- Email: [email protected]
An innovation model of alumni relationship management: Alumni segmentation analysis
Kasetsart Journal of Social Sciences xxx (2017) 1e11
Contents lists available at ScienceDirect
Kasetsart Journal of Social Sciences journal homepage: http://www.elsevier.com/locate/kjss
An innovation model of alumni relationship management: Alumni segmentation analysis Natthawat Rattanamethawong a, Sukree Sinthupinyo b, *, Achara Chandrachai c a
Technopreneurship and Innovation Management Program, Chulalongkorn University, Bangkok, Thailand Faculty of Computer Engineering, Chulalongkorn University, Bangkok, Thailand c Faculty of Commerce and Accountancy, Chulalongkorn University, Bangkok, Thailand b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 October 2016 Accepted 22 February 2017 Available online xxxx
The purpose of this study was to cluster alumni into segments to better understand the alumni's characteristics, lifestyles, types of behavior, and interests. A sample of 300 university alumni records was used to obtain their respective attribute values consisting of demographics, preferred communication channels, lifestyle, activities/interests, and expectation from university, needed information, donation willingness, and frequency of contact. The researcher used logistic regression and the k-mean clustering technique to analyze the data from the survey. Five segments could be derived from the analysis. Segment 3, the so-called “Mid Age Religious” contained the highest portion while segment 5, the so-called “Elaborate Cohort” had the least portion. Most of the population under these two segments was female. Differences were identified in age, marital status, education, occupation, position, income, experience, and field of work. The Elaborate Cohort segment represented young females having a bachelor degree, with low experience and low income, working for their first employer, and still enjoying being single. Another segment with similar values of attributes as the Elaborate Cohort was segment 1, the socalled “Activist Mainstreamer” whose field of work was computer technology. The segment called “Senior League” consisted of members older than 41 years like the Mid Age Religious segment, however all members were male. The last segment, the so-called “Passionate Learner” had members aged between 31 and 40 years. In conclusion, the results of this study can assist in formulating strategic marketing by alumni associations to satisfy and engage their alumni. © 2017 Kasetsart University. Publishing services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/ 4.0/).
Keywords: cluster data mining segmentation analysis university alumni
Introduction Alumni Relationship Management (ARM) is an essential agenda item for every alumni association or university. Education institutions are understanding the importance of services with the right information and communication * Corresponding author. E-mail addresses: [email protected] (N. Rattanamethawong), [email protected] (S. Sinthupinyo), [email protected] (A. Chandrachai). Peer review under responsibility of Kasetsart University.
(Ahmadi, Nilashi, Ibrahim, Rad, & Pourhashemi, 2013). Facebook and other social media are being used to solicit and thank alumni in order to reach out and have a greater impact (Masterson, 2012). The distribution of universities varies throughout Thailand. Han, Kamber, and Pei (2012) stated that data mining detection is useful in this alumni communication because there are very likely natural groupings that may represent segments of the alumni who have much in common, and for which customized marketing is justified (Le Blance & Rucks, 2009).
http://dx.doi.org/10.1016/j.kjss.2017.02.002 2452-3151/© 2017 Kasetsart University. Publishing services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
2
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Since Thailand's first modern institution, namely Chulalongkorn University opened by King Chulalongkorn back in 1902, alumni associations have been established; however, involvement and collaboration are still rare mainly because: 1) the alumni do not feel that universities have provided them with useful information and benefits, 2) the alumni are tracked only to seek funding contributions to universities, 3) the universities tend to treat successful and older alumni members better than new and less successful ones, and 4) the relationship among professors, existing students, and the alumni is very weak as there is no attachment to the university after graduation. There are additional reasons blocking the relationship between a university and its alumni. The main objective of this research was to overcome the above situation and strengthen the relationship between the university and its alumni. The author believes that the social network capability is able to support collaboration among universities and alumni members. The social network is a new channel for alumni members to use to communicate and to collaborate. However the author needed to study and analyze which attributes offer value to the alumni and will attract them to join and to be connected. The questions which the author aimed to answer in this research were:
1) How many different segments are alumni in Thailand clustered into? 2) Which attributes influence the correlations among the alumni members? 3) What are the characteristics and behavior of each segment? Literature Review Currently, some universities focus on alumni relationship management. Researchers have studied the relationship between a university and its alumni in order to improve and engagement the alumni. Hanson (2000) studied the relationship of selected student demographics, student academic involvement, student social involvement, alumni demographics, alumni social involvement, and alumni attitudinal measures with alumni supportive behavior. The researcher developed a conceptual model from his previous research to predict the alumni's support. The university uses the model to predict student and alumni supportive behavior from when students entered the university. Other engagement research examined the relationship between alumni engagement and two categories of variables, alumni characteristics and alumni giving behavior (Radcliffe). The researcher gathered information such as alumni demographics, experiences, and attitudes. Data from the survey were analyzed to improve the ability to identify alumni donors and promoters. The results illustrated that there was no difference in the engagement scores based on alumni characteristics and that the engagement score had a positive correlation to a variety of types of giving behavior, donor status, recent donor status, annual giving behavior, and adjusted lifetime giving.
There are several models used for data clustering. One study provided a framework for market segmentation to determine the right target customers (Larsen, 2010). The author investigated two models: 1) the Minerva model which divided customers into five different lifestyle segments based on their values. These segments was designated a color; blue, green, rose, violet, and gray; and 2) the Mosaic model combined variables from the geographic segmentation and the demographic segmentation. These two models were then used to analyze and determine the right target customers. Data mining using a neural network clustering technique has been used to identify high profit, high value, and low risk customers (Rajagopal, 2011). The results identified four clusters based on value. Cluster 3 had high revenue, high cost customers and was classified as the high value cluster. Cluster 1 had high revenue and high cost and was classified as medium value. Cluster 2 had low revenue and low cost and was classified as low value. The clustering results suggested strategies to retain or move the customers from the lower band to the upper band. Prasanna and Peddyreddy (2015) identified high profit, high value, and low risk customers using a data mining technique called clustering. The research delivered information by extracting data from transactions classified by the customer's revenue cluster. The result was very useful in increasing sales. Durango-Cohen and Balasubramanian (2014) presented a finite mixture model framework to classify the alumni population of a university in the Midwestern United States based on the monetary value of annual contributions. The results were supportive of the design of segment tailored solicitations and supported the allocation of resources for fundraising. Rattanamethawong, Sinthupinyo, and Chandrachai (2015) studied the factors which impacted the relationship between the university and students/alumni and proposed the conceptual framework illustrated in Figure 1. The model in Figure 2 agrees with this model especially regarding the component “Identify your alumni members as individuals and let them feel you know them” and the component “Information to alumni, about your ability to cope with their requirement”. These two components are similar to this study's ARM framework. However, the ARM framework is different from the new CRM strategy in “Identify your alumni members as individuals and make them feel you know them” and “Information to alumni, about your ability to satisfy their requirement”.
Materials and Methods Sample Sampling in this study engaged 300 alumni in Thailand from all around Thailand with 222 (74%) from the Bangkok Metropolitan area, 25 (8.33%) from central, 19 (6.33%) from northeastern, 15 (5%) from northern, 15 (5%) from eastern, and 4 (1.33%) from southern Thailand. All respondents successfully completed the survey which was launched via an online Google form survey.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Figure 1 Alumni relationship management conceptual framework
Figure 2 New CRM strategy model. Note. From Alumni Liaison Unit (ALU), University Technology Malaysia (UTM)
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
3
4
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Instruments The survey in this study comprised six sections. The first section asked for general information (gender, age range, marital status, number of children, education level, faculty, occupation, job position, field of work, income range, experience (years), and region of living). The second examined habits regarding the use of social media. The third, fourth, fifth, and sixth sections of the survey assessed the level of alumni lifestyle, interests/activities, expectations from the alumni association or university, and preferred information to receive. Methods The researcher used a two-step segmentation approach to determine the optimal number of alumni segments and the unique characteristics of each segment. Step 1: Pre clustering. The researcher used logistic regression statistics for data set deduction. Step 2: Clustering. A k-mean clustering technique was used to determine the optimum number of clusters and to classify the alumni database into segments.
The dataset was then re structured based on the selected cluster model and all the variables were tested against each other to determine the key differences between segments. Discriminant analysis was used to check the goodness-of-fit of the model that the cluster analysis had found and of the profiles of the clusters. In almost all analyses, discriminant analysis follows cluster analysis as the cluster analysis does not have any goodness-of-fit measures or tests of significance. Cluster analysis relies on discriminant analysis to check if the groups are statistically significant and if the variables significantly discriminate between the groups. The significance test probability value reflects the strength of the evidence against the null hypothesis. If the probability is below .01, the data provide strong evidence that the null hypothesis is false. If the probability value is below .05 but larger than .01, then the null hypothesis is typically rejected, but not with as much confidence as it would be if the probability value were below .01. Probability values between .05 and .10 provide weak evidence against the null hypothesis and, by convention, are not considered low enough to justify rejecting it, and therefore are also considered when determining the differences among clusters, although with lesser priority. Analysis and Results
Figure 3 presents the alumni clustering procedure which the researcher applied. The research started by distributing the questionnaire to 300 alumni members from multi-universities around Thailand in order to gather information such as demographics, lifestyle, and interests/activities. In the pre clustering step, the author scanned all variables one by one and decided which variables contributed to the likelihood to respond to participation in social media activities. The researcher used a logistic regression methodology and the determining variables were selected using p-value statistic (<.05). In the clustering step, all the determining variables from the pre clustering steps were set as variables to determine the optimum number of clusters through the k-means methodology or centroidbased segmentation, where the k clusters centers are assigned with the objects to the nearest center, such that the squared distances from the clusters are minimized. The author tested 3, 4, and 5 clusters and concluded with the 5 cluster model because it had the lowest SSE (sum of squared errors) and the smallest segment from the analysis (segment 5) still fulfilled the smallest sample size (>30) requirement for the significance test among clusters.
After conducting pre clustering, the author tested the possible maximum numbers of groups which would still provide a minimum SSE. Moreover the author considered three components: 1) a good distribution of respondents across the segments, to avoid having 90 percent of all people in one segment; 2) an output with a lower SSE (doesn't need to be the lowest, but avoid the highest); and 3) most importantly, look for means per segment that make sense with the variability covering all scales. The author ran the k-means analysis and the output is shown in Tables 1e4 (from two segments to five segments, respectively). The output shown in Table 1 does have a well-defined population distribution. Segment 1 consists of 62 percent of respondents which was considered to be too high, while the results in Table 2 still define segment 2 as having 44.7 percent of all respondents. The results for four clusters and five clusters are shown in Tables 3 and 4 and these were subjected to further analysis. The values of SSE were compared for the results with four and five clusters and five clusters offered a lower SSE (1,118.1) than that for four clusters (1,239.3) from Tables 4
Figure 3 Alumni clustering step
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
5
Table 1 Output for two clusters Mean/centroid
Age
Segment 1 Segment 2 AVERAGE Respondents Segment 1 Segment 2 Total
2.91 3.49 3.13 Number 186 114 300
Education 2.53 2.49 2.51 % 62.0 38.0 100.0
Experience
Travelling
Game/social
Gardening
3.45 3.93 3.63 SSE/segment 1189.4 609.5
3.34 2.44 3.00
3.80 1.63 2.98
3.30 2.05 2.83
SSE total 1789.7
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 2 Output for three clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 AVERAGE Respondents Segment 1 Segment 2
2.33 3.97 2.71 3.13 Number 110 134
Education
% 36.7 44.7
Segment 3 Total
56 300
18.7 100.0
2.51 2.60 2.32 2.51
Experience
Travelling
Game/social
Gardening
2.90 4.52 2.95 3.63 SSE/segment 529.2 685.5
3.54 2.96 2.02 3.00
4.27 2.42 1.77 2.98
3.08 3.16 1.52 2.83
SSE total 1445.7
231.1
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 3 Output for four clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 Segment 4 AVERAGE Respondents Segment 1 Segment 2 Segment 3 Segment 4 Total
2.31 3.96 3.73 2.25 3.13 Number 94 89 73 44 300
Education 2.48 2.61 2.60 2.25 2.51 % 31.3 29.7 24.3 14.7 100.0
Experience
Travelling
Game/social
Gardening
2.90 4.52 4.36 2.20 3.63 SSE/segment 410.2 393.2 231.3 204.6
3.49 3.22 2.37 2.52 3.00
4.39 2.74 1.86 2.27 2.98
3.27 3.89 1.64 1.70 2.83
SSE total 1239.3
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 4 Output for five clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 AVERAGE Respondents Segment 1 Segment 2
2.09 3.09 4.05 3.75 2.09 3.13 Number 65 66
Education
% 21.7 22.0
Segment 3 Segment 4 Segment 5 Total
76 59 34 300
25.3 19.7 11.3 100.0
2.40 2.76 2.54 2.59 2.06 2.51
Experience
Travelling
Game/social
Gardening
2.60 3.74 4.55 4.39 2.03 3.63 SSE/segment 274.2 222.0
3.40 3.70 3.07 2.17 2.15 3.00
4.54 3.52 2.66 1.58 2.09 2.98
3.66 2.15 4.07 1.66 1.79 2.83
SSE total 1118.1
330.0 134.5 134.5
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
6
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
and 3, respectively. The last analysis confirmed the author's decision as the means per segment in Table 4 had better variation than those of Table 3; therefore, this analysis showed that there were five clusters among the sampled alumni. In conclusion, Table 4 shows five clusters with an appropriate distribution of segments (21.7% in segment 1, 22.0% in segment 2, 25.3% in segment 3, 19.7% in segment 4, and 11.3% in segment 5). The five segments also produced the lowest SSE and had the right variation of k-mean in each segment. Alumni Segmentation Five segments were derived from the research. The characteristics of each segment are summarized below and in Figures 4e10. Segment 1 contained mixed gender alumni aged less than 30 years, enjoying their single status and currently pursuing a further degree while still working for their first employer and being paid less than THB 30,000 per month. The members of this segment work in computer technology with less than 3 years' working experience. The author named this segment “Activist Mainstreamer”. Segment 2 contained mixed gender alumni aged between 31 and 40 years, enjoying their single status and most of them undertaking their Master degree with their salary being THB 50,000e200,000 per month. The members of this segment had worked in many fields for more than 10 years but less than 19 years. The author named this segment “Passionate Learner”. Segment 3 contained female alumni aged more than 41 years, enjoying their married status and currently working at a management level for a government entity. This
segment consisted of Bachelor and Master degree holders with salary range of THB 50,000e200,000 per month, work experience of more than 20 years, and the majority were nurses. The author named this segment “Mid Age Religious”. Segment 4 contained male alumni aged more than 41 years, enjoying married life. The members of this segment held either a Bachelor or Master degree, had been running their business for more than 20 years and had earnings of more than THB 200,000 per month. The author named this segment “Senior League”. Segment 5 contained female alumni aged less than 30 years, enjoying their single status and currently pursuing a further degree while still working for their first employer and being paid less than THB 30,000 per month. The members of this segment worked in a variety of fields and had less than 3 years' working experience. The author named this segment “Elaborate Cohort”. The Mid Age Religious segment had the highest number of alumni (25%) followed by the Passionate Learner and Activist Mainstreamer (22%) segments. The least alumni were in the Elaborate Cohort segment which accounted for only 11 percent, while the Senior League segment contained the remaining members. Other research has reported that each segment has different interests and preferences (Ahmadi et al., 2013). The Activist Mainstreamer segment loves to associate with social media like Facebook, Line, and Instagram. The members of this segment are active in many key activities including: exercise, walking, running, cycling, football, basketball, volleyball, tennis, badminton, table tennis, watching sport, traveling, ecotourism, temple tourism, dinner parties, cinema, concert, theater, music, karaoke,
Figure 4 Alumni demographics segment highlight
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
7
Figure 5 Five alumni segment highlight
photography, painting, social games, playing music, gardening, and household duties. The members are interested in seeking information on hot news, travel news, business and marketing news, entertainment news, football news, lifestyle news, food and beverage news, and job news. They are also willing to support universities with new graduate recruitment,
special lectures, and/or sport participation. They also expect universities to offer sports cantering service, free parking, venue rental, job search service, and innovation hub. The members of the Activist Mainstreamer segment are very active people seeking knowledge and they love to socialize with others. The author would recommend adopting the members in this segment as the first priority.
Figure 6 Activist mainstreamer segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
8
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Figure 7 Interests of passionate learner segment
The Passionate Learner segment has similar characteristics to the first segment. However, the members of this segment are less active due to their older age. Activities which they like are: exercise, walking, running, cycling,
adventure, watching sport, traveling, ecotourism, temple tourism, dinner parties, and social games. The members of this segment are seeking information on business and marketing news as well as news related to
Figure 8 Mid age religious segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
9
Figure 9 Senior league segment
economics. They expect the university to offer a consultancy service. The author would recommend universities adopt these members. The Mid Age Religious segment consists of members who mostly are house wives. The key activities are related to the
family such as traveling, dinner parties, cooking, gardening, and charity work. They do not expect any support or information from universities but prefer to have reunions so that they can meet old friends and professors. The author thinks that the members of this segment will be interested in
Figure 10 Elaborate cohort segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
10
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
associating with the university's professors and other alumni members even though it is in a passive mode. The Senior League segment consists of elder members who once-in-a-while utilize social media. However, they are not interested in any activities and do not expect anything from universities. The author feels that it is very hard to get these members to associate with universities mainly because they have limited free time since most of them are business owners and family heads. However they are good for donations to charity (Clotfelter, 1985). The Elaborate Cohort is another segment with members who do not want to participate in any activities and do not expect anything from universities. Based on their demographics (gender, age, marital status, position, income, experience), they are similar to the Activist Mainstreamer segment. However the educations level of this segment contains more at the Bachelor level compared to the Activist Mainstream segment who had graduated with both Bachelor and Master degrees. Alumni in the Activist Mainstreamer segment mostly work in the computer and technology field. Moreover the Elaborate Cohort segment has less interest in playing on social media compared to other segments. The author considers that universities should give these members a very low priority regarding their current association with alumni activities or on the other hand, universities should pay attention more to engage this segment back.
alumni more deeply by clustering the customers or alumni into segments based on demographics and alumni behavior. The results of alumni clustering can help a university to communicate with its alumni using appropriate information and the proper channels to get a high response rate. In conclusion, each segment has different needs and wants. The research has clearly answered the questions mentioned during introductory part. Clustering alumni into segments and understanding their characteristics/behavior are important factors in obtaining a successful outcome. However, the author will conduct further studies to complete the objectives including developing a collaboration model and adoption strategies for each segment. Conflict of interest The authors declare that there is no conflict of interest regarding the publication of this paper. Acknowledgments This research was undertaken using a grant by the Graduate School of Chulalongkorn University, Thailand under the program “the 90th Anniversary of Chulalongkorn University Fund” (Ratchadaphiseksomphot Endowment Fund).
Discussion This research aimed to classified alumni with the same preferences into segments. The methodology designed the instruments to extract the alumni's demographics, lifestyles, expectations, and interests. The researcher expected that the designed instruments could collect data and enumerate the data using the chosen data mining technique. The results from simple k-mean clustering provided the alumni's preferences in segments and was consistent with the concept of the research. For instance the majority of the Activist Mainstreamer cluster consisted of singles under 30 years old. They are interested in a variety of activities such as exercise, running, cycling, football, and tennis and expected the university to provide them with a sport center service. Most of them were a first jobber and they were interested in recruiting graduates activities by the alumni association and expected a job search service as well. They used Facebook and LINE to communicate. The choice of answers in the instruments highlighted the Mid Age Religious cluster. Alumni in this cluster or segmentation were mainly females aged more than 41 years. The lifestyles choices they chose are in the survey were temple tourism, chapel visiting, being involved in charities, and praying. They were not interested in any activities proposed by the university. Moreover, they expected nothing from the university. These results were able to answer the research questions. Conclusion and Future Research Alumni clustering is one methodology that a company or university can use to understand their customers or
References Ahmadi, H., Nilashi, M., Ibrahim, O., Rad, M. S., & Pourhashemi, S. O. S. (2013). Customer relationship management model in higher education: A case of Malaysia. Today Science Journal of Humanity, 1(1), 1e10. Retrieved from http://www.tsjournals.com/Article/2013/1/Customer% 20Relationship%20Management%20Model%20in%20Higher% 20Education%20A%20Case%20of%20Malaysia.pdf. Clotfelter, C. T. (1985). Federal tax policy and charitable giving. Chicago, IL: University of Chicago Press. Retrieved from http://papers.nber.org/ books/clot85-1. Durango-Cohen, E. J., & Balasubramanian, S. K. (2014). Effective segmentation of university alumni: Mining contribution data with finitemixture models. Research in Higher Education, 56, 78e104. Retrieved from http://www.scopus.com/inward/record.url?eid¼2-s2.0-84901580676&partnerID¼40&md5¼b271fa0bfc55ea84b325b4fab5f01157. Han, J., Kamber, M., & Pei, J. (2012). Data mining concepts and techniques (3rd ed.). Burnaby, BC, Canada: Morgan Kaufmann Publishers. Retrieved from https://web.engr.illinois.edu/~hanj/bk2/toc.pdf. Hanson, S. K. (2000). Alumni characteristics that predict promoting and donating to alma mater: Implications for alumni relations (Unpublished doctoral dissertation). Grand Forks, ND: University of North Dakota. Larsen, N. (2010). Market segmentation e A framework for determining the right target customer (Bachelor’s thesis). Aarhus: Aarhus University. Retrieved from http://pure.au.dk/portal/files/11462/BA.pdf. Le Blance, L. A., & Rucks, C. T. (2009). Data mining of university philanthropic giving: Cluster-discriminant analysis and Pareto effects. International Journal of Educational Advancement, 9(2), 64e82. Retrieved from https://www.researchgate.net/profile/Louis_Le_Blanc/publication/ 263326511_Data_mining_of_university_philanthropic_giving_Clusterdiscriminant_analysisand_Pareto_effects/links/56ab70d208aed5a0135 c1598.pdf?origin¼publication_list. Masterson, K. (2012). 5 Colleges that inspire alumni giving, and how they do it. Chronicle of Higher Education, 1, 1e4. Retrieved from http://www. fulbright.fi/sites/default/files/Liitetiedostot/Palvelut-asiantuntijoille/ Opintomatka- alumni/recent-news-articles-chronicle-alumni-relationsand-institutional-giving.pdf. Prasanna, K. V., & Peddyreddy, P. B. (2015). Mining customer transaction data using clustering. International Journal of Electronics and Computer System, 3(11), 41e44. Retrieved from www.irdindia.in/journal_ijeecs/ pdf/vol3_iss11/10.pdf.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11 Radcliffe S.K., A study of alumni engagement and its relationship to giving behaviors, (Master's thesis). Bucknell University, Lewisburg, PA. Retrieved from http://digitalcommons.bucknell.edu/cgi/viewcontent. cgi?article¼1001&context¼masters_theses Rajagopal, S. (2011). Customer data clustering using data mining technique. International Journal Database Management System, 3(4),
11
1e11. Retrieved from http://airccse.org/journal/ijdms/papers/ 3411ijdms01.pdf. Rattanamethawong, V., Sinthupinyo, S., & Chandrachai, A. (2015). An innovation systems that can quickly responses to the needs of students and alumni. Procedia e Social and Behavioral Sciences, 182, 645 e652. Retrieved from www.sciencedirect.com.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
Contents lists available at ScienceDirect
Kasetsart Journal of Social Sciences journal homepage: http://www.elsevier.com/locate/kjss
An innovation model of alumni relationship management: Alumni segmentation analysis Natthawat Rattanamethawong a, Sukree Sinthupinyo b, *, Achara Chandrachai c a
Technopreneurship and Innovation Management Program, Chulalongkorn University, Bangkok, Thailand Faculty of Computer Engineering, Chulalongkorn University, Bangkok, Thailand c Faculty of Commerce and Accountancy, Chulalongkorn University, Bangkok, Thailand b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 October 2016 Accepted 22 February 2017 Available online xxxx
The purpose of this study was to cluster alumni into segments to better understand the alumni's characteristics, lifestyles, types of behavior, and interests. A sample of 300 university alumni records was used to obtain their respective attribute values consisting of demographics, preferred communication channels, lifestyle, activities/interests, and expectation from university, needed information, donation willingness, and frequency of contact. The researcher used logistic regression and the k-mean clustering technique to analyze the data from the survey. Five segments could be derived from the analysis. Segment 3, the so-called “Mid Age Religious” contained the highest portion while segment 5, the so-called “Elaborate Cohort” had the least portion. Most of the population under these two segments was female. Differences were identified in age, marital status, education, occupation, position, income, experience, and field of work. The Elaborate Cohort segment represented young females having a bachelor degree, with low experience and low income, working for their first employer, and still enjoying being single. Another segment with similar values of attributes as the Elaborate Cohort was segment 1, the socalled “Activist Mainstreamer” whose field of work was computer technology. The segment called “Senior League” consisted of members older than 41 years like the Mid Age Religious segment, however all members were male. The last segment, the so-called “Passionate Learner” had members aged between 31 and 40 years. In conclusion, the results of this study can assist in formulating strategic marketing by alumni associations to satisfy and engage their alumni. © 2017 Kasetsart University. Publishing services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/ 4.0/).
Keywords: cluster data mining segmentation analysis university alumni
Introduction Alumni Relationship Management (ARM) is an essential agenda item for every alumni association or university. Education institutions are understanding the importance of services with the right information and communication * Corresponding author. E-mail addresses: [email protected] (N. Rattanamethawong), [email protected] (S. Sinthupinyo), [email protected] (A. Chandrachai). Peer review under responsibility of Kasetsart University.
(Ahmadi, Nilashi, Ibrahim, Rad, & Pourhashemi, 2013). Facebook and other social media are being used to solicit and thank alumni in order to reach out and have a greater impact (Masterson, 2012). The distribution of universities varies throughout Thailand. Han, Kamber, and Pei (2012) stated that data mining detection is useful in this alumni communication because there are very likely natural groupings that may represent segments of the alumni who have much in common, and for which customized marketing is justified (Le Blance & Rucks, 2009).
http://dx.doi.org/10.1016/j.kjss.2017.02.002 2452-3151/© 2017 Kasetsart University. Publishing services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
2
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Since Thailand's first modern institution, namely Chulalongkorn University opened by King Chulalongkorn back in 1902, alumni associations have been established; however, involvement and collaboration are still rare mainly because: 1) the alumni do not feel that universities have provided them with useful information and benefits, 2) the alumni are tracked only to seek funding contributions to universities, 3) the universities tend to treat successful and older alumni members better than new and less successful ones, and 4) the relationship among professors, existing students, and the alumni is very weak as there is no attachment to the university after graduation. There are additional reasons blocking the relationship between a university and its alumni. The main objective of this research was to overcome the above situation and strengthen the relationship between the university and its alumni. The author believes that the social network capability is able to support collaboration among universities and alumni members. The social network is a new channel for alumni members to use to communicate and to collaborate. However the author needed to study and analyze which attributes offer value to the alumni and will attract them to join and to be connected. The questions which the author aimed to answer in this research were:
1) How many different segments are alumni in Thailand clustered into? 2) Which attributes influence the correlations among the alumni members? 3) What are the characteristics and behavior of each segment? Literature Review Currently, some universities focus on alumni relationship management. Researchers have studied the relationship between a university and its alumni in order to improve and engagement the alumni. Hanson (2000) studied the relationship of selected student demographics, student academic involvement, student social involvement, alumni demographics, alumni social involvement, and alumni attitudinal measures with alumni supportive behavior. The researcher developed a conceptual model from his previous research to predict the alumni's support. The university uses the model to predict student and alumni supportive behavior from when students entered the university. Other engagement research examined the relationship between alumni engagement and two categories of variables, alumni characteristics and alumni giving behavior (Radcliffe). The researcher gathered information such as alumni demographics, experiences, and attitudes. Data from the survey were analyzed to improve the ability to identify alumni donors and promoters. The results illustrated that there was no difference in the engagement scores based on alumni characteristics and that the engagement score had a positive correlation to a variety of types of giving behavior, donor status, recent donor status, annual giving behavior, and adjusted lifetime giving.
There are several models used for data clustering. One study provided a framework for market segmentation to determine the right target customers (Larsen, 2010). The author investigated two models: 1) the Minerva model which divided customers into five different lifestyle segments based on their values. These segments was designated a color; blue, green, rose, violet, and gray; and 2) the Mosaic model combined variables from the geographic segmentation and the demographic segmentation. These two models were then used to analyze and determine the right target customers. Data mining using a neural network clustering technique has been used to identify high profit, high value, and low risk customers (Rajagopal, 2011). The results identified four clusters based on value. Cluster 3 had high revenue, high cost customers and was classified as the high value cluster. Cluster 1 had high revenue and high cost and was classified as medium value. Cluster 2 had low revenue and low cost and was classified as low value. The clustering results suggested strategies to retain or move the customers from the lower band to the upper band. Prasanna and Peddyreddy (2015) identified high profit, high value, and low risk customers using a data mining technique called clustering. The research delivered information by extracting data from transactions classified by the customer's revenue cluster. The result was very useful in increasing sales. Durango-Cohen and Balasubramanian (2014) presented a finite mixture model framework to classify the alumni population of a university in the Midwestern United States based on the monetary value of annual contributions. The results were supportive of the design of segment tailored solicitations and supported the allocation of resources for fundraising. Rattanamethawong, Sinthupinyo, and Chandrachai (2015) studied the factors which impacted the relationship between the university and students/alumni and proposed the conceptual framework illustrated in Figure 1. The model in Figure 2 agrees with this model especially regarding the component “Identify your alumni members as individuals and let them feel you know them” and the component “Information to alumni, about your ability to cope with their requirement”. These two components are similar to this study's ARM framework. However, the ARM framework is different from the new CRM strategy in “Identify your alumni members as individuals and make them feel you know them” and “Information to alumni, about your ability to satisfy their requirement”.
Materials and Methods Sample Sampling in this study engaged 300 alumni in Thailand from all around Thailand with 222 (74%) from the Bangkok Metropolitan area, 25 (8.33%) from central, 19 (6.33%) from northeastern, 15 (5%) from northern, 15 (5%) from eastern, and 4 (1.33%) from southern Thailand. All respondents successfully completed the survey which was launched via an online Google form survey.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Figure 1 Alumni relationship management conceptual framework
Figure 2 New CRM strategy model. Note. From Alumni Liaison Unit (ALU), University Technology Malaysia (UTM)
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
3
4
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Instruments The survey in this study comprised six sections. The first section asked for general information (gender, age range, marital status, number of children, education level, faculty, occupation, job position, field of work, income range, experience (years), and region of living). The second examined habits regarding the use of social media. The third, fourth, fifth, and sixth sections of the survey assessed the level of alumni lifestyle, interests/activities, expectations from the alumni association or university, and preferred information to receive. Methods The researcher used a two-step segmentation approach to determine the optimal number of alumni segments and the unique characteristics of each segment. Step 1: Pre clustering. The researcher used logistic regression statistics for data set deduction. Step 2: Clustering. A k-mean clustering technique was used to determine the optimum number of clusters and to classify the alumni database into segments.
The dataset was then re structured based on the selected cluster model and all the variables were tested against each other to determine the key differences between segments. Discriminant analysis was used to check the goodness-of-fit of the model that the cluster analysis had found and of the profiles of the clusters. In almost all analyses, discriminant analysis follows cluster analysis as the cluster analysis does not have any goodness-of-fit measures or tests of significance. Cluster analysis relies on discriminant analysis to check if the groups are statistically significant and if the variables significantly discriminate between the groups. The significance test probability value reflects the strength of the evidence against the null hypothesis. If the probability is below .01, the data provide strong evidence that the null hypothesis is false. If the probability value is below .05 but larger than .01, then the null hypothesis is typically rejected, but not with as much confidence as it would be if the probability value were below .01. Probability values between .05 and .10 provide weak evidence against the null hypothesis and, by convention, are not considered low enough to justify rejecting it, and therefore are also considered when determining the differences among clusters, although with lesser priority. Analysis and Results
Figure 3 presents the alumni clustering procedure which the researcher applied. The research started by distributing the questionnaire to 300 alumni members from multi-universities around Thailand in order to gather information such as demographics, lifestyle, and interests/activities. In the pre clustering step, the author scanned all variables one by one and decided which variables contributed to the likelihood to respond to participation in social media activities. The researcher used a logistic regression methodology and the determining variables were selected using p-value statistic (<.05). In the clustering step, all the determining variables from the pre clustering steps were set as variables to determine the optimum number of clusters through the k-means methodology or centroidbased segmentation, where the k clusters centers are assigned with the objects to the nearest center, such that the squared distances from the clusters are minimized. The author tested 3, 4, and 5 clusters and concluded with the 5 cluster model because it had the lowest SSE (sum of squared errors) and the smallest segment from the analysis (segment 5) still fulfilled the smallest sample size (>30) requirement for the significance test among clusters.
After conducting pre clustering, the author tested the possible maximum numbers of groups which would still provide a minimum SSE. Moreover the author considered three components: 1) a good distribution of respondents across the segments, to avoid having 90 percent of all people in one segment; 2) an output with a lower SSE (doesn't need to be the lowest, but avoid the highest); and 3) most importantly, look for means per segment that make sense with the variability covering all scales. The author ran the k-means analysis and the output is shown in Tables 1e4 (from two segments to five segments, respectively). The output shown in Table 1 does have a well-defined population distribution. Segment 1 consists of 62 percent of respondents which was considered to be too high, while the results in Table 2 still define segment 2 as having 44.7 percent of all respondents. The results for four clusters and five clusters are shown in Tables 3 and 4 and these were subjected to further analysis. The values of SSE were compared for the results with four and five clusters and five clusters offered a lower SSE (1,118.1) than that for four clusters (1,239.3) from Tables 4
Figure 3 Alumni clustering step
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
5
Table 1 Output for two clusters Mean/centroid
Age
Segment 1 Segment 2 AVERAGE Respondents Segment 1 Segment 2 Total
2.91 3.49 3.13 Number 186 114 300
Education 2.53 2.49 2.51 % 62.0 38.0 100.0
Experience
Travelling
Game/social
Gardening
3.45 3.93 3.63 SSE/segment 1189.4 609.5
3.34 2.44 3.00
3.80 1.63 2.98
3.30 2.05 2.83
SSE total 1789.7
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 2 Output for three clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 AVERAGE Respondents Segment 1 Segment 2
2.33 3.97 2.71 3.13 Number 110 134
Education
% 36.7 44.7
Segment 3 Total
56 300
18.7 100.0
2.51 2.60 2.32 2.51
Experience
Travelling
Game/social
Gardening
2.90 4.52 2.95 3.63 SSE/segment 529.2 685.5
3.54 2.96 2.02 3.00
4.27 2.42 1.77 2.98
3.08 3.16 1.52 2.83
SSE total 1445.7
231.1
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 3 Output for four clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 Segment 4 AVERAGE Respondents Segment 1 Segment 2 Segment 3 Segment 4 Total
2.31 3.96 3.73 2.25 3.13 Number 94 89 73 44 300
Education 2.48 2.61 2.60 2.25 2.51 % 31.3 29.7 24.3 14.7 100.0
Experience
Travelling
Game/social
Gardening
2.90 4.52 4.36 2.20 3.63 SSE/segment 410.2 393.2 231.3 204.6
3.49 3.22 2.37 2.52 3.00
4.39 2.74 1.86 2.27 2.98
3.27 3.89 1.64 1.70 2.83
SSE total 1239.3
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Table 4 Output for five clusters Mean/centroid
Age
Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 AVERAGE Respondents Segment 1 Segment 2
2.09 3.09 4.05 3.75 2.09 3.13 Number 65 66
Education
% 21.7 22.0
Segment 3 Segment 4 Segment 5 Total
76 59 34 300
25.3 19.7 11.3 100.0
2.40 2.76 2.54 2.59 2.06 2.51
Experience
Travelling
Game/social
Gardening
2.60 3.74 4.55 4.39 2.03 3.63 SSE/segment 274.2 222.0
3.40 3.70 3.07 2.17 2.15 3.00
4.54 3.52 2.66 1.58 2.09 2.98
3.66 2.15 4.07 1.66 1.79 2.83
SSE total 1118.1
330.0 134.5 134.5
Respondents mean people in which cluster answered questionnaire; Number of people who answered questionnaire in each cluster; % mean percentage of people who answered questionnaire in each cluster; SSE/segment means Sum of Squared Error per Segment; SSE total means Total of SSE/segment.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
6
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
and 3, respectively. The last analysis confirmed the author's decision as the means per segment in Table 4 had better variation than those of Table 3; therefore, this analysis showed that there were five clusters among the sampled alumni. In conclusion, Table 4 shows five clusters with an appropriate distribution of segments (21.7% in segment 1, 22.0% in segment 2, 25.3% in segment 3, 19.7% in segment 4, and 11.3% in segment 5). The five segments also produced the lowest SSE and had the right variation of k-mean in each segment. Alumni Segmentation Five segments were derived from the research. The characteristics of each segment are summarized below and in Figures 4e10. Segment 1 contained mixed gender alumni aged less than 30 years, enjoying their single status and currently pursuing a further degree while still working for their first employer and being paid less than THB 30,000 per month. The members of this segment work in computer technology with less than 3 years' working experience. The author named this segment “Activist Mainstreamer”. Segment 2 contained mixed gender alumni aged between 31 and 40 years, enjoying their single status and most of them undertaking their Master degree with their salary being THB 50,000e200,000 per month. The members of this segment had worked in many fields for more than 10 years but less than 19 years. The author named this segment “Passionate Learner”. Segment 3 contained female alumni aged more than 41 years, enjoying their married status and currently working at a management level for a government entity. This
segment consisted of Bachelor and Master degree holders with salary range of THB 50,000e200,000 per month, work experience of more than 20 years, and the majority were nurses. The author named this segment “Mid Age Religious”. Segment 4 contained male alumni aged more than 41 years, enjoying married life. The members of this segment held either a Bachelor or Master degree, had been running their business for more than 20 years and had earnings of more than THB 200,000 per month. The author named this segment “Senior League”. Segment 5 contained female alumni aged less than 30 years, enjoying their single status and currently pursuing a further degree while still working for their first employer and being paid less than THB 30,000 per month. The members of this segment worked in a variety of fields and had less than 3 years' working experience. The author named this segment “Elaborate Cohort”. The Mid Age Religious segment had the highest number of alumni (25%) followed by the Passionate Learner and Activist Mainstreamer (22%) segments. The least alumni were in the Elaborate Cohort segment which accounted for only 11 percent, while the Senior League segment contained the remaining members. Other research has reported that each segment has different interests and preferences (Ahmadi et al., 2013). The Activist Mainstreamer segment loves to associate with social media like Facebook, Line, and Instagram. The members of this segment are active in many key activities including: exercise, walking, running, cycling, football, basketball, volleyball, tennis, badminton, table tennis, watching sport, traveling, ecotourism, temple tourism, dinner parties, cinema, concert, theater, music, karaoke,
Figure 4 Alumni demographics segment highlight
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
7
Figure 5 Five alumni segment highlight
photography, painting, social games, playing music, gardening, and household duties. The members are interested in seeking information on hot news, travel news, business and marketing news, entertainment news, football news, lifestyle news, food and beverage news, and job news. They are also willing to support universities with new graduate recruitment,
special lectures, and/or sport participation. They also expect universities to offer sports cantering service, free parking, venue rental, job search service, and innovation hub. The members of the Activist Mainstreamer segment are very active people seeking knowledge and they love to socialize with others. The author would recommend adopting the members in this segment as the first priority.
Figure 6 Activist mainstreamer segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
8
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
Figure 7 Interests of passionate learner segment
The Passionate Learner segment has similar characteristics to the first segment. However, the members of this segment are less active due to their older age. Activities which they like are: exercise, walking, running, cycling,
adventure, watching sport, traveling, ecotourism, temple tourism, dinner parties, and social games. The members of this segment are seeking information on business and marketing news as well as news related to
Figure 8 Mid age religious segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
9
Figure 9 Senior league segment
economics. They expect the university to offer a consultancy service. The author would recommend universities adopt these members. The Mid Age Religious segment consists of members who mostly are house wives. The key activities are related to the
family such as traveling, dinner parties, cooking, gardening, and charity work. They do not expect any support or information from universities but prefer to have reunions so that they can meet old friends and professors. The author thinks that the members of this segment will be interested in
Figure 10 Elaborate cohort segment
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
10
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11
associating with the university's professors and other alumni members even though it is in a passive mode. The Senior League segment consists of elder members who once-in-a-while utilize social media. However, they are not interested in any activities and do not expect anything from universities. The author feels that it is very hard to get these members to associate with universities mainly because they have limited free time since most of them are business owners and family heads. However they are good for donations to charity (Clotfelter, 1985). The Elaborate Cohort is another segment with members who do not want to participate in any activities and do not expect anything from universities. Based on their demographics (gender, age, marital status, position, income, experience), they are similar to the Activist Mainstreamer segment. However the educations level of this segment contains more at the Bachelor level compared to the Activist Mainstream segment who had graduated with both Bachelor and Master degrees. Alumni in the Activist Mainstreamer segment mostly work in the computer and technology field. Moreover the Elaborate Cohort segment has less interest in playing on social media compared to other segments. The author considers that universities should give these members a very low priority regarding their current association with alumni activities or on the other hand, universities should pay attention more to engage this segment back.
alumni more deeply by clustering the customers or alumni into segments based on demographics and alumni behavior. The results of alumni clustering can help a university to communicate with its alumni using appropriate information and the proper channels to get a high response rate. In conclusion, each segment has different needs and wants. The research has clearly answered the questions mentioned during introductory part. Clustering alumni into segments and understanding their characteristics/behavior are important factors in obtaining a successful outcome. However, the author will conduct further studies to complete the objectives including developing a collaboration model and adoption strategies for each segment. Conflict of interest The authors declare that there is no conflict of interest regarding the publication of this paper. Acknowledgments This research was undertaken using a grant by the Graduate School of Chulalongkorn University, Thailand under the program “the 90th Anniversary of Chulalongkorn University Fund” (Ratchadaphiseksomphot Endowment Fund).
Discussion This research aimed to classified alumni with the same preferences into segments. The methodology designed the instruments to extract the alumni's demographics, lifestyles, expectations, and interests. The researcher expected that the designed instruments could collect data and enumerate the data using the chosen data mining technique. The results from simple k-mean clustering provided the alumni's preferences in segments and was consistent with the concept of the research. For instance the majority of the Activist Mainstreamer cluster consisted of singles under 30 years old. They are interested in a variety of activities such as exercise, running, cycling, football, and tennis and expected the university to provide them with a sport center service. Most of them were a first jobber and they were interested in recruiting graduates activities by the alumni association and expected a job search service as well. They used Facebook and LINE to communicate. The choice of answers in the instruments highlighted the Mid Age Religious cluster. Alumni in this cluster or segmentation were mainly females aged more than 41 years. The lifestyles choices they chose are in the survey were temple tourism, chapel visiting, being involved in charities, and praying. They were not interested in any activities proposed by the university. Moreover, they expected nothing from the university. These results were able to answer the research questions. Conclusion and Future Research Alumni clustering is one methodology that a company or university can use to understand their customers or
References Ahmadi, H., Nilashi, M., Ibrahim, O., Rad, M. S., & Pourhashemi, S. O. S. (2013). Customer relationship management model in higher education: A case of Malaysia. Today Science Journal of Humanity, 1(1), 1e10. Retrieved from http://www.tsjournals.com/Article/2013/1/Customer% 20Relationship%20Management%20Model%20in%20Higher% 20Education%20A%20Case%20of%20Malaysia.pdf. Clotfelter, C. T. (1985). Federal tax policy and charitable giving. Chicago, IL: University of Chicago Press. Retrieved from http://papers.nber.org/ books/clot85-1. Durango-Cohen, E. J., & Balasubramanian, S. K. (2014). Effective segmentation of university alumni: Mining contribution data with finitemixture models. Research in Higher Education, 56, 78e104. Retrieved from http://www.scopus.com/inward/record.url?eid¼2-s2.0-84901580676&partnerID¼40&md5¼b271fa0bfc55ea84b325b4fab5f01157. Han, J., Kamber, M., & Pei, J. (2012). Data mining concepts and techniques (3rd ed.). Burnaby, BC, Canada: Morgan Kaufmann Publishers. Retrieved from https://web.engr.illinois.edu/~hanj/bk2/toc.pdf. Hanson, S. K. (2000). Alumni characteristics that predict promoting and donating to alma mater: Implications for alumni relations (Unpublished doctoral dissertation). Grand Forks, ND: University of North Dakota. Larsen, N. (2010). Market segmentation e A framework for determining the right target customer (Bachelor’s thesis). Aarhus: Aarhus University. Retrieved from http://pure.au.dk/portal/files/11462/BA.pdf. Le Blance, L. A., & Rucks, C. T. (2009). Data mining of university philanthropic giving: Cluster-discriminant analysis and Pareto effects. International Journal of Educational Advancement, 9(2), 64e82. Retrieved from https://www.researchgate.net/profile/Louis_Le_Blanc/publication/ 263326511_Data_mining_of_university_philanthropic_giving_Clusterdiscriminant_analysisand_Pareto_effects/links/56ab70d208aed5a0135 c1598.pdf?origin¼publication_list. Masterson, K. (2012). 5 Colleges that inspire alumni giving, and how they do it. Chronicle of Higher Education, 1, 1e4. Retrieved from http://www. fulbright.fi/sites/default/files/Liitetiedostot/Palvelut-asiantuntijoille/ Opintomatka- alumni/recent-news-articles-chronicle-alumni-relationsand-institutional-giving.pdf. Prasanna, K. V., & Peddyreddy, P. B. (2015). Mining customer transaction data using clustering. International Journal of Electronics and Computer System, 3(11), 41e44. Retrieved from www.irdindia.in/journal_ijeecs/ pdf/vol3_iss11/10.pdf.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002
N. Rattanamethawong et al. / Kasetsart Journal of Social Sciences xxx (2017) 1e11 Radcliffe S.K., A study of alumni engagement and its relationship to giving behaviors, (Master's thesis). Bucknell University, Lewisburg, PA. Retrieved from http://digitalcommons.bucknell.edu/cgi/viewcontent. cgi?article¼1001&context¼masters_theses Rajagopal, S. (2011). Customer data clustering using data mining technique. International Journal Database Management System, 3(4),
11
1e11. Retrieved from http://airccse.org/journal/ijdms/papers/ 3411ijdms01.pdf. Rattanamethawong, V., Sinthupinyo, S., & Chandrachai, A. (2015). An innovation systems that can quickly responses to the needs of students and alumni. Procedia e Social and Behavioral Sciences, 182, 645 e652. Retrieved from www.sciencedirect.com.
Please cite this article in press as: Rattanamethawong, N., et al., An innovation model of alumni relationship management: Alumni segmentation analysis, Kasetsart Journal of Social Sciences (2017), http://dx.doi.org/10.1016/j.kjss.2017.02.002