Statistics & Academic Publishers, Norwell. Data Analysis 38: 161-180. South East European Journal of Economics and Business 41
South East European Journal of Economics and Business – de Gruyter
Published: Nov 1, 2013
Keywords: self-organizing maps; segmentation; banking; neural networks; data mining
Access the full text.
Sign up today, get DeepDyve free for 14 days.