Home Depot Renovates Its Data Warehouse

How could it be that Home Depot, which ironically bills itself as a home-improvement warehouse, doesn't have a data warehouse? The lack of one is particularly incongruous considering most major retailers spent much of the past decade or more building and using such data repositories to gain a competitive advantage by providing a granular view of sales, creating forecasts and monitoring a variety of other aspects of the business.

The truth is, the company has fared quite well without one. Home Depot, which will celebrate its 25th anniversary next year, experienced rapid growth in its relatively short history, becoming the world's second-largest retailer,with $53.5 billion in revenue last year,eclipsed only by Wal-Mart.

But with the global economy in decline and archrival Lowe's moving into markets where Home Depot had been the only game in town, new CEO Robert Nardelli is leading the Atlanta-based retailer's charge. Nardelli, the longtime General Electric executive who was passed over to replace Jack Welch in favor of Jeff Immelt, was recruited last year to help fix up the retailer by boosting revenue and adding to the bottom line. Key ways he plans to make that happen are by reacting in real time to customer-buying patterns, utilizing better price-optimization methods, fine-tuning merchandising, inventory management and assortment planning, and,perhaps most important,looking for ways to improve customer service and maximize inventory management.

"It's important to know what products our customers are buying, what other products our customers may want to buy and how to sort like products that are complementary," says Kevin Murphy, Home Depot's vice president of information management. It's not that Home Depot hasn't saved that data,it just hasn't been easy to crunch that data because it resides in an IBM DB2 database running on an operational mainframe platform.

Silver Lining

There is, however, a major upside to the fact that Home Depot doesn't have a data warehouse. Many companies that have built data warehouses in the past have done so with older technologies or may have done so incorporating only specific silos of the business.

"[Home Depot has the ability to build it with not only the latest technology, but to build it in a way where they can get immediate return on the warehouse," says Karen Parrish, vice president of business intelligence solutions at IBM. "What we find with a lot of customers who built warehouses 10 or 15 years ago is they had the data collected in a particular place," she says. "But the real value of a data warehouse is not in placing the data inside the warehouse, but actually in pulling the data out of the warehouse, analyzing it and gaining insight. And that's where the warehousing opportunities of the past seem to fall short of really delivering returns to our customers on their investments."

By creating a data warehouse in an offline repository, managers will be able to use their preferred analytical tools, whether they be from Brio or Business Objects, among

others, to view data and create reports in real time, says Murphy, who was tapped for the effort by recently hired CIO Bob DeRodes. Murphy engaged in a similar effort for DeRodes at Delta Air Lines before the two came to Home Depot earlier this year.

According to Murphy, the issue facing Home Depot is that it doesn't have the data in an operational data store that lets managers easily and in near-real time query specific information. To make that possible, the new data warehouse will enable an instance of DB2 running on IBM's Unix-based pSeries Regatta server. The first version of the warehouse will operate on two 32-node pSeries 690 servers running AIX 5.1 with up to 60 TB of storage.

"In order to really do analytics properly, we felt we would get a lot more accomplished [by moving that information from the operational environment to one tailored for analytics, so you can model data differently, look at different views and have the horsepower to run some of the more sophisticated statistical-behavioral models and pattern-recognition models," Murphy says.

The fact that Home Depot has a clear goal for what it wants to gain from the data warehouse improves its chances of success, IGS' Parrish says. "They know they want to increase customer service, they know they want to increase the [ratio between the sales associate and the customer, and they know they want to focus on their merchandise to have the right selection in the right store based on customer-buying patterns. When you know what your goal is and you build a warehouse to achieve those goals, you do it right the first time."

Phased Implementation

Still, Gartner analyst Kevin Strange says while Home Depot could gain significant business advantages, the retailer is taking on significant risk. "This may not work out the way they are hoping because DB2 is not a proven database as an enterprise data warehouse," Strange says.

That risk, however, is mitigated due to the fact that the data warehouse is being implemented on an application-by-application basis. "Absolutely, the only way you can do data-warehouse implementations successfully is to do them in a phased implementation," Strange says.

Those issues are not foreign to Home Depot's Murphy, who agreed with the phased-in approach, but also wanted to move quickly. In August, when the project first got under way, Murphy promised to deliver the first tangible use of the data warehouse within 120 days. With help from IGS, which is providing project management and retail IT support, and solution providers, such as Doublestar, that have data-modeling and decision-support expertise and are helping cleanse the data residing in the existing database, Murphy says the first application to help assess employee productivity and other metrics should be ready this month.

The first step was to tackle HR and personnel management. With approximately 300,000 employees and a high level of attrition that is typical with retail, Home Depot opens, on average, three new stores per week. As a result, Home Depot in a given year hires as many as 50,000 new employees. Not only is human resources a ready-and-willing participant, it is technically simple to start with because making a data warehouse available to other groups requires involvement from multiple functional areas and data sources.

"With HR, I had a very well-prepared and focused customer," Murphy notes. "In an enterprise data warehouse, you are only as good as your business relationship. You just can't build this to what IT wants to build,you have to have a firm base of business knowledge. Quite frankly, our HR department is very skilled, and they knew what they wanted."

The HR data is being fed into the new data warehouse, where employees can use their own extract transformation and load tools to analyze the data. Currently, the data is coming from Home Depot's legacy HR system, called Tesarec. But the company is deploying a PeopleSoft HR system, which will be fed into the data warehouse, Murphy says. That system will let managers associate employee attributes with up-to-date sales records and other metrics.

The choice of analytic tools will be a key factor in whether the implementation pays off, Strange says. While Murphy says those decisions are still being made, there is no shortage of online analytic processing and reporting tools that let individuals from sales managers to business analysts slice and dice data pulled in from the data warehouse. In addition to Business Objects and Brio, key suppliers of data analytic tools include Cognos, Crystal Decisions and Hyperion Solutions. In addition, ERP and CRM suites have their own reporting tools.

Murphy says the current crop of business-intelligence tools on the market are much easier for novices to use. "If Home Depot had built a data warehouse earlier and decided to update it, I would have the complexity of first doing a pretty big conversion from the tool I had selected previously," Murphy says. "From that instance, we will have the advantage of not having gone first."

Making sure data is cleansed properly is also important, adds Doug Laney, an analyst at the Meta Group. "The biggest hurdle is data quality," Laney says. "The data integration can be strong, but trying to cleanse all the data and deal with different infrastructures and semantics can be difficult."

Following the HR stage, Home Depot plans to give other parts of the business access to the data warehouse, with a goal toward providing real-time sales forecasting, pricing, space planning and optimization, and inventory management.

If all goes well, this could prove the most productive warehouse Home Depot has constructed in a long time. The coming year will tell whether Home Depot's own home improvement project pays off. n