Hudson's Bay Company conquers the information wilderness

Flexibility and a thorough knowledge of its business keep Canada's oldest corporation strong through the centuries.

by Eric Schoeniger

A lot can happen in 335 years.

Hudson's Bay Company (Hbc) has seen its share of change. The venerable Canadian retailer traces its roots to 1670, when it began as a fur trader in the North American hinterlands. In the early 20th century it transformed its far-flung trading posts into Canada's first department stores. By the 1980s, following a series of acquisitions, Hbc was well on its way to becoming Canada's largest department store retailer.

But a successful history doesn't mean there won't be challenges in the present and the future. Like many retailers, Hbc is focusing on the future, which promises to bring more rapid and total change.

In North America and across the globe, retailers are grappling with flat sales, fierce competition and a diversification of sales channels. In particular, customers are more demanding, and their buying habits have become more eclectic. The same customer who drives across town to save 10 cents on laundry soap will pay a premium for a latte.

That means retailers everywhere must better understand customer behavior so they can make the most profitable decisions about product, price, promotion and placement. And that calls for a new approach to data management, one that gives a broad, comprehensive view of the business.

"It's one thing to know which products you're selling," says Mary-Jane Jarvis-Haig, senior manager of business intelligence development and support for Hbc. "But you also need to understand who you're selling them to and in what locations ... And you need to do that across all your channels."

Toward that end, Hbc has been working with Teradata on a major initiative to better understand customer behavior, anticipate customer needs and ensure that the right products are in the right place at the right time. The strategy began several years ago with the development of data marts built on a Teradata Database. More recently, Hbc and Teradata have collaborated to unify those data marts into an enterprise data warehouse to give Hbc a single view of its customers and operations.

One customer, one retailer

With four distinct banners and an e-commerce program, Hbc has a unique opportunity to bolster its claim that it offers "everything Canadians need to live their lives." In fact, 90% of Canadians shop at least one Hbc store. The trouble is, Hbc's customers are not easily sorted into neat categories. The discriminating consumer who demands a high level of service at The Bay, Hbc's upscale department store channel, may also seek out lower prices at Zellers, the company's mass merchandising unit. At times, that same customer might even expect to meet all of his or her needs at a single Hbc banner.

That means Hbc must distribute its merchandise according to customer demand, not channel definition. And it must do so while continuing to differentiate each banner so customers know what to expect from their shopping experience.

Bringing data home

In the late '90s, Hbc began to build the data infrastructure needed to better understand and serve each customer across all touch points. In part, the strategy involved presenting a common face to the customer through initiatives such as a credit card and customer-rewards program. The Hbc Credit Card is now accepted at all Hbc banners and helps the company capture data on customer demographics and spending patterns. Hbc Rewards, through which shoppers earn points redeemable toward merchandise, provides Hbc with valuable information about the program's 8 million participants.

At the heart of the data infrastructure is a Teradata Warehouse. The data warehouse, which was designed by Teradata and third-party consultants, was initially managed off-site. Then in 1999, "based on tangible business benefits of increased gross profit, the data warehouse was brought in-house," Jarvis-Haig says.

The physical implementation started with a single banner and then grew to reflect all banners—with what was essentially a separate data mart for each channel. Although Hbc shared data where it could, differing business rules and operating practices required that it keep much of the data in separate tables.

The initial focus was on inventory management, merchandising and sales. "We started capturing detailed information about inventory and sales transactions," Jarvis-Haig explains. "We then offered that information to buyers and planners—the people who make sure the supply chain is full and business processes are effective."

The data generated by Hbc's 500 stores is spread across multiple operational systems. In the past, merchants relied on the IT department to produce reports to support merchandising and product management decisions. But the reports could take up to a week to generate, which meant the information was obsolete before it was even delivered. As a result, inventory and buying decisions were made instinctively rather than on the basis of solid analysis.

Today, the data warehouse provides Hbc's merchants with a clear view of the business. "We can understand demographic differences so that we can offer a different assortment in different stores," Jarvis-Haig says. "We can also make decisions around price management so that we can sell more products and sell them in the most profitable way. And we can understand sales trends to answer questions such as, 'Are sales down because customers aren't buying, or because inventory is low?'"

Hbc has quantified a 300% return on its initial investment in the Teradata Warehouse. Business process improvements supported by data warehouse information have resulted in reduced inventory requirements, lower labor costs and fewer lost sales during promotions. Product in-stock rates have improved significantly, while cross-banner shopping rose by 16%.

Getting smart about business intelligence

Buoyed by its early success, Hbc was ready to further its infrastructure transformation to better support corporate strategies. Its early efforts, while highly beneficial, did not fully advance Hbc's current strategic goal of operating as a single enterprise.

In 2003, Hbc embarked on an initiative to unify the banner-specific data marts into a single, enterprise-wide Teradata Warehouse.

The first major rollout took place in October 2004, and the first phase of the business intelligence initiative is now about 70% complete. While the existing data marts served 600 users, the enterprise data warehouse will deliver information—about sales and returns, inventory, pricing, product hierarchies, markdowns, gross margin and more—to 4,500 employees and partners.

"Until now, we've been good at managing information about our individual banners," Jarvis-Haig says. "But our intention is to make enterprise-level information available across the company."

What's more, the enterprise data warehouse provides a foundation for data consolidation and process standardization. Jarvis-Haig explains: "Until information is consolidated and business processes are standardized, people are doing things differently. Decisions are made based on past experience or gut feel, rather than on data and facts."

When the rollout is complete, Hbc will be able to practice exception-based decision making. "Instead of spending time trying to find problems, our business leaders can focus on solving them," Jarvis-Haig notes.

The business intelligence initiative was designed specifically to support Hbc's business strategy of operating as a single enterprise—One Hbc. One reason for that alignment between IT activities and business objectives is that Hbc views IT as a competitive differentiator. "Our CIO sits at the table with the business strategists," Jarvis-Haig explains. "At the highest levels, the IT organization understands and influences strategic decision making."

It's a mindset that seems to penetrate Hbc's culture. Although Jarvis-Haig is part of the IT organization, she knows retailers' concerns and clearly understands Hbc's core business.

Working the best-laid plans

That's not to say Hbc doesn't face the same challenges of any large organization using technology to help support business goals. One of the biggest hurdles, Jarvis-Haig recalls, was ensuring business participation. "Business people are busy trying to run the company," she says. "But a project of this scale absolutely requires ongoing involvement from the business."

It took several months to assemble representatives from the business who could help define requirements for business intelligence functionality. A team of 30 people eventually contributed to the effort, but even then, Jarvis-Haig notes, "the reality is that business intelligence requirements can be difficult to define precisely. ... (Also,) in an undertaking this large, how do you predict the effort required, the amount of time and the potential roadblocks? It's not a science and building BI applications require a more iterative approach than traditional transaction processing applications."

The lessons learned? Carefully manage expectations and proceed in stages. But Jarvis-Haig admits that such an approach is easier said than done. "Our intention was to get the middleware in place first and stabilized, then define requirements and build a little to see if that's what the business wanted, and so on. But delays happen, business priorities change and new issues become urgent. Even though we had an advantage of an existing data model from which to migrate, the advantage was quickly reversed as unexpected delays on several fronts found us migrating data and defining requirements simultaneously."

It helped to have a team of Teradata consultants working alongside Hbc's IT staff. "Teradata was integral to the design and implementation of the data model," Jarvis-Haig emphasizes. "They helped us understand what the new data model should be, and they helped us transition from the old model to the new one."

Teradata experts also helped with Business Intelligence application performance challenges. And when Hbc began to reach its existing capacity limits, "Teradata worked with us to upgrade the infrastructure and increase capacity."

Results to grow on

With the first phase of the business intelligence initiative nearing completion, Hbc is already realizing positive results. "For example, we're approaching assortment differently, and some brands are now offered in multiple banners," Jarvis-Haig explains. "We're putting product where our customers want to buy it as opposed to where we want to sell it."

What's more, the company can track the effectiveness of such brand reallocation. "It helps us understand where the strategy is working well, so we can repeat that success," Jarvis-Haig says. "And it helps us identify where it isn't working, so we can determine whether it was a good strategy that wasn't well-executed or just a bad strategy."

In addition, Hbc can better analyze trends and compare the performance of peer-group stores to share successful strategies and increase sales. "We can view data at a very atomic level," Jarvis-Haig says, "but we can also understand the business at a higher level, across all the banners. That's enabling us to become more customer-centric."

When the project is complete, executives, merchants, stores and suppliers will have a single data warehouse, with one view of the business and the ability to report on all Hbc data. Business rules will be aligned across all banners, with decision making focused on the most promising opportunities. The solution will enable exception reporting, and information will be tied to key performance indicators for sales, inventory and gross margins.

But Jarvis-Haig believes that Hbc is just getting started with business intelligence. "We've established a center of excellence that we'll use to develop new applications to support other business functions. Right now the focus is on merchants and the supply chain. But we're also looking at CRM, financials, real-estate management and more. They're all anxious to take advantage of the platform."

After 335 years of change, Hbc has learned how to use new tools to its advantage, and the company's embrace of data warehouse and business intelligence technology is no exception. T

Taking a bite out of retail fraud

RETAILERS LOSE BILLIONS of dollars to theft and fraud—dollars that, in effect, come right off the bottom line. And much of that lost revenue—$16 billion a year, according to the National Retail Federation—is attributable to returns fraud. Returns fraud occurs when a customer brings back a shoplifted item or reuses a sales receipt to "return" an item multiple times. Many retailers are addressing the growing problem by changing their returns policies—shortening the time period in which customers can return an item after purchase, say, or offering store credit instead of cash refunds. But as Hudson's Bay Company began development on its enterprise data warehouse, it realized it could use technology to take a bite out of fraud. Hbc combined a Teradata Warehouse, MicroStrategy business intelligence software and IBM's WebSphere middleware to create an electronic link between customer purchases and returns activity. Now, within moments of a transaction at the point of sale, the data warehouse is updated with sale, return, exchange and void data, allowing instantaneous system response to returns. This means a receipt can't be reused to fraudulently return merchandise. And it means that legitimate returns are processed with greater speed and accuracy, improving customer service. Within hours of the first pilot installation, the solution stopped a criminal ring that had used a single receipt to defraud Hbc of thousands of dollars. Within five months of rolling out the solution to all Hbc stores in October 2004, the reduction in losses had delivered a 100% return on investment. "The cost savings have been huge," says Mary-Jane Jarvis-Haig, senior manager of business intelligence development and support for Hbc. "We're already exceeding our targeted benefits." One could say Hbc is also anticipating many happy returns.

Behind the solution

Teradata Warehouse powered by: Teradata Database V2R5.1.1, 10-node NCR 5380 Servers (eight active nodes, two capacity-on-demand nodes) Storage: Total Disk: 9TB (5TB currently in use) Operating System: UNIX MP-RAS Teradata Utilities: ARCmain, BTEQ, FastLoad, MultiLoad Tools/Apps.: Teradata Database Query Manager, Teradata Priority Scheduler, Teradata SQL Assistant, Teradata Manager, Teradata Visual Explain

Eric Schoeniger is a journalist whose work has appeared in Business Week, Chief Executive and Future.