Analytic Engineer Has a Passion for Solving Puzzles
This column is written for statisticians with master’s degrees and highlights areas of employment that will benefit statisticians at the master’s level. Comments and suggestions should be sent to Keith Crank, the ASA’s research and graduate education manager, at email@example.com.
Justin Rowland came out of the hospitality industry to pursue a career in analytics. He is an analytic engineer at the Advanced Analytics Lab at SAS, working on applications of social network analysis, specifically fraud detection and customer retention.
Across all industries, organizations are generating and collecting massive amounts of data to improve operations and make better decisions. Some companies already have their own analytical teams; others are just beginning to realize the value of investing in analytic capabilities. Regardless of how an organization uses its collected data, there remains a tremendous benefit in having external data specialists provide a fresh perspective. In 2007, SAS CEO Jim Goodnight created the Advanced Analytics Lab (AAL) to further assist organizations make the most of their data, leveraging cutting-edge SAS tools and a team of individuals who live and breathe data (like me).
Here at AAL, we work on a variety of interesting problems. Some colleagues are forecasting how much a cable company stands to gain or lose from lowering its rates. Others have analyzed factors that cause inmates to return to jail after release. Perhaps the most common problem we are working on is fraud detection. Many banks, insurance companies, health-care, and welfare organizations have fraud detection systems in place, but they have no methods for linking individuals to known fraudulent activity.
We use SAS Social Network Analysis to connect the dots between individuals, revealing fraud rings that once went undetected. However, the software can be applied to any data set in which there are connections between different entities. For one project, we are using SNA to identify influential customers for a telecommunications company.
A typical project begins with the client providing data and a set of objectives. The client usually provides data in the same form that it is captured and stored—in multiple files from a relational database. The first task is exploratory data analysis, in which the analyst gains an understanding of the variables and discovers how the tables relate to each other. Examining the quality of the data is also necessary (for example, determining how many missing values each variable has).
The next step is to combine the multiple files into as few files as possible. For fraud projects, this usually means having one data set for transactions, one for accounts, and one for individuals. Once the tables have been cleaned and combined, the analytical work can begin. (A general rule of thumb is that about 85% of a project is preparatory work.) The analytical work involved varies from project to project, depending mostly on the objectives provided by the client. Some examples include predictive modeling, clustering, survival analysis, and time-series forecasting.
For students with an enthusiasm for math and statistics, working with data would be a great career choice. Job openings are expected to continue to rise, salaries are substantial, and the work is rewarding—it’s like working on complicated puzzles with a team of smart people. I especially like that I am able to use my creativity and problemsolving skills every day.
My advice to students interested in this line of work is to attend a program that not only teaches theory but also offers a chance for practical experience. I was fortunate enough to have attended two of these programs, receiving a bachelor’s in discovery informatics from the College of Charleston and an MS in analytics from North Carolina State University. Both programs allowed me to apply the analytical techniques that I learned in the classroom to real-world projects, which greatly helped prepare me for my career.
Lastly, I recommend that students either attend a program that teaches SAS programming or learn it on their own. There are many tools available for working with data, but none that I have tried come close to offering the power and control that SAS does—and no, I’m not required to say that.