Another early project, which aimed to assist admissions, fed current applicants' data into the data warehouse. Assessing applications—who was applying, how desirable candidates were in comparison to others, who had accepted and so on—had been a matter of looking at outdated information. Through business intelligence reporting, decision makers can see daily changes to the application mix. This has enabled the school to choose more selectively based on the right combination of students they want to admit according to such factors as academic excellence, leadership, diversity of experience, geography, gender and ethnicity.

Create one version of data truth.

Rensselaer created cross-functional groups to establish data accuracy, including teams which created common data definitions. Such activity, no matter how difficult it may be to get agreement, is crucial for the success of a business intelligence project. Kolb lists "one version of the truth" as the top reason why the effort has borne fruit.

Colin White, founder and president of BI Research, agrees. End-user-facing business intelligence tools and dashboards may seem to be the sexier aspects of BI, but data warehouses and the data governance to ensure clean and consistent data are the foundation of successful business intelligence. "Garbage in, garbage out: Unless you've got accurate data you don't get the benefits," he says. Deploying an enterprise data warehouse is often the one way you can guarantee clean and consistent data.

To ensure that data stays clean, you must put processes in place to make sure people find mistakes and correct them, says Ora Fish, project manager of data warehouse and business intelligence for Rensselaer. The university designated "data stewards" and "data experts" early in the project and made them accountable for the cleanliness of the data. "People are not going to do it out of the goodness of their heart," says Fish. Functionally, the data stewards are associate vice presidents and are responsible for setting procedures and policies for data management. Data experts are typically senior managers who report to the stewards. The data experts meet to set campuswide data definitions and take responsibility for any necessary cleanup of existing information.

In addition, Rensselaer has other ways to enforce data cleanliness. One way is to define business rules for the data warehouse so that erroneous data is rejected (for example, if an expenditure is submitted against a fund that is not active or if a student is given two acceptances) and an e-mail is sent to the owner to correct it. Another is to mark nonstandard data (such as a grant that is submitted without a code, with permission from a key business user). A third is to have reports and queries analyzed regularly by data experts.