Few are the companies who, if given the choice, wouldn’t do their data warehouse project over again, knowing what they do after the process. A quick survey of published case studies of even successful data warehouse implementations turns up recurring lessons for those embarking on a project:
Start with business needs. Many businesses build data warehouses on the “it’s the thing to do these days,” or “everyone else has one” rationale. Some enterprises don’t even need data warehouses, which a careful assessment of business needs should show.
“My advice to anyone who is just starting to embark on data warehousing is to focus on what business needs data warehousing can serve,” says Gene Alvarez, data administrator at women’s footwear and accessories maker Nine West Group, Inc. “Build a data warehouse to support those needs and grow from there. If you try to do it all up front, it will delay your implementation as well as endanger success.”
Allow for project ownership. Handing off the project to new leadership during its implementation can needlessly delay or even scuttle a data warehouse project. David Thompson, vice president of information warehouse services for PCS Health Systems, Inc. says, “Before beginning a major data warehouse initiative, it’s important to designate a single group or department to follow the project from start to finish, from data extraction to the application running on the workstation.” He also advises: “For the sake of efficiency and control, none of these tasks should be outsourced.”
Maintain strong IS/user communication. Data warehousing is not “an IS project.” It’s an enterprise project that will be used by the business side of the company. Beth P. Hassinger, manager of information infrastructure for BankBoston says, “through trial and error, [we have] discovered several factors that are key to making a data warehouse successful. First is a strong IS/user partnership.”
Companies that don’t bother to get end user input at every stage of the process run the real risk of wasting money on a data warehouse end users either won’t or can’t benefit from.
Keep the project flexible. Data warehouses should be built to handle queries that aren’t even conceived of at the time of construction. “Flexibility is critical in a decision support system environment, as users’ needs can change rapidly,” says Tony Marshall, spokesperson for Hallmark Cards. Marshall says, based on Hallmark’s experience, flexibility is crucial in “getting functionality into the end users’ hands as soon as possible.”
Have realistic time goals. Not only should you not rush a project, but conversely, you shouldn’t delay rollout. Ron Clark, data warehouse specialist for the investor-owned Louisiana power provider CLECO, says, “When making time estimates on that first project, do not forget to allow time for the learning curve. Keep the implementation in a three- to six-month range.” If it takes longer, users will start to lose interest. “We found that we could maintain the interest level of our users if we quickly delivered smaller components of the data warehouse,” says Clark.
