When considering BI (business intelligence) projects, organisations would be better off in the long run to set up an enterprise-wide data warehouse, rather than running different data marts for individual projects, advised Michael Jones, senior vice president of IT for the Gap clothing retailer.
Jones spoke at the National Retail Federation's annual conference, held in New York last month.
As a result of running a single enterprise data warehouse, rather than maintaining a series of data marts set up to answer a specific set of questions, the Gap's managers can get a wider array of metrics and reports about what is happening across all its stores and different brands.
"We're able to answer many more questions as if we had the data in silos, which we have had in the past," Jones said.
The Gap has multiple brands to keep track of, including Old Navy, Banana Republic, The Gap and the online-only PiperLime. Over 3,100 of its stores are spread out across North America, Europe and Japan, and the company has franchises in 25 countries. The company reported US$14.5 billion in sales in 2008, including more than $1 billion in online sales.
"Our challenge is to try to provide information across an entire enterprise," Jones said.
The company's first experience with BI was to deploy a number of individual data marts, each designed primarily to answer one set of questions for one business unit. "We used projects to drive what went into our data marts," Jones said.
Over time, however, that data was reused in larger, enterprisewide assessments, which had varying degrees of fidelity to the business as a whole. So in 2000, the company kicked off an initiative to build an enterprise-wide data warehouse, one that would draw data from all the operational systems. Though it would cost more, it would provide a wider array of answers.
The overall goal was to get updated business information that could be used to measure the results of business strategies. "We have to have up-to-date information, across our enterprise in an almost real-time methodology," Jones said.
The key to this job was the ability to integrate data across different business units. "Learn once, use many times," Jones said.
That Gap contracted Teradata to set up the enterprise data warehouse. Teradata's retail data model served as a template for its own data model. A data czar was also appointed to work with each of the brands to make all the data and the definitions for that data were consistent across the different segments.
"We're taking a much more data-centric approach to the business," said Mark Brennan, vice president of infrastructure and data warehouse services.
The two major categories of data the company keeps is on sales and inventory. In simple terms, the company has about 15 different questions it needs to ask on a regular basis on how well its products are selling, and about 25 concerning its inventory.
Here is where the efficiencies of scale kick in, however. By aligning data from all the business units into this single data warehouse, the company is able to answer a total of 96 questions, more than the number of questions it could answer by keeping the sales and inventory data in separate silos.
Plus, the company can now answer new questions and allow its managers to produce unique reports.
"We are able to answers many questions that we were not able to do before," Jones said. The Gap uses query software from MicroStrategy, so if a business manager requires a unique analysis of data, one can be served up fairly easily.
Overall, setting up an individual data mart only costs about 10 percent to 20 percent of the price of setting up a full-fledged data warehouse, but the extra flexibility is well worth the cost, according to the Gap.
For instance, the Old Navy stores do a lot of promotions. With the data warehouse in place, the brand managers can get "a lot more immediacy and feedback on the those promotional events," Jones said.