To be Successful in Business Intelligence, Build Four Pillars
Effective Business Performance Measurement, Business Intelligence, and Data Warehousing requires more than simply ‘throwing a BI tool’ at some data. After 20 years in BI, we’ve come to the realization that success rests on four ‘pillars’. If you’ve got these covered, you’re on the right path. These pillars are Culture & Business Process, Data, Architecture, and Business Intelligence.
These four pillars are defined below:
CULTURE & BUSINESS PROCESS
As I’ve noted here and here, BI success is much more than just rolling out a system that works (oh, and here, too). Real BI success means rolling out a system that gets used. In many cases BI is optional - people do their jobs today without it so just throwing data at them doesn’t mean they’ll use it. On the other hand, understanding where BI can enhance the business and then designing business processes with BI at their heart helps ensure that the new BI system will get used. Thus, while the tool might be BI, the goal is not BI - the goal is improved business processes. And, to ensure that your BI system will support those business processes, you need to think them through BEFORE you design it.
At the same time, some organizations are data oriented / data friendly and some are not. In designing your BI system you must be realistic about your culture. Will your people take to it? If not, are there some people who will take to it and can feed it to their coworkers? Are there managers willing to consider data-drivenness (yes, not a word but better than what the spell checker recommended - data-drunkeness) when evaluating their staff? If not, perhaps you need to start with with lower BI expectations.
It doesn’t really matter whether you’re using a Kimball approach, a normalized approach, some hybrid, a punched card approach… What does matter is that you have a data modeling and storage approach that will work and can be justified, beyond simply what some book tells you to do.
In the end, we’re largely talking about data models: do they correctly model reality and still provide the necessary information and query performance? The data pillar isn’t talking about technology but, instead, about your ability to represent the business in data models. Along those same lines, do you even need a data warehouse at all? While everyone enjoys a good data warehouse project, perhaps you can address your specific situation by applying a BI tool directly to your OLTP data.
Architecture represents the technologies underpinning your system - do you have the processes and tools necessary to reliably get clean data to the places where it is needed?
Business Intelligence is the piece that the user sees: can you develop and deliver reports, analyses, and predictive analytics in forms required by the business users & processes? Remember, BI isn’t just one thing. It really encompasses a wide variety of technologies such as prompted reporting, slice and dice OLAP, data mining and predictive analytics, geographic information systems…
Remember, too, that it usually makes sense to set BI standards for categories of users but not for whole organizations. Why? Well consider the case of the BI tool SAS. SAS is used by analysts, frequently with a statistical background, to build predictive models. Would you put it in front of your C-level executives? Probably not. The point being that SAS may be the standard for your business analysts but it shouldn’t be your sole BI standard.
What do do with this framework? Well, I’m afraid it gets a touch more complex now. For each of the pillars, examine your people, your processes and your technologies in light of your goals. Are they sufficient? Where are the cracks?
In future posts I’ll provide more details on the kinds of tasks and staffing each pillar requires as well as the consequences of failing to build that pillar sufficiently strong.
Let me know your thoughts.