Success Factors of Data Warehousing

1. Expectations are communicated to the users

IT is often unwilling or afraid to tell the users what they will be getting and when. Users should be told about the following:

• Performance – what to expect for response time
• Availability – scheduled hours and days
• Function – the data that will be accessible and what pre-defined queries and reports are available. The level of detail data, as well as how the data is integrated and aggregated
• Historical data
• The expectations of accuracy for both the cleanliness of the data and an understanding of what the data means
• Timeliness – when the data will be available and how frequently the data is loaded/updated/refreshed
• Schedule expectations involve when the system is due for delivery

2. User involvement is ensured

There are three levels of user involvement, as follows:

• Build it; they will use it
• Solicit requirements from the users
• Have the users involved all the way through the project

The last level is by far the most successful approach, while the first almost always results in failure.

3. The project has a good sponsor

The best sponsor is from the business side, not from IT. Most importantly, the sponsor should be in serious need of the data warehouse’s capabilities to solve a specific problem or gain some advantage for his or her department.

4. The team has the right skill set

Without the right skills dedicated to the team, the project will fail. The emphasis is on “dedicated to the team.”

5. The schedule is realistic

The most common cause of failure is an unrealistic schedule, usually imposed without the input or the concurrence of the project manager or team members. Most often, the imposed schedules have no rationale for specific dates, but are only means to “hold the project manager to a schedule.” A realistic schedule will include all the required tasks to implement the project along with their durations, assigned resources and task dependencies.

6. The right tools have been chosen

The first decisions to be made are the categories of tools: Extract/Transform/Load, data cleansing, OLAP, ROLAP, data modeling, administration, and so on. The tools must match the requirements of the organization, the users, and the project. The tools should work together without the need to build interfaces or write special code.

7. Users are properly trained

In spite of what the vendors tell you, users must be trained and the training should be geared to the level of user and the way they plan to use the data warehouse. All users must learn about the data, and power users should have additional in-depth training on the data structures.