Case Study: Cabela's
Case Study: Cabela's
The Buck Starts Here
Can a niche market leader benefit from advanced analytics? Cabela’s says absolutely.
Ever since Gordon Moore predicted in 1965 that computing power would double approximately every two years, Moore’s law has essentially held true. But as anyone in the IT industry can attest, data growth easily outstrips Moore’s law—some say by a factor of 10—thanks to emerging data sources such as social media, mobile applications and e-commerce. For companies like Cabela’s, that’s a good thing.
Big data analytics promises to mine vast data sets to uncover hidden business insights. Currently the domain of behemoths like Google and Yahoo, big data analytics is capturing the attention of smaller-scale, niche market leaders like Cabela’s, one of the world’s top retailers for hunting, fishing and outdoor gear.
Teradata Magazine spoke with Cabela’s Dean Wynkoop, director of data science, and Craig Bruner, data warehouse senior architect, about how the company is adding big data analytics to its analytical ecosystem to better understand the interrelationships among its key marketing channels.
Cabela’s has an omni-channel marketing department. What is the ‘omni’ approach?
Wynkoop: The idea of “get a catalog and call an 800 number” is embedded in people’s minds when it comes to Cabela’s. But we know that catalogs influence purchase behavior in brick-and-mortar stores, online and through the traditional call center. We also know that digital marketing influences purchases in all three channels. Omni marketing is about developing a holistic view of media options and understanding how they drive sales agnostic of the channel.
Have new data sources affected this approach?
Wynkoop: They provide several challenges. One is the sheer volume of available data that has to be sorted. You also need architecture in place to load that data and pull out insights efficiently and quickly. Another major challenge is cross-referencing transactional data with some of the semi-structured data that’s coming in. For example, how do you tie a banner ad to a brick-and-mortar purchase?
How are you overcoming these challenges?
Wynkoop: We’ve defined a central ID that ties all of the data to a given household and/or an individual. Then we’ve taken steps to ensure that, as unobtrusively as possible, we can make those identifications throughout the system.
Bruner: We have to apply that identity to each of our channels in a consistent and cohesive manner. This way, when we’re looking at everything from clickstream data to the retail ads in newspapers or catalogs, we can cohesively tie the information together to one household.
Wynkoop: That has its own set of challenges because we’re dealing with multiple applications that, out of the box, don’t support this type of functionality. We have to augment the applications to make those connections and then move everything into a single analytical ecosystem so we can look at the data. We’re planning to expand our capabilities into more digital opportunities, which will require us to develop some type of big data environment that goes hand-in-hand with our Teradata environment.
How far along are you in this process?
Bruner: Currently, we’re using clickstream data to match the experience that people are having on the website with their experiences across other channels. We can see if someone from the website purchases through the call center, or if they browse online and then go to a brick-and-mortar store to buy. We can tie everything together to identify which marketing touches influenced a purchase.
Wynkoop: I would characterize what we’re doing as reporting and analytics. We’re not currently driving actions to the execution side, although that’s in the plans. But the big breakthrough has been understanding all these connections and being able to measure it. Real decisions are being made based on this information. It’s more than just a report.
How is this process driving business?
Wynkoop: We receive regular reports from our business partners that define the results of ongoing digital programs. Through deeper analysis, we’ve been able to augment those results. When you start making those connections across channels, you can describe the value in a more realistic way. Sometimes you discover that the real drivers for a program’s success aren’t what you thought. It’s pretty exciting to capture those insights and apply them to future programs.
That’s a big change from Cabela’s early catalog days.
Wynkoop: In the traditional catalog world, we had statisticians who used regression models to determine who should receive catalogs. What we’re doing now is different, but it’s still an extension of that same mindset of modeling, understanding, testing, revising and iteratively working through these programs.
Bruner: Our biggest insight was figuring out that there are multiple ways to influence customer behavior. We used to tie specific media to certain channels to measure effectiveness. But now we know there’s a complex interrelationship among all advertising that affects specific channels and how customers interact with us across multiple channels.
Wynkoop: For instance, Cabela’s “Deer Nation” is an omni-channel marketing effort that targeted deer hunters during last fall’s hunting season. When we analyzed the results, we looked at the flow of data online—the path that customers took while viewing the page, how that generated sales, what worked and what didn’t. We also looked at the relationship with the retail store and how people came to the site from email and through other sites like YouTube. By overlaying that with our overall performance last year as well as over time, we gained valuable insights that will allow us to take a very good concept and improve it.
Where will you go from here?
Bruner: We already have granular, atomic-level data living in our data warehouse. We need to move to a big data environment to leverage the capabilities, performance gains and return on investment that come with it.
Wynkoop: We’re trying to implement the principles of data science without introducing Java because finding people with those skills is difficult. If we can take the data analyst’s or statistician’s existing skill set and augment it with something like MapReduce or Hive, then we can accomplish the same goals without competing against the Googles and Yahoos of the world for those resources.
Bruner: We see big data analytics as a tool to take advantage of data that we can’t take advantage of today. It will also position us to gain more insight into information that we didn’t even know was there.
Wynkoop: The solution is an analytic environment that will help us with our omni-channel marketing efforts. There are other skills—knowledge, expertise, experience and, dare I say, intuition—at play, and this is just one more tool in the arsenal that will help us get the quantitative knowledge needed for decision making.