Right-Sized Analytics

Rooms To Go finds a data warehouse-driven solution that fits its midsized business

Ever wonder if all the brochures and flyers that spill out of your Sunday newspaper represent a careful, cost-effectively targeted marketing investment? Rooms To Go, a regional furniture retailer based in Seffner, Florida, wondered too.

With USD1.5 billion in annual revenue and 150 showrooms across the Southeast organized into three districts with multiple distribution centers, Rooms To Go is a leading independent furniture company in the United States. However, it has not been able to compete using the kind of information technology that drives marketing campaigns for the industry’s giants. In fact, says CIO Russ Rosen, “We have been outsourcing our targeting efforts and mailing expensive brochures to too many people. Response rates being what they are, we were spending a lot of money on costly brochures with little linkage to results, and it was difficult to steer.”

Rosen wanted to detect and predict the buying patterns of customers beyond simple metrics, such as which furniture pieces and categories are the best sellers by state and customer type. With better insights, Rooms To Go would be able to target repeat customers more effectively in its campaigns.

Sound familiar? Customer analytics fueled by data warehouses are one of the not-so-secret weapons of the world’s largest retailers. The benefits, even if measured in only a point or two of improvement per store, can be massive. And as Rosen discovered, effective analytics are no longer out of reach or years away for small and midsized businesses.

A big lesson learned for us was identifying the need for data we had not been collecting. I spent a few weeks with an IBM modeling consultant mapping the model to the data we actually had, and we figured out what we were missing.
—Russ Rosen, CIO, Rooms To Go

Prebuilt data models offer best practice expertise

Like many others in his position, Rosen quickly discovered that his team lacked both the critical expertise and the requisite time for building its own data warehouse from scratch. For example, each of the company’s three districts is a separate distribution area and has individual data collection systems for both supply and sales. “Before I could do anything, I needed to roll the data into one place,” says Rosen. “But we attempted a data modeling effort and realized that it was over our heads.”

Rooms To Go solicited proposals from consultants; the bids were quite large, and the allotted budget was beginning to look inadequate for the task. The IBM InfoSphere team offered a different solution: InfoSphere Balanced Warehouse, including IBM retail business solution templates. The package included not only the hardware and software for database, analytics, and data warehouse design and administration, but also a pre-developed, ready-to-go retail model.

Such models have proliferated in recent years, as specialists codify years of experience in building systems within specific industries, pointed at specific tasks. Rosen believed that these best practices could shorten his development time, but was unsure whether they would be too different from his current systems to be easily adapted. “When we went over the business solution templates, I knew we were on the right track,” says Rosen. “The model defined which attributes would be needed for market basket analysis and for campaign promotion measurements. We weren’t modeling experts, so this seemed like it would save us a lot of time.”

The prebuilt model also addressed another of Rosen’s concerns. “We knew that the first time we analyzed our data would not be the last,” he says. “Our biggest fear was that we’d solve one problem and then have to rebuild as soon as we identified the next one. The promise of using a model, with the experience that had gone into it, was that we’d be able to build the next piece as a small increment without tearing everything down.”

Rosen initiated the project and quickly came to grips with one of the biggest challenges: mapping real data to business needs, in this case as specified by the data model. “A big lesson learned for us was identifying the need for data we had not been collecting,” he explains. “I spent a few weeks with an IBM modeling consultant mapping the model to the data we actually had, and we figured out what we were missing. We had to build some new tables to collect information that had not been needed to support the original transactional requirements.”

The project took a pause to get all of the necessary data in place. “We actually went back and changed some of the transactional systems,” Rosen says. “As the sponsor of the project, I knew the delay meant much better information.” The changes to the transactional system were additions; because they didn’t alter any of the existing data structures, no line-of-business approval was needed.

Legacy systems put pressure on timelines

Another key lesson, and one that many smaller firms will encounter when deploying data warehouses, was the vital role of the extract, transform, and load (ETL) stage, and its potential cost in money and time. Rooms To Go runs on internally built transactional systems, using a Rocket Software UniVerse database with multi-value fields and distributed files—a specialized architecture that, while robust, is hardly representative of today’s typical data formats. The UniVerse data had to be converted before it would work with the new IBM DB2-based system.

As the project got underway in earnest, the ambitious three- to four-month time frame appeared threatened by the time the conversion might take. It’s not uncommon for this step to cause lengthy delays when legacy formats are involved: the skills to develop the data transfers may be in short supply, and commercial ETL products don’t handle every necessary task. Factoring in the time to convert complex data formats requires some careful assessment, and even well-planned projects can hit snags.

Rosen decided to bring in a consulting firm rather than hiring additional staff; the extra short-term cost was a better investment than adding in-house resources that would not be needed once the project was complete. Fortunately, the IBM solution included a “light” version of InfoSphere DataStage, which proved sufficient—with some tweaking—to move the data.

Once the data issues were resolved, Rosen turned his attention to delivery. “From start to finish, the first project took two months, and once it was done, we delivered our first executive dashboards in a week or two. Now we’re able to combine the market basket analysis with an understanding of customers’ linear decisions in repeated visits to the showrooms.”

The system meets not only Rooms To Go’s current IT needs, but has plenty of options for expansion—an important consideration for any business with an eye on growth. For example, the 3 TB IBM System Storage DS3200 has enough space to eliminate any scalability concerns for the near future. “We’ve loaded three years of data and we haven’t come close to using half the space yet,” says Rosen. “We’re not using DB2 Compression yet; we’ll investigate that, but we don’t need it so far.”

Rosen didn’t have to be told that few project variables are more important to success than payback. Even before the analyses begin to deliver results in campaign design, the internal systems are expected to deliver results in less than two years by eliminating the costs of the outsourced marketing systems. The entire package, including hardware and software, cost less than some of the design-and-develop–only proposals Rosen received. But the real payoff will come from Rooms To Go’s ability to develop campaigns that compare favorably with those of larger organizations—adding to the top line.

Solution components

  • IBM InfoSphere Balanced Warehouse
  • IBM Retail Data Warehouse
  • IBM InfoSphere DataStage
  • IBM Cognos Business Intelligence
  • IBM DB2 for Linux, UNIX, and Windows
  • IBM System x3650 server with quad-core Intel Xeon processors running at 3.0 GHz with 4 GB RAM
  • IBM System Storage DS3200 with 3 TB preconfigured storage