Nuts & Bolts - Case Study: AAA South Makes, Models Analytics Efficiency
Challenge: Improve analytics efficiency.
Solution: Build a table for the analytics team to use for modeling and load that data into a self-service analytics tool for non-analysts to generate their own counts.
Results: The analytics team can build more sophisticated models three times faster and, because the tools are user-friendly, non-analysts can perform tasks with the self-service tool.
Fantasyland is a great place for its members to go, but AAA Auto Club South's analytics team wants to ensure that its data—enabling, for example, insight into customers' propensity to buy the discounted Walt Disney World tickets it sells—reflects reality.
That's why in June 2009, the Tampa, Fla.-based affiliate of AAA chose to improve its analytics capability and efficiency by hiring Portrait Software, which is now part of Pitney Bowes Business Insight.
Kristin Rahn, the affiliate's advanced analytics director, remembers how she explained the problem to the vendor: "I have seven people on my team. And three of them are trained statisticians and the other ones are not, but they're good analysts.
"So my challenge is that I have seven people that need to deliver complex analytics results and models," continues Rahn, "but they're not all trained to do that. And I need to do that across a wide range of businesses: including membership, insurance, travel and financial services. And I can't be a bottleneck to the organization, in terms of [business users] understanding how their business works and implementing programs to improve business."
The affiliate first built an analytics table with 6.9 million records and about 500 columns of household-level aggregate transactional data. Then, by the beginning of 2010, Rahn's team could send requested data to non-analysts within AAA Auto Club South so they could perform some of the simpler tasks themselves.
As of presstime, Rahn says her team already has 20 predictive models built, half of which are actively deployed. Because of the goals set for the models—to increase renewals, reduce churn and cross-sell and upsell members—Rahn says the affiliate waits a full year to judge results, so sales figures aren't available.
However, the models are already yielding insights. For instance, they show that members from Georgia and Tennessee travel to Disney during the summer, but Floridians go see Mickey year-round.
Rahn also says the affiliate has already gained efficiency. The 10 analysts trained on the analytics table can build predictive models in a third of the time it once took. Mostly, that's because the tool presents a user-friendly face to the modeler and does most of the complex work behind the scenes. Before, Rahn says analysts had to pull together data from more than 200 tables—instead of this new, single one—for each project. Now analysts still have the option of handling the details themselves, but only if they choose to do so.
Also, the analytics team isn't a bottleneck. Rahn says 50 non- analysts, from the affiliate's president to the marketing strategists, can already answer questions on their own when the analytics team sends data to them—in minutes, instead of the previous two-week turnaround. She also says another 100 executives, from regional office managers to marketing professionals, will soon have the self-service tool.
"It's like having another person on my team," Rahn says. "The time-savings has been a huge benefit to us ... We can produce more and it's all of better quality than what we used to produce."