Hudson's Bay builds a data warehouse to improve inventory levels

Established in 1670, Canadian retailer Hudson's Bay Company (HBC) operates more than 500 department, mass merchandise and specialty stores under the Bay, Zellers and Home Outfitters banners with 70,000 employees. HBC's stores generate vast amounts of data across its formats and systems, creating a challenge to quickly access information.

To make appropriate merchandising and product management decisions, associates used to rely on reports that could take up to a week to produce. Although useful, associates found the reports were not very thorough, so they often based inventory and buying decisions on instinct, as opposed to analysis.
"We saw data warehousing as a way to consolidate and share data," says Mary-Jane Jarvis-Haig, senior manager of data warehousing at Hudson's Bay Company.

In order to improve inventory levels and increase gross profit, in conjunction with other large operational system and business process improvements, HBC created a data warehouse with the help of NCR/Teradata.

Developing the Database
Since initial installation, the data warehouse has grown from 245 gigabytes to more than 1.7 terabytes of user data with HBC adding two nodes to the system each year as new stores and more data are added. The warehouse stores detail and summary sales and inventory data, with information refreshed nightly from more than 100 feeds. More than 400 users access the system to perform ad hoc queries and view standard reports through front-end access tools. "The data warehouse has contributed to improvements in inventory, cost reductions in specific areas of marketing, advertising, and understanding of customer trends and behaviors," says Jarvis-Haig.

The implementation of the data warehouse enables the retailer to perform fact based decision making. "This eliminates different interpretations of outcomes as the data provides a single version of the truth," says Jarvis-Haig.

The centralized data warehouse lets HBC analyze its business via multiple attributes rather than being limited to its traditional product and organizational hierarchies. In addition, cross enterprise analysis is now possible rather than only banner-by-banner.

HBC has planned a full platform configuration replacement for December 2004 "in order to exploit a new enterprise data model and build towards a true business intelligence capability," concludes Jarvis-Haig.