THE ASIA-PACIFIC CENTRE FOR RESEARCH, INC.

home > articles >

Data Mining Applications in Higher Education
Source: www.spss.com
Copyright SPSS, Inc. 2004

Part 3 of 3

Case study three: predicting alumni pledges

Challenge

What are the most cost-effective and efficient means for universities to solicit alumni pledges? For a typical urban university of 25,000 the alumni population can be as many as ten times its enrollment. Though the university may never hear from most of them, most universities
continue to send out regular mailings. These mailings typically cost over $100,000 a year. This case study shows how data mining helps universities develop a cost-effective method targeting the alumni most likely to make pledges.

Solution

Often it is difficult to determine if the mailings directly affect the volume and sum of alumni pledges. Data mining helps to resolve this quandary. In other words, given the same type of mailing, Alumna Mary may contribute regularly or she may not contribute at all. Or, what if
the reverse is true: those who receive mailings are not the ones who contribute? Another phenomenon often observed in conducting data mining in this type of setting is the presence of outliers. There are few alumni who contribute in large sums and they show up unexpectedly, such as Gordon Moore donating $500 million to his alma mater. How do we
identify them and cultivate a relationship with them? This is an issue similar to dealing with outliers, who often arouse intense interest.

In a hypothetical mass mailing for alumni pledges, the dotted line is the optimal return rate (alumni sending in contributions) as predicted by data mining. The solid 45-degree line is the result if the entire population received the mailing. If data mining is used, when the mailing reaches the 30th percentile of the population predicted to be “responsive,” 80 percent would respond with a pledge. Without data mining, only 40 percent would respond. If every percentage point = $2,500, savings = (70% * $2,500) – (30% * $2,500) = $175,000 - $75,000 = $100,000). Without data mining, it would cost $100,000 more to reach all 80 percent.

Results

Data mining serves as a cost-effective means of increasing alumni pledges. There are typically more people who contribute in small amounts, which is also of interest to data miners. Data mining may reveal a particular pattern associated with them. Moreover, data mining enables universities to make their mailings more effective. This is best described using the concept called “lift.” If 20 percent of alumni respond to a pledge request, it behooves the university to concentrate on those 20 percent. If data mining can quickly identify them by a ratio of two to four (correctly predicting two out of four who will donate), then the university only needs to mail out 40 percent, thus saving money by not mailing the other 60 percent. Better yet, the university may use the money saved to recruit additional and/or larger donations.

Conclusion

Thanks to data mining, educational institutions have a powerful analytical tool that enables them to better allocate their resources and staff to manage student and alumni relations. With the ability to uncover hidden patterns in large databases, community colleges and universities can build models that predict — with a high degree of accuracy — the behavior of population clusters. By acting on these predictive models, educational institutions can now effectively address thorny issues from learning outcomes, transfers and retention, to marketing and alumni relations.