"The most important step in successful data mining is setting your objectives," says Rob Gerritsen, president of Philadelphia- based data mining consultancy Exclusive Ore, Inc. In particular, objectives must be actionable, and, ultimately, profitable. According to Gerritsen, mined data should "lead you to being able to change the way you do business."
Gerritsen and Exclusive Ore vice-president Estelle Brand share their secrets on how to prepare for successful data mining projects:
- Make it a team project. In almost every successful data mining project, the end-users, those who will ultimately be using the data, are the impetus for the project. But, others must be involved. "Convince the appropriate people in the organization that it needs to be done," Gerritsen says, and point out that involvement from groups other than end-users will be profitable for the organization. Keep in mind that the call for a data mining project will come from whichever group in a company has an agenda for using the data, but successful projects are always team efforts between the business users and I.T. department.
- Pick your team based on your goals. Both Gerritsen and Brand stress to "look at what you're going to do at the deployment stage, since this indicates what people need to be on board the project." This is completely dependent upon how the data will be used - if the data you mine is going to be used by the sales and engineering departments, involve those people in the early stages of the planning process.
- Don't forget Team Marketing Department. Having input early on from the marketing department is vital. They view data mining as an important way of being able to target customers, says Gerritsen, hence the marketing department "is the early major adopter of this technology," and probably will even be the instigators.
- Involving the whole organization prepares it for change. Since successful data mining usually results in changing the way your company does business, well-received and widely used data mining projects involve as many departments as possible. "People who aren't involved are going to feel completely blindsided," Brand warns, and probably end up resisting the process at crucial moments.
- Create a game plan. Have the business users and the technology department create a "strategy and tactics" breakdown of the project, Brand says. The overall strategy should explain how the mined data will help the company's performance - this is attained from the end users. Creating the framework for the tactics, or exactly how the data mining will be done, comes from those within the information technology department.
- Inventory your data. Create and organize an inventory of your preexisting data. Figure out where it is - if it's already in a warehouse, you're ahead. This will help you determine what is missing and what isn't, where it your data lies, who's responsible for its maintenance, how much of it there really is, and give you an indication if you need to add external data.
- Choose your plays based on your goals. Select your hardware and software based on the organization's goals and end-usage requirements that are firmly in place ahead of time. In other words, don't let technology drive you, make technology fit your needs.
Rob Gerritsen and Estelle Brand will be presenting a seminar on this topic June 26th as part of the Database and Client Server World Boston event, and will present the seminar July 26th at the Data Warehouse World Conference in New York.