Keep CRM In Stock With Data Mining

When the proprietors of The Vermont Country Store (Manchester Center, VT) mailed the company's first catalog in 1945, its customer relationship strategy was simple: the mailing list included every name on the family Christmas card list. More than 50 years later, the company mails 50 million catalogs full of the "practical, functional, and hard to find" each year.

Early on, the family learned of the delicate balance between healthy growth and overselling. "We learned how to control the pace at which we grew with a simple equation," explains Lyman Orton, the company's proprietor. "We needed to expand our capacity in the stockroom and shipping department, and then send out enough extra catalogs to increase the orders, but not so many that we could not ship today's orders tomorrow."

To help control its mailings and more accurately predict customer needs and behavior, the company purchased Enterprise Miner from SAS. "Enterprise Miner is a big part of why we've been able to consistently find better ways to mail to our customers," says Larry Shaw, The Vermont Country Store's vice president of marketing and creative. "When you mail 50 million catalogs a year, if you can improve results by 1 to 3 percent, that's a big increase."

Before turning to SAS, the store's mailings were based solely on customers' buying histories (how recently and how frequently they purchased and how much money they spent). But while this RFM (recency-frequency-monetary) approach is a fairly reliable method, models are limited only to those three variables. "RFM targets always tell you to send catalogs to the people who have made recent purchases and who have spent the most money," explains Shaw. But that's not always the best strategy. Shaw and his team wanted to include other variables, to make sure they were targeting the right customers with the right products and not over-saturating the market by sending too many catalogs to the same customers.

Since they began using Enterprise Miner in 1998, market researchers at The Vermont Country Store have expanded their modeling methods to incorporate new variables, including the number of catalogs a customer receives, the buying patterns for each season, and the products that are popular with each customer segment.

"With Enterprise Miner, we can actually take raw data and create new variables, which tell us the biggest drivers for our customers. We can look beyond which month purchases were made, or how much money they spent, and start looking at which customer is more likely to spend their money in one quarter over another." With this information, marketers can determine the ideal times to send new catalogs to each customer. The software allows the company to bring in outside data - including credit and demographic data and lifestyle codes - to aid in predictive modeling.

Reactivate Customers
The Vermont Country Store maintains a customer database with more than 4 million names and addresses. About 20 percent are active customers who have recently bought items from the store's catalogs. The remaining dormant customers may have gone 5 to 10 years without making a purchase. "In the past, we wouldn't mail catalogs to these customers," explains Shaw. "But with predictive modeling, we've been able to reactivate customers who haven't ordered from us in years."

The company mails small versions of its catalogs to dormant customers. "We rank customers in terms of productivity, then mail large catalogs to better customers and small versions to prospects," says Shaw.

Maximize Mailing Efficiency
Today, the balance between healthy growth and overselling at The Vermont Country Store is understood through circulation efficiency. "Circulation efficiency is probably the biggest driver of profitability for a catalog company. How can we mail more without compromising the productivity of what we mail?" asks Shaw.

"This year, we've been able to mail more catalogs without lowering our productivity. In fact, our productivity has increased," he says. "Enterprise Miner has allowed us to find new customers that we normally wouldn't have mailed to, and it helps us make sure we're not over-mailing our existing customers."

The Vermont Country Store also has a Web catalog, the performance of which, Shaw points out, is affected by the company's physical catalog mailings. "When we mail a physical catalog, online sales go up," he says. "So, we started a 'quick shop' feature that leads catalog recipients to our Web page to buy." In the next six months the company will use the SAS IntelliVisor ASP (application service provider) e-commerce reporting program to build models and algorithms on the Web site. This will allow automated up- and cross-sells to occur based on clickstream data gathered from customers as they shop. IntelliVisor gives the company the ability to track and pattern its online customers in even more detail than its traditional catalog customers, recording most visited pages, sales conversion rates, search engine performance, and effectiveness of search terms, for example. With a detailed view of both cyber and traditional shoppers, Shaw is confident his company will continue to improve and tailor its direct-to-consumer marketing approach. "If you want to build a strong, profitable catalog or Web business, you have to have detailed analytic programs," he says. "It's not a choice."