Analytical Techniques for Marketing - Research Paper Example

Module Assignment and number Degree Word count 976 Table of Contents Section Page number Introduction 1 2. Theory 1 3. Summary of a Published on Cluster Analysis 2 4. Method 3 5. Results 4 6. Marketing Implications of Results 8 7. Summary /Conclusions 8 List of References 10 Appendices Appendix 1 SSS Output 12 List of Tables Table 1. Average Factor Scores for Final Cluster Centres 5 Table 2. Cluster Identity and Behavioural Characteristics 6 Table 3. Summary of Cluster Profiles 7 1. Introduction The aim of this paper is to establish the factors that measure students’ attitudes to the importance of supermarket features. This has been achieved through cluster analysis of data. Cluster analysis can be defined as a technique for classified large amounts of information into manageable and meaningful units. It is a statistical method for partitioning a sample into homogenous classes to produce an operational classification (Nourani et al. 2012). Preliminary analysis using hierarchical clustering will identify respectively the possibility solutions in the range five-clusters to two-clusters. This analysis will employ the K-Means optimisation method and will specify a three-cluster solution. The results will be important in identifying the factors that measure students attitudes to the importance of supermarket features and therefore improve the effectiveness of marketing to student shoppers This paper therefore organised as follows. First, section 2 is the theory and presents a structured, critical account of the theory of cluster analysis. Section 3 summarises published studies on the application of cluster analysis. Section 4 explains the methodology. Section 5 presents and interprets the results of the analysis. Section 6 discusses the marketing implications of the results. Section 7 summarises the study, evaluates the study and suggests further research. 2. Theory The main objective of cluster analysis technique is to classify data into some meaningful groups. This method does not distinguish between dependent and independent variables. It reduces the number of observations by putting them into smaller clusters. Hierarchical cluster analysis is usually used to attain an optimum number of clusters. This is done using Ward’s method by applying squared Euclidean Distance. Then a hierarchical cluster analysis is rerun with selected clusters from the first analysis in order to allocate every case in the sample to a specific cluster. The Euclidean distance is the most commonly used distance measure in SPSS but there are other methods. Ward’s method is the most commonly used for clustering algorithms. 3. Summary of a Published Articles on Cluster Analysis Gough & Sozou (2005) sought to understand peoples motivation and behaviour as regards to the provision for their retirement. Data was collected through questionnaires administered to 540 respondents. Ten measures were used. The ten measures were concerned with demographic and economic variables or direct behavioural measures related to saving and pensions. The first five measures were concerned with factual information while the remaining five were concerned with respondents agreement or disagreement with some statements concerning behaviour and attitudes measured on a five point Likert scale (1=strongly agree; 5=strongly disagree). The cluster analysis results showed six clusters. K-means method in SPSS was applied to cluster the 540 cases. These were interpreted as young cluster; old cluster; female cluster; high level of education and pension membership rate cluster; younger, more highly educated, lower income female cluster; and male and young cluster. The results are important as they show that people need not be classified only based on their gross levels of income as some people are usually in debt while they save in retirement through pensions or regular savings. The cluster analysis used in this study was therefore able to group the large data into six clusters which can inform practice. Ho and Hung (2008) used analytical hierarchy process (AHP), cluster analysis, and correspondence analysis to examine how a graduate institute at can develop effective marketing strategies. The analytic hierarchy process was used to identify the requirements of prospective students. The method identified fourteen factors and reduced them into five categories. Cluster analysis was performed using Ward’s linkage clustering analysis. The analysis was meant to understand the underlying dimensions of student choice based on the AHP selection criteria, and to segment students into identifiable clusters. Five clusters were therefore identified and named appropriately. Then a correspondence analysis was done in order to select target markets and to accomplish market positioning. 4. Method The data are based on a survey of student food shoppers. The survey employed a questionnaire designed to include nominal measures of shopping behaviour, a scale of students’ attitudes to the importance of supermarket features, and nominal measures of students’ characteristics. The survey was conducted at Newcastle University using face-to-face interviews with full-time undergraduates. A quota sample was specified to approximate student representation by gender and faculty. The data consist of 731 valid responses. An analysis applied factor analysis to data in the form of a 14-item scale to measure students’ attitudes to the importance of supermarket features with 708 valid responses. The analysis employed the principal component method with Varimax rotation using the criterion that factors with eigenvalues greater than one would be extracted. The rotated solution established five factors and five factor scores were saved as economy, payment, range and quality of products, friendly staff, and accessibility. The five factor variables are the target variables for cluster analysis. Preliminary analysis using hierarchical clustering identified respectively the possibility solutions in the range five-clusters to two-clusters. As discussed in SPSS (2010), this analysis employed the K- Means optimisation method and specifies a three-cluster solution named storecard, budget, and weekly expenditure. All the behavioural variables were nominal variables. Thus profiles were obtained from crosstabs with the cluster identity variable clusid as the row variable and each of the behavioural variables as column variables 5. Results This chapter presents the results of the study. The chapter begins with a presentation of cluster analysis results. This is followed by a presentation of cluster profiles. Lastly, a summary table of profiles is shown. 5.1 Cluster analysis results Initial five factors namely economy, payment systems, range of quality of products, friendly staff, and accessibility were iterated to give the final three clusters that explain the shopping behaviour. The ANOVA test showed that the cluster center values were equal for all the five factors, p < .01. The cluster size results show three clusters. Cluster 1 is the smallest while cluster 2 is the largest. The results show that 27% of the respondents were grouped into cluster 1, 44% into cluster 2 and 29% into cluster 3. The results show three clusters namely storecard ownership, use of budget and weekly expenditure. 5.2 Cluster profiles From the average factors scores for final cluster centres, it can be noted that cluster 1 members were driven by the payment systems and range of quality products in supermarkets. The important factors to this cluster payments systems, range and quality products, and accessibility. Economy and friendly staff were not important to them. The results show that cluster 2 members were driven more by the economy as well as friendly staff in supermarkets. The important factors to them were economy, friendly staff, payment systems, and range and quality of products. It is also observed that accessibility was not important to this group. Further, the results show that cluster 3 members were driven by accessibility. To this group, payment systems, range and quality of products and friendly staff were unimportant in influencing their shopping behaviour. Economy and accessibility were however important to cluster 3 members. These results are shown in Table 1. Table 1. Average Factor Scores for Final Cluster Centres Factor Cluster 1 2 3 Economy -1.05774 .44846 .28904 Payment systems .40466 .13830 -.58140 Range and quality of products .42505 .16458 -.64004 Friendly staff -.27364 .60530 -.66864 Accessibility .17900 -.27888 .25951 5.4 Profiles Table 2 shows the chi-square results testing the hypothesis that cluster identity and behavioural characteristics were independent. The cross-tabulation results are also discussed for each of the behavioural characteristics. Table 2. Summary of Tests for Cluster Identity and Shopping Behavioural Characteristics Behavioural Characteristic Chi-square Statistic and Significance1 Null Hypothesis Storecard Ownership 2 (2)= 12.387, Sig = 0.002 Reject Use of Budget 2 (2)= 34.602, Sig = 0.000 Reject Weekly Expenditure 2 (4)= 28.595, Sig = 0.000 Reject A hypothesis that storecard ownership was independent among the three clusters was tested. The chi-square test led to the rejection of null hypothesis, χ2(2, n=678) = 12.89, p < .05. There is therefore an association between cluster identity and storecard ownership. This can be explained by the results of cross-tabulation which showed that 50.8% of those in cluster 1 had scorecard ownership, 66.2% of those in cluster 2 had them while 55.5% of those in cluster 3 hard them. A hypothesis that cluster identity and use of budget were independent was tested. The chi-square test for cluster identity and use of budget leads to the rejection of null hypothesis, χ2(2, n=705) = 34.60, p < .05. Cluster identity is therefore associated with use of budget. This can be explained by the results of cross-tabulation which showed that 77.1% of those in cluster 1 did not use the budget, 58.5% of those in cluster 2 did not use the budget while 51.5% of those in cluster 3 used the budget. The hypothesis that cluster identity and weekly expenditure was independent was tested. The chi-square test for cluster identity and weekly expenditure leads to the rejection of null hypothesis, χ2(4, n=691) = 28.60, p < .05. Cluster identity is therefore associated with weekly expenditure. This can be explained by the results of cross-tabulation which showed that there was variance in weekly expenditures. Table 4. Summary of Cluster Profiles Profile Cluster 1 (27%) Cluster 2 (44%) Cluster 3(29%) Descriptive label  Storecard Ownership Use of Budget Weekly Expenditure Importance of store features factors Economy -1.05774 .44846 .28904 Payment systems .40466 .13830 -.58140 Range and Quality of Products .42505 .16458 -.64004 Friendly Staff -.27364 .60530 -.66864 Accessibility .17900 -.27888 .25951 Shopping behaviour measures: Storecard Ownership Payment systems, range and quality of products Use of Budget Economy, friendly staff Weekly Expenditure Accessibility 6. Marketing Implications of Results The results have shown that there are three clusters that explain the shopping behaviour namely storecard ownership, use of budget, and weekly expenditure. As the results showed that there is an association between the cluster identities, marketers can therefore use this to infer that shopping behaviour is driven by these three basic issues of card ownership, budgets, and expenditure. Thus supermarkets that need to their sales should give their shoppers loyalty cards as this drives some customers to shop. Secondly, the marketers in retail business especially supermarkets should understand that some customers are driven by the use of budgets and therefore should have end month sales promotions where customers that use budgets can be driven to shop in the supermarkets. Further, marketing strategies such as the use weekly coupons can be used for those who are driven by weekly expenditures. 7. Summary /Conclusions The aim of this research was to establish the factors that measure students’ attitudes to the importance of supermarket features. This was achieved through a cluster analysis method using hierarchical clustering to identify the range of clusters and the K- Means optimisation method to specify the cluster solution. The results showed that there were three clusters which were named storecard ownership, use of budgets, and weekly expenditure. The storecard ownership members were characterised by payment systems and range and quality of products. The use of budget members were characterised by economy and friendly staff while the weekly expenditure members were characterised by accessibility. The chi-square tests showed that cluster identities were independent among the shopping characteristics. These results are very useful to marketers and especially those in retail business. More specifically they are important to supermarkets as they show the shopping behaviour of customers in supermarkets. This study has therefore presented three clusters upon which supermarket shoppers can be described. The focus on students may be the limitation of the study. More research should be done in the future in this are with a sample drawn from other cities or a nation-wide study in order to establish whether the results hold. Further, a multi-method study can also be performed to establish the shopping behaviours. Thus the use of content analysis and more specifically the correspondence analysis can be done together with cluster analysis. 8. List of References Gough, O, & Sozou, PD 2005, “Pensions and retirement savings: cluster analysis of consumer behaviour and attitudes", International Journal of Bank Marketing, Vol. 23 Iss: 7, pp.558 - 570 Ho, H-F & Hung, C-C 2008, "Marketing mix formulation for higher education: An integrated analysis employing analytic hierarchy process, cluster analysis and correspondence analysis", International Journal of Educational Management, Vol. 22 Iss: 4, pp.328 – 340 Ness, M. R. (2012b). ‘Cluster analysis lecture 2012-2013’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). Ness, M. R. (2012b). ‘Practical 4a: Hierarchical cluster analysis on SPSS 19.0’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). Ness, M. R. (2012c). ‘Practical 4b: K-means cluster analysis on SPSS 19.0’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). Ness, M. R. (2012d). ‘Assignment 2 Cluster Analysis 2012-2013’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). Ness, M. R. (2012e). ‘Guide to the Presentation of Written Work’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). Ness, M. R. (2012f). ‘Guide to the Cluster Analysis Assignment 2012-2013’. MKT3004 Analytical Techniques for Marketing [Online] Available at: http://blackboard.ncl.ac.uk (Accessed: March 11 2013). SPSS 2010, SPSS for Windows (Version 19.0), Chicago, IL, USA: SPSS Inc. 9. Appendices Table 1. Initial Cluster Centres Factor Cluster 1 2 3 Economy -4.77899 1.40149 1.00605 Payment systems -2.06711 1.15148 -3.01248 Range and quality of products 2.59659 1.85017 -2.48593 Friendly staff -.34331 1.73208 -1.80568 Accessibility .03768 1.49838 2.23241 Table 2. Iteration History Iteration Change in Cluster Centers 1 2 3 1 3.750 3.187 3.530 2 .511 .188 .292 3 .327 .137 .145 4 .173 .083 .085 5 .138 .086 .046 6 .098 .065 .039 7 .071 .046 .020 8 .045 .024 .016 9 .021 .018 .012 10 .023 .014 .000 Table 3. Final Cluster Centres Factor Cluster 1 2 3 Economy -1.05774 .44846 .28904 Payment systems .40466 .13830 -.58140 Range and quality of products .42505 .16458 -.64004 Friendly staff -.27364 .60530 -.66864 Accessibility .17900 -.27888 .25951 Table 4. Descriptive ANOVA Factor Cluster Error F Sig. Mean Square df Mean Square df Economy 145.807 2 .589 705 247.467 .000 Payment systems 53.284 2 .852 705 62.564 .000 Range and quality of products 63.507 2 .823 705 77.195 .000 Friendly staff 110.464 2 .689 705 160.217 .000 Accessibility 22.137 2 .940 705 23.549 .000 Table 5. Cluster Size Number Percent Cluster 1 189.000 27 2 313.000 44 3 206.000 29 Valid 708.000 100 Missing 23.000 Table 3. Final Cluster Centres Factor Cluster 1 2 3 Economy -1.05774 .44846 .28904 Payment systems .40466 .13830 -.58140 Range and quality of products .42505 .16458 -.64004 Friendly staff -.27364 .60530 -.66864 Accessibility .17900 -.27888 .25951 Table 6. Crosstab Cluster Identity by Storecard Ownership Storecard Total Yes No Cluster identity Cluster 1 Count 91 88 179 % within Cluster identity 50.8% 49.2% 100.0% Cluster 2 Count 198 101 299 % within Cluster identity 66.2% 33.8% 100.0% Cluster 3 Count 111 89 200 % within Cluster identity 55.5% 44.5% 100.0% Total Count 400 278 678 % within Cluster identity 59.0% 41.0% 100.0% Table 7. Chi-square Test for Cluster Identity and Storecard Ownership Value df Asymp. Sig. (2-sided) Pearson Chi-Square 12.387a 2 .002 Likelihood Ratio 12.447 2 .002 Linear-by-Linear Association .632 1 .427 N of Valid Cases 678 Table 8. Crosstab Cluster Identity by Use of Budget Budget Total Yes No Cluster identity Cluster 1 Count 43 145 188 % within Cluster identity 22.9% 77.1% 100.0% Cluster 2 Count 129 182 311 % within Cluster identity 41.5% 58.5% 100.0% Cluster 3 Count 106 100 206 % within Cluster identity 51.5% 48.5% 100.0% Total Count 278 427 705 % within Cluster identity 39.4% 60.6% 100.0% Table 9. Chi-square Test for Cluster Identity and Use of Budget Value df Asymp. Sig. (2-sided) Pearson Chi-Square 34.602a 2 .000 Likelihood Ratio 35.959 2 .000 Linear-by-Linear Association 33.201 1 .000 N of Valid Cases 705 Table 10. Crosstab Cluster Identity by Weekly Expenditure Weekly Expenditure Total £0-15 £16-30 £31+ Cluster identity Cluster 1 Count 36 112 40 188 % within Cluster identity 19.1% 59.6% 21.3% 100.0% Cluster 2 Count 112 142 44 298 % within Cluster identity 37.6% 47.7% 14.8% 100.0% Cluster 3 Count 80 107 18 205 % within Cluster identity 39.0% 52.2% 8.8% 100.0% Total Count 228 361 102 691 % within Cluster identity 33.0% 52.2% 14.8% 100.0% Table 11. Chi-square Test for Cluster Identity and Weekly Expenditure Value df Asymp. Sig. (2-sided) Pearson Chi-Square 28.595a 4 .000 Likelihood Ratio 30.497 4 .000 Linear-by-Linear Association 22.649 1 .000 N of Valid Cases 691 Read More