Part 2: Is Your Data Warehouse on Life Support?
In an earlier blog, I argued that the data warehouse is far from dead and gave three reasons why. This article explains how your data warehouse may not be dead, but could be on life support. If you are in this boat and you want to pull the plug on your implementation and relieve it of its misery, this article will shed light on ways to improve your implementation.
In the last article, I mentioned the benefits of the enterprise data warehouse (EDW), but what are some of the pitfalls? So many BI projects crash and fail, how do you keep your EDW out of the emergency room?
1. Bad Implementations A big reason EDWs end up on chemotherapy, is the implementation itself. Many are poorly executed from the beginning—underfunded, understaffed, and or unsupported.
2. What versus why. Many times, when IT meets with the business, either prior to the meeting or shortly after, there is a mound of paperwork in the form of requirements that more often than not, serve very little purpose. Other times, the BI team chases its tail, creating multiple iterations of a dashboard to please the customer. I’ve presented 436 versions of a dashboard to an executive before he finally says, “That’s it!”
One issue is that IT teams focus on the “What”. You may say, “How can we start building something if we do not know ’what’ the customer wants?” To which I say, “When have you EVER written a BI solution for which the customer knew exactly what they wanted? “ I submit that the real question isn’t “What?” but “Why?” Although the “what” of a project or a customer expectation may change over time, the “why” usually remains static.
3. Lack of customer collaboration. Get in the stream with the customer, understand their business, allow their mindset to help drive the solution. Do not ignore this step, or ignore the business. In order to provide the business with the proper solution, you must collaborate, which is different than just meeting with the customer.
4. No expertise/talent. Another reason EDW projects fail or have a “do not resuscitate” proxy is due to the people who created the warehouse to begin with. You need smart, experienced individuals to lead this effort on a full-time basis. You will not achieve success with part time staff, or by cross training newbies to handle the gig.
5. No Roadmap. You must have vision, and you must start small. Where there is no vision, the people and the BI implementation perish. We know plans and people change. However, at any given point (especially at the inception of the effort), you should have a clear definition of where the team is going, supporting the goals of the company.
Many data warehouses were built without a full understanding of what a data warehouse is, or what is truly needed to accomplish the task. This is not a slam towards our industry, simply a fact that must be stated to understand the following point: well established, well-built data warehouses are a critical component of corporate data infrastructure. . These systems run businesses, enabling business users to make data-driven decisions and ask questions about any part of the organization.
With the influx of siloed reporting systems, there has never been a time when the enterprise data warehouse is more needed. However, it must be done right. You should take measures to ensure that your BI program and data warehouse takes the following concepts into consideration:
- Conformed dimensions
- Staging area/database
- Well defined metrics and attributes
- Supported by IT
- Well-tuned database
- Data from across the enterprise
- Matching metrics (from different domains, like date or release date)
- The right team, dedicated
- The right process
- The right questions
- The right tools
- The right rollout plans
- The right training
- The right support
- The right funding
- The right sources
- The right attitude
Keep these concepts first place in building your business intelligence team and you are well on your way to success.