1. Use what has already been purchased - Most installations have an RDBMS which will serve them well as the store for the data warehouse as well as a query or report writer. By using what is already available, the cost of the data warehouse is reduced and the organization does not have to spend as much time and money on the training and the learning curve required for bringing in as many new software products.
2. Don't let management set an artificial deadline - The data warehouse project manager needs to be proactive, develop a project plan complete with tasks, deliverables, durations and assignment of responsibilities. The project plan will produce a completion date that management can now see and, hopefully, will accept.
3. Assemble a small team of good people for the data warehouse project. Since the data warehouse is new, and all the good people are already actively involved with other projects, management will often try to staff with people who have nothing to do. Don't accept incompetents or nay sayers on the team.
4. Chose an application of reasonable size. The initial pilot should not be so large that the inevitable performance problems of a very large data base will jeopardize the success of the project. Nor should it be so small that little is learned about managing a data warehouse.
5. Choose an application that makes a real contribution to the organization. Every organization has a wealth of data warehouse opportunities. There is never a reason to choose a data warehouse application that will be a throw away or will make no contribution to the company.
6. If a consultant is engaged, be sure that Job One is skills transfer. Too often consultants come in, work their magic and leave without teaching the IT staff how to develop the next data warehouse application.
7. The sponsor should desperately want and need the capabilities of the data warehouse. This helps to ensure their support when problems arise (they will).
8. Ask the vendors to back their claims with seasoned references and written assurances of their product's capabilities. Don't be snowed (or let your users be snowed) by flashy demos.
9. Allow adequate time for training - Even if the tools are intuitive, training is still required to understand the data and know how to validate the query results.
10. Clean the data - The users will walk away if their reports and queries are wrong. Dirty data will deliver incorrect reports. Measure results, usage and report benefits - If you don't know how the data warehouse is being employed and you don't know how it benefits the users, it will be difficult to ask for the budgets and resources necessary to maintain and enhance what has already been built.
Edited by Sunfish, 20 November 2009 - 06:18 PM.