BUSINESS PROBLEM-SOLVING CASE: Can HP Mine Success from an Enterprise Data Warehouse?
Hewlett-Packard, the $98.5 billion manufacturer of personal computers, server computers, printers, and provider of consulting services, is in the middle of a business transformation. The company is trying to reduce yearly spending by its information technology department by 30 percent over five years. HP expects information technology expenditures to drop from just over $3 billion in 2003 to $2.1 billion in 2008. HP is reducing its information system applications from 5,000 to 1,500 and consolidating 85 computer centers to six. HP’s current IT infrastructure employs between 19,000 and 22,000 servers. The consolidation will decrease the total by 8,000 to 9,000.
The success of HP’s business transformation may hinge on one particular project. HP is building a 400-terabyte data warehouse to serve the entire enterprise. If successful, the data warehouse will dispose of 17 different database technologies and unite 14,000 databases currently in use. If the initiative fails, HP would join a long list of organizations that have been confounded by the complexity of implementing enterprise-wide databases.
From an internal perspective, HP’s data warehouse aims to give its workforce access to data in real time with no departmental or geographic boundaries. HP’s numerous systems and applications had serious data management problems. CEO Mark Hurd had difficulty collecting and analyzing “consistent, timely data spanning different parts of the business.” Some systems tracked sales and pricing by product, while others tracked sales information geographically. Commonly used financial information, such as gross margins to measure profitability, were calculated differently from business unit to business unit. The company was obtaining information from more than 750 data marts.
Lack of data consistency dragged down sales and profits. Compiling information about the business from various systems could take up to a week, so managers had to make decisions based on relatively stale data. Seemingly simple questions, such as how much the company was spending on marketing across its different businesses, were difficult to answer. Without a consistent view of the enterprise, senior executives struggled with decisions on matters such as the size of sales and service teams assigned to particular systems.
HP CIO Randy Mott began consolidating the data marts in November 2005 into a single data warehouse serving the entire enterprise. He created a team composed of 300 people who were running the data marts and charged them with modeling the enterprise wide database that would be the foundation of the data warehouse. They had to make sure that the data would always be up to date, consistent for the entire enterprise, and complete.
The company launched its enterprise data warehouse in May 2006 to coincide with its consolidation of applications and data centers. To date, HP has consolidated hundreds of data marts into just over 200. The data warehouse contains 180 terabytes of raw data and 75 terabytes of functional data. At some point in 2008, the size should double and the data warehouse will be complete. Fifty thousand HP workers will utilize the data warehouse. All HP financial data will be able to be accessed via the data warehouse.
HP believes so strongly in its development of the system that the company is trying to sell its expertise to other companies that are seeking data warehouse technology. HP is marketing a product called Neoview, which has been developed from the proprietary work that the company has done in creating its own data warehouse.
More than 100 database specialists and software developers at HP are perfecting the system’s dexterity with table joins and giving it the ability to perform analysis functions at the same time that it is managing new incoming data. HP is also enhancing Neoview’s management and monitoring tools.
The servers employed by the Neoview system utilize Itanium processors from Intel, so they meet industry standards and are far more versatile than servers with proprietary technology. The system is also highly scalable and promises availability 99.999 percent of the time.
HP’s first customer for the Neoview database system was Bon-Ton Stores, which operates 272 department stores and 7 furniture stores in 23 states. Bon-Ton purchased a 7-terabyte Neoview system for a data warehouse that includes merchandise, customer, and supplier data for merchandise analysis and marketing. One of the database tables for the warehouse has more than 4 billion rows. Bon-Ton’s CIO James Lance reported that the Neoview system exceeded expectations.