Skip to main content
LinkedIn Live | Oct. 22 @ 1PM ET | Paving a clear path: A guide to the future of tolling Register now

The Mathematics of Profitable Marketing

Cabela’s uses predictive analytics to formulate a customer interaction strategy that maximizes the spending of each individual customer.

The Mathematics of Profitable Marketing

In the book Moneyball: The Art of Winning an Unfair Game, which chronicles the Oakland Athletics’ analytical approach to assembling a competitive baseball team, author Michael Lewis illustrates in a few words why data overload can be disastrous. He quotes baseball writer, historian, and statistician Bill James: “I wonder if we haven’t become so numbed by all these numbers that we are no longer capable of truly assimilating any knowledge which might result from them.”

This quote still resonates today because it represents the state of most enterprises: having the ability to collect rafts of customer data across myriad touchpoints, only to be paralyzed by the vastness of that data and unable to forecast how customers will behave.

Using predictive analytics to mine their data, companies can determine the best action for a given situation, find patterns that can guide decision making, and uncover actionable predictions for each customer. These predictions can encompass all channels—both online and offline—foreseeing which customers will buy, click, respond, convert, or cancel. When considering thousands of factors, and a universe of millions of customers, predictive analytics connects the dots, guiding each decision to greater success.

Retailer Cabela’s, a U.S.-based outdoor outfitter and the largest direct marketer of hunting, fishing, camping, and related outdoor merchandise in the world, leverages predictive analytics to bring rules and reason to its marketing by analyzing individual customer information to determine responsiveness to catalogs and other mailings, and to select appropriate product mixes for each. Cabela’s strategy reveals a world of opportunity that advanced analytics can deliver.

A data-driven organization
Many consider Cabela’s to be one of the great American success stories. Founded in 1961 around their kitchen table, husband and wife Dick and Mary Cabela started crafting direct mail to advertise fly-fishing products. Over the years the company grew from that kitchen office into what many call the “Disneyland” for outdoor enthusiasts, serving more than 60 million customers. The company now operates 30 retail stores and produces 100 different types of catalogs, mailing a total of 120 million every year to customers in 120 countries. This complexity makes predictive analytics essential to ensuring the accuracy of its mailing lists and for targeting various customer groups.

Corey Bergstrom, director of market research and analyses, calls analytics the core of Cabela’s business. As a result, he says, the organization has become data driven, with departments ranging from merchandising to customer service relying on data and analytics to make daily business decisions (see sidebar, pg. 20). “It’s something we’ve been doing for a long time,” Bergstrom says. “Especially in this [economic] environment, having that data allows [marketing] to know when they should make a decision.”

Before Cabela’s became the data-driven organization that it is today, it pulled its customer information from about five sources; over time the data sources increased in number.

The divergence of data turned the process of mailing catalogs into a time-consuming ordeal. Gathering reports and analyzing customer data lasted weeks and marketers had limited access to the information. “Our biggest challenge was…the people performing the marketing selection did not have access to the same information as the people building the models,” Bergstrom says. “As a result it became a bottleneck.”

Realizing the gravity of disparate data on its mailing efforts, Cabela’s started integrating its data sources four years ago. The company dismantled data silos, implemented Teradata’s data warehouse, and applied SAS analytics to help score customers. Simultaneously, Bergstrom’s team spent a year training the data management group, statisticians, and marketing analysts on how to leverage the integrated data warehouse and analytics solution to solve their business challenges. The integration, solutions deployment, and initial training lasted two years and ultimately enabled Cabela’s to perform analytics functions in seconds rather than weeks, and reduced the need to copy data from one system to another.

Ranking customers by individual value
Today Cabela’s operates in a dynamic query environment, relying on variables, scores, and rankings to determine relevant content to send to individual customers via their preferred channel with the goal of improving response rates, buying behavior, customer retention, and overall profitability. In short, Cabela’s applies advanced analytic methods to help make better business decisions.

Like the Oakland team that made analytical observations that often flew in the face of conventional baseball wisdom, Cabela’s deploys a bold analytical strategy: Every customer receives a unique score. “We model down to a unique customer and a unique customer has his own score and that score represents what our revenue projection would be if we mail him a catalog,” Bergstrom explains.

The process entails the use of linear and logistic regression to ensure that Cabela’s sends the most appropriate customer mailings. Two statisticians on Bergstrom’s team analyze catalog, store, and website purchases weekly and then build models that rank each customer from best to worst based on a combination of his prior purchases and propensity to move to another category.

The statisticians then apply five to 15 predictive variables that work together to determine the customer’s score on a 100-point scale. The score then ranks the customer according to his revenue projection based on if Cabela’s had mailed him a catalog. “We take a concept of media drivers versus execution drivers. We know that a percentage of Internet sales is driven by catalogs. So we try to send the media that drives that customer to make a purchase versus looking at the execution channel,” Bergstrom says.

He explains that a customer’s scores could differ across categories. For instance, a customer may be ranked first in the archery category, but ranked 3.4 millionth in women’s apparel. The analytics team shares the scores with the marketing group. From the combination of rankings, marketing determines the depth and the frequency with which Cabela’s should mail based on the company’s business goals: to drive sales and profit.

The number one ranked customer in archery, for instance, may receive a 100-page catalog, but the best overall ranked customers who move across categories will receive the company’s 1,700-page hard-bound book. The process creates internal cost efficiencies and ultimately helps to enhance the customer experience by personalizing customer interactions. “We know the return of every dollar we put out and we optimize that spend,” Bergstrom says. “If we don’t know, we find out.”

Since Cabela’s began scoring customers, it’s seen a 300 percent increase in catalog performance and a “significant reduction” in catalog production costs. “It’s important to optimizing the marketing spend [per customer] to realize what media we need to send to get them to buy, and it’s important from an operations spend to determine if [customers] will execute it on the Internet or through a catalog so we have the right resources in place,” Bergstrom says.

Analytics as an enterprise function
In addition to optimizing its catalog spending, Cabela’s uses predictive analytics to streamline and target its production of retail flyers, which the company inserts into newspapers to support its retail sales. Leveraging an optimization strategy rather than a blanket insertion strategy, Cabela’s relies on analytics to determine which zip codes within a 120-mile radius of a store respond best to direct mail. Then Cabela’s inserts flyers into the local newspapers in those geographies that will likely generate the most sales. Bergstrom says the company initially saw a 60 percent increase in the performance of the flyers. “We looked at where the sales are coming from and we were able to reduce circulation, but increase sales,” he says.

Cabela’s also recently started to calculate the customers’ value to the company to help the organization determine how to best deliver each customer’s overall customer experience. The company uses a five-star rating system for scoring its high-value customers and applies specific treatment strategies based on each level of stars. Customer service receives training about how to deliver service to each level of stars.

The success from customer scoring has helped to evolve the analytics team from a purely marketing support function to an enterprise support function. Today hundreds of employees, not just statisticians, can access the data warehouse, and Bergstrom continues to teach them how to leverage the data. His team works with the various departments to define a business challenge and then, together, they develop an analytics solution. “We bring the whole area into meeting rooms and show them the tool and how to leverage it to make better decisions,” Bergstrom explains. “They are making great strides in using that information and more of the company has shifted to a data-driven organization.”

Cabela’s also has reallocated staff to work on other projects because now it takes two statiticians to conduct the customer scoring rather than the previous six. “We’ve seen resources become what they were designed to be,” Bergstrom says. “A merchant can focus on a merchandise strategy more, a statistician can focus on building models more, and there’s an outright cost savings.”

Bergstrom’s team soon will begin analyzing the click-stream patterns of customers shopping online, with the goal of presenting customers with the most appropriate offers. The team is also supporting the company in its retail expansion efforts by helping to select locations with the highest growth potential based on predictive data modeling of multi-factor trade areas. Bergstrom relies on the models to help with planning the sizing of the outlets, as well. He says, “We are very much aligned with our strategic initiatives in our company and, in fact, we will be helping with each and every one of them.”