Data warehouse implementation strategies and practices

Data Warehousing at Wal-Mart

The intent of this paper is to evaluate the data warehousing project Wal-Mart implemented in June, 2007 to streamline business analytics and reporting from its 4,000 stores in addition to synchronizing demand with suppliers through its Retail Link system (Weier, 32). Implicit in the design of this data warehousing project is the integration with existing databases throughout Wal-Mart, many of which are stored in Teradata data warehouses that have not been synchronized to the Retail Link system (Havenstein, 14), and also need to be incorporated into the process workflows for the Radio Frequency Identification (RFID) projects that are piloted throughout the 20,000 Wal-Mart suppliers during 2007 (Wang, 45-66).

Wal-Mart's Selection of Data Warehousing Strategies

For Wal-Mart, creating data warehouses that provide demand visibility back to its supply chain is critical if the key measures of its internal efficiency are going to be attained. These include inventory turns and delivering orders that are flawless in execution and quality. The timeline for Wal-Mart's data warehousing project began in 2005 when the CIO and VPs in the it Division realized that there were several hundred stores that were not specifically having their data read into the Master Data Management (MDM) hierarchy that had been created. Further, these retail locations, many of them superstores only had limited visibility into what products would be available for delivery in the next 72 hours (Duff, 14) which made the many processes required for forecasting demand quite time consuming and imprecise (Foote, Krishnamurthi, 15). All of these factors combined with the growing reliance on the Retail Link system in Wal-Mart for planning new store expansions made an enterprise-level data warehousing project a high priority. The urgency surrounding the project was accentuated by the need for having a higher level of accurate of initial products in the superstores when they opened, based on demographic analysis of the surrounding areas. The data warehouse was needed to align specific customer segments with the mix of products in the initial superstores when they opened.

Wal-mart realized that the mainstay and most valuable customer segment they have is the value shopper, typically a head of household that relies on Wal-Mart's prices to make ends meet. Getting the right mix of products at the opening of a new store was crucial to attracting new value shoppers to the newer stores. In addition to all these factors was the growing importance of RFID as a means of tracking inventory positions and the velocity of inventory through its distribution centers. Prior to June, 2007 when this data warehouse was rolled out there was no centralized database to track and report on the sales of RFID-based products by retail locations. Without this data, the RFID pilots that Wal-Mart had mandated were not showing any Return on Investment (ROI), and this was causing ill will with some of the larger suppliers including Proctor & Gamble, and Unilever, who wanted to know the ROI of their system investments in RFID. All of these factors combined to form the catalyst of the data warehouse project being made a higher priority than the many other competing projects within Wal-mart at the time.

At a cost of approximately $3M in software and $9M in services and training, Wal-Mart partnered with Hewlett-Packard and became one of the first companies to adopt the HP Neoview data warehousing system. One of the key reasons for Wal-Mart choosing the HP Neoview data warehousing system was its high level of compatibility with their existing investments in the Teradata databases, which were used as the foundation of their Business Intelligence applications. Wal-Mart also wanted to create a more consistent integration platform to support their Retail Link system so that the RFID data could be used for calculating the ROI of partner investments in the technology.

Lastly, HP Neoview runs on the existing Tandem NonStop operating system and kernel platform to ensure the highest levels of performance possible. Despite the HP Neoview being a relatively new enterprise-level application for data warehousing, Wal-Mart evaluated the performance of 750 data marts that HP was managing with Neoview prior to formally launching the application for sale. The result of HP thoroughly testing Neoview internally helped to convince Wal-Mart the application was tested well enough to be used in the demanding data warehousing, RFID analysis and Retail Link integration.