Are there specific features or capabilities that companies should look for in a BI platform? Are there certain technological aspects or road map questions that organizations should be asking vendors that they’re evaluating?
Absolutely, one of the keys to a successful BI implementation is integration. And out of the many layers of business intelligence architecture, what we call the BI stack, the most important layer in my mind is meta data because ultimately you can actually utilize data integration from one vendor, data modeling tools from a second vendor and presentation from a third vendor as long as all of them are tied with either the same or an integrated meta data architecture.
And as long as I can change a business rule or a name of an entity only in one place, and as long as I trace my data linage from a report that I see on my screen all the way back to the source system, which is what meta data does, that is really the key.
So in terms of evaluating the business intelligence software, obviously the simplest decision would be let me look at all the products that cover every single layer of the BI stack from a single vendor. That works very well in SMB market, where it’s a smaller organization, and there are not as many IT resources, and we are not really looking for best of breed solutions. 80 percent of what we’re looking for is good enough, and therefore going with a single vendor is perfect.
|#1 Ranking: Read how InetSoft was rated #1 for user adoption in G2's user survey-based index
But in larger organizations where we have many departments, many lines of business, many different business requirements, sometimes it is just reality that a single vendor will not fulfill all of the requirements. That’s where most of the challenges lie, and if in those cases the implementers of technology insist on integrated meta data architecture, I think they will be a step closer to the right solid business intelligence architecture.
Let’s delve a little more into meta data. It’s a concept that’s been around for a long time and techy people are very familiar with it. People who have worked in database administration are familiar with it. How do you explain meta data to a business person?
Well, the traditional answer is meta data is data about data. It’s actually everything, but the data itself. It’s where did the data come from? When did it come here? What steps did the data take on the way from its ultimate source to the transactional system to the report that I’m seeing on the screen. Who had access rights to the data? Who prepared the report? When was that report last refreshed? What was the formula that was used to calculate a particular metric that I see on my screen?
So it’s everything but the data itself. Meta data repositories are a key part of meta data architecture, but it’s actually much more than that because very few organizations other than multi billion organizations, have the luxury of implementing large complex, centralized meta data repositories. And these efforts are typically so complex that the benefits do not outweigh the effort. So it is not the central meta data repository that is important. It’s the integration between all of the meta data capabilities, meta data portions of all of the BI tools that one puts together. It’s the integration part.
Moving back to our discussion about evaluating software, what are some key steps organizations take when they evaluate business intelligence software. They find something they think they like. What are their next steps?
I’d like to answer that by taking a step backwards a little bit. I do recommend that business initiatives always precede any kind of BI technology evaluation because if an organization does not have such a concept, such a methodology, such a process as data governance or data stewardship then every technological evaluation or implementation of BI tool will ultimately fail.
So what one needs to do first is make sure there is a strong senior executive sponsorship of a data governance initiative. The data governance initiative has representation from all of the major lines of businesses and not just from a front office, but front, middle, and back because requirements may be different. It’s important that the very first step that this organization takes is to create a common data dictionary. A set of metrics that the organization will be measured on and common definitions of calculations and methodology behind arriving on those metrics.
Once that is done the next step is still not technological. The next step is coming up with a business intelligence vision. Business intelligence is never a tactical project, never a tactical implementation. It is always a road map. It is typically takes no less than 2 or 3 years to implement the comprehensive enterprise wide, business intelligence environment. So it has to be vision.
There has to be a strategy, a vision of the ultimate target state and a very clearly defined set of steps or phases with dependencies on how to get there. Once that is defined at that point, I say now you are ready to start evaluating the business intelligence tools because now you have a very clear set of steps and a clear set of requirements, and a clear set of metrics. Now you take those, and you map them against BI vendors’ capabilities, and that’s how you evaluate them.