Doing IT With Data: Learn.org Speaks with Kevin MacDonell

Kevin MacDonell is an annual giving officer at Dalhousie University in Halifax, Nova Scotia. He's also the author of CoolData, a blog where he shares his experiences teaching himself to use data mining and analysis in higher education. Learn.org caught up with Kevin to learn about his innovative uses of data in higher ed and just what makes the field so cool.

Kevin MacDonell Cool Data

Learn.org: What's your educational and professional background, and how did you develop a passion for data analysis?

Kevin MacDonell: My background seems a little unlikely for numbers work: I hold a degree in journalism and for the first part of my career I worked as a writer and editor, mostly for primary-industry trade magazines.

I've always loved numbers, but I was never proficient in math and have never even taken a statistics course. I think had I taken stats I would have fallen in love with the concepts when I was a lot younger; I'm 41 now.

The beginning of my passion for data analysis resulted from the combination of having a hard problem to solve at work (in higher education fundraising) and reading one excellent book, Data Mining for Fundraisers by Peter Wylie. I recognized that here was a powerful tool I could use that nobody I knew was using.

Learn.org Please describe some of the data mining and modeling tools that you feel are essential for fundraising in higher education.

KM: That's a difficult question to answer because the techniques are not dependent on the tools. There are many excellent statistics and modeling software packages available and it comes down to what type of interface you prefer or are most comfortable with. For example, not everyone agrees, but I don't think Excel is up to the task of doing serious data work.

At minimum you need specialized software. Staff at universities and hospitals probably have free access to SPSS or Minitab, two of the most popular packages. I have both of those but I don't use them, not because I don't like them but because I was trained to do data mining in Data Desk. I like the 'feel' of working in Data Desk, the drag-and-drop nature of it.

It comes down to personal choice, mostly. The missing ingredient is not software, but training: Training in data mining techniques as applied in the software package to which one has access.

Learn.org In what ways can data-driven decision making be applied to the world of university education?

KM: The possibilities are endless, really. I work in fundraising, so that is what I think of first. We have models for propensity to give by phone and by mail, and a future possibility will be online giving. These models allow us to be more targeted in who we call or send mail to. I have also created models for propensity to give at higher levels and to give via Planned Giving (bequests and certain structured, long-term forms of giving).

Expanding beyond fundraising to alumni relations, I have created models for likelihood to attend an event, which is used for slimming down invitation mailings. Other models for different aspects of engagement - such as volunteering - are possible as well.

Outside university advancement, data analysis can also be applied in admissions and recruitment and in student retention, such as the likelihood to accept an offer of admission, the risk of dropping out, etc.

And these examples only consider predictive modeling. Data analysis considered more generally could include forecasting, fundraising campaign planning, alumni engagement scoring, data visualization and relationship mapping - you name it.

Learn.org Do you see a role for some of these models in public education, and if so, what would that look like?

KM: I know a pair of consultants who are working on a project to predict academic success, including which students are most likely to complete successfully and which are most at risk of failure. If predictive modeling could give us an early warning of which students would benefit most from intervention and extra help, think of how important that would be, regardless of whether it's at the high school or university level.

These analysts/consultants do a lot of work with fundraisers, but one of them told me that he feels this work is potentially more significant, and I would have to agree. Anyone working in university advancement is ultimately working to enhance the student's opportunities for learning, after all.

Learn.org You also run a blog, CoolData. Can you tell our readers more about your blog, and how the blogging process has informed your use of data analysis in your work?

KM: Although CoolData might sometimes read like an advice column written by someone claiming to be an expert, the one person who has benefited most is me. The blog has played a big role in my development as a data thinker since I started writing it in 2009. In nearly every post I am writing about something that I am either learning about, working on or gaining a deeper knowledge of. This gives the writing a certain energy, I think, because I am more 'into' something at the time that it is beginning to make sense than I am once I have it figured out - then it's boring and just part of 'background knowledge.'

Doing the work to understand a concept well enough to explain it to someone else has really given my own development a big boost. I don't always fully understand what I'm writing about - I hope I am always honest enough to admit there are gaps in my knowledge and to say 'I'm not sure' or 'I don't know' - and I've been known to change my mind on some issues over time. That's a strength, not a weakness.

The blog has also led to invitations to speak at conferences and to many fruitful conversations with people who are smarter than me.

Learn.org What are some of the coolest or most interesting tools out there for someone who's just starting to get into data analysis?

KM: The software for analysis and building models is getting more sophisticated, and yet easier to use, all the time. I build models using stats software, but there are modeling software packages that can build models for you - just add variables and stir.

I do think it's better that people learn from the ground up so that they know what they are doing - any model will score the constituents in your database, but do those scores mean something, or are they just random noise?

Some of the cooler stuff isn't for stats but for data visualization. I don't do a lot of visualization, but I've always been interested in relationship mapping. (If you've mapped out your linkages in Facebook or LinkedIn, that's relationship mapping.) The free NodeXL plugin for Excel is one cool tool that I encourage people to play with.

Learn.org Many of our readers are students or educators. What advice would you give someone who's interested in bringing these techniques to his or her own institution?

KM: I would start by working with what you've got. Find out if your school has a site license for SPSS or Minitab, download it and get someone to help you learn how to use it. Even if all you've got is Excel, there's a free data analysis plugin you can download and install which will extend Excel's statistical capabilities. Then hit the library and see what you can find there.

The Internet is, of course, a vast trove of resources and advice. It can be very confusing to chart a path through all this stuff, unfortunately. There are dozens of approaches and tools in just predictive modeling alone - data analysis is a big field. For now, I have deliberately chosen to focus just on predictive modeling using multiple linear regression and binary logistic regression. They are not the newest, sexiest tools out there, but I think they are the most powerful.

Someone else might chart a different path, but you need to be focused. The way to get focused is to ask yourself: What is the really big question I have that I need to come up with an answer for? Then research what aspect of data analysis might help you answer that question. My big questions are always about identifying what segments of our alumni population are most likely to engage in desired behaviors such as donating, volunteering and attending events. People working in retail businesses have similar big questions: What segments of the customer base are most likely to be upgraded, to re-subscribe, etc.? For someone else, the big questions might be of a totally different nature, and therefore different tools and techniques will be appropriate.

Learn.org Finally, I'd like to give you the opportunity to share anything you'd like about data analysis and higher education.

KM: Data analysis has been around a long time, and its use in the private sector is long-established. Higher education has been slower to catch up, but I foresee that more and more nonprofit institutions are going to create new, well-paid positions dedicated to using data analysis to serve the entire enterprise. It's my perception that young statisticians graduating today have never had so many interesting opportunities for employment. Working in the private sector has been an option for a long time, but fulfilling work in the not-for-profit sector is going to be an option now, too. That is my hope, at least.

For those of us without formal training in statistics but a passion for data, the time has never been better to find opportunities to learn on the job. That means reading books and playing with one's own data, but more importantly, it means getting hands-on training and finding a mentor to help you choose a path.

To continue researching, browse degree options below for course curriculum, prerequisites and financial aid information. Or, learn more about the subject by reading the related articles below:

The schools in the listing below are not free and may include sponsored content but are popular choices among our users. Tuition and costs will vary across programs and locations. Be sure to always request tuition information before starting a program.

Popular Schools

  1. 1
  2. 2
  3. 3
  4. Next »