Williams-Sonoma Understands Teens as Well as Sophisticates

Team of Three Analysts Builds More than 100 Models per Year

Williams-Sonoma Inc., retailer of fine home furnishings for every room in the home, has introduced a hot new catalog – Pottery Barn Teen – which offers everything from fringed lamps and pillows to chic furniture and accessories. But how did this sophisticated retailer decide how to market its new brand to the trendy teenage audience? With the help of SAS.

Using SAS Enterprise Miner, senior modeling analysts Sumedha Dolan and Marianna Dizik and modeling analyst Sam Pederson easily combined demographic data on the teenage market with traditional Pottery Barn sales data to create an original mailing strategy for the Pottery Barn Teen catalog. The new data model, similar to those used for Williams-Sonoma, Pottery Barn and Hold Everything catalogs, segments customers into groups and details the potential profitability of each segment. The results help marketing executives decide which customers should receive each new catalog.

The data modeling process has become second nature at Williams-Sonoma, where this team of three modeling analysts builds more than 100 models every year for the company's multiple catalog mailing campaigns. They say the advanced data mining solution from SAS expands their analytic capabilities and streamlines the modeling process. In particular, Dolan says Enterprise Miner provides:

• A broad set of data mining techniques for modeling and analysis.
• Advanced visualization tools for examining data and comparing results.
• A complete generation of scoring code for all stages of model development.

The best part, say the analysts, is that now they have more time to explore and analyze their data with advanced modeling techniques. "The quality of our models depends not only on the data we use but also on how we use that data," explains Dizik. "SAS really helps us make the most of the company's data, so it can have the largest impact for the model build."

Straightforward, Repeatable
According to Dolan, data mining has two main components: building the model and applying that model to the data. She says SAS helps on both fronts. "Enterprise Miner is used when building the model and when applying the models to different data sets. Enterprise Miner really makes a difference when we're working with new initiatives."

Using an active database of more than 33 million households, the Williams-Sonoma modeling team begins each new model build by selecting data from the previous year's mailing results. Enterprise Miner determines the relative importance of each variable and allows the analysts to segment similar customers into groups that include 30,000 to 50,000 households. The model predicts profitability for each group, based primarily on the previous year's total average purchases per household.

Throughout the entire process, the simple point-and-click interface of SAS Enterprise Miner tracks and records their work on every model. Before using Enterprise Miner, Pederson says he used to keep Excel spreadsheets with complex rows of notes to remind him about each step of the data flow.

"But now all the information we need is in the flowchart. Everything is set up as projects, so we can easily pull up the template for a particular model build and determine if we want to make changes. It's all right there for us."

Dizik says SAS also helps improve project continuity by allowing her to view the data flow and the process flow simultaneously. "It gives us the flexibility to check our work, to see the results and to be consistent from year to year," she says. "We can have the process standardized but still make changes easily."

Modeling Improvements, Advanced Techniques
By standardizing the data mining process, Enterprise Miner not only reduces errors and improves learning curves; it also opens the door to more complicated analyses.

"Enterprise Miner gives us a lot of flexibility to try new modeling approaches," says Dizik. Before using Enterprise Miner, she and her coworkers primarily used regression for every model, but Enterprise Miner has made it easy to try more advanced techniques such as logistic regression, two-stage modeling and neural networks.

"It's a major strength too," adds Dolan, "Because we can spend time working on new techniques rather than trying to rebuild the same things we did before."

For instance, Dizik recently tried a bagging and boosting technique which allowed her to improve and stabilize a model that analyzes inactive customers. By nature, retail companies have less information available about their inactive clients, so they are traditionally one of the toughest segments to model. With Enterprise Miner, however, Dizik was able to smooth the projections and come up with a more accurate prediction of which customers to target.

Analysts and managers at Williams-Sonoma also look forward to a new series of modeling developments with SAS. In addition to introducing new brands like Pottery Barn Teen and the successful West Elm catalog, the company also plans to enhance their models with data from third party sources – using information such as demographic and lifestyle data to more cost-effectively tailor the right catalogs and messages to the right customers.

Other exciting projects that have been initiated with SAS include multichannel modeling efforts and customer lifetime value initiatives. The multichannel modeling effort uses data from Williams-Sonoma Web sites and Williams-Sonoma retail stores to understand the way customers shop all three channels. Likewise, measuring the lifetime value of its customers will help the company establish more long-term goals for its catalogs.

"The catalogs are the backbone of our advertising strategy," explains Pederson. "We're always looking for ways to redistribute them and use them in more advantageous ways." As they look to the future, Williams-Sonoma analysts and managers know SAS will help them understand the many ways their catalog mailings can contribute to long-term and short-term profitability within every channel.

The results illustrated in this article are specific to the particular situations, business models, data input, and computing environments described herein. Each SAS customer’s experience is unique based on business and technical variables and all statements must be considered non-typical. Actual savings, results, and performance characteristics will vary depending on individual customer configurations and conditions. SAS does not guarantee or represent that every customer will achieve similar results. The only warranties for SAS products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with SAS as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of SAS software. Brand and product names are trademarks of their respective companies.

Copyright © SAS Institute Inc. All Rights Reserved.