Holistic vs Incremental approaches to BI

There is a strong link here to my Vision vs Pragmatism article. In this I argued that Vision and Pragmatism are both essential for the success of any project, be that related to change, to IT, and certainly when using IT to drive change. Unsurprisingly, similar comments apply to whether a holistic or incremental approach to BI is the superior route. However, in this case, I will come down more firmly on the side of one of the options.
The benefits of an incremental approach

Of course the secret of the success of many projects is their incremental nature. Incremental deliveries, particularly those early on in a project, enable you to do a number of things, including: –

  1. Proving that business value can be added the work that you are doing
  2. Showing tangible evidence of progress
  3. Demonstrating that the project team is responsive to business priorities
  4. Chopping up funding into more digestible parts
  5. Providing early exposure to change management issues; allowing time to learn from mistakes when still operating at on a smaller scale

Overall incremental work can enhance the credibility of a project team and thereby made it easier to secure senior management support. Such work is indispensable to any project.
How does the sum of the parts measure up?

However there is a point to be made here in favour of a holistic approach which goes beyond my previous preference for always having an overarching vision. This is something that is specific to business intelligence and relates to the nature of information delivery. In a nutshell the sum of several incremental BI developments may be considerably less than the whole if each is not part of an overall strategy.

BI is about having the information necessary to run the business. However, it is also about how that information is delivered and how internally consistent it is. Often BI projects aim to address a fragmentation of existing reporting systems that leads to confusion amongst users and even a general distrust of figures. It is entirely possible to perpetuate this situation, simply replacing older reporting technology with shiny new ones. Each of these new systems may be easier to use that its predecessor and offer significantly greater access to information, but the fragmented nature of information provision will not have been addressed; it may even have been made worse.
A single platform

The ideal for a BI solution is to have a single platform which supports all pertinent reporting needs. There will undoubtedly be different segments of this, tailored to different groups of users, but these should use subsets of the same dimensions and measures and the same reporting and analysis tools should be used. Adhering to these precepts means that when users of one part of the system need to employ another part, they are not taking a step into terra incognita, but instead are familiar with their surroundings and get the sense that the same logic pervades all of the system.

On a practical level, this approach minimises costs due to software licenses and simplifies your technical architecture, again keeping a lid on expenditure. Fewer people are also needed to both build and maintain a single, central system than many divergent ones. Just as importantly, a single-platform approach means that training becomes focussed on business issues rather than the functionality of a different reporting suites. My experience suggests that, after an initial investment in thorough training for users, introduction of new reporting capabilities can be very smooth and efficient in such a set-up.

Of course developing good BI takes time and effort. Getting to the eventual ideal state that I have described above will undoubtedly take some time (in my most recent BI project it took five years to fully realise). This means that there is no real alternative to the incremental approach that I described at the beginning. However, taking a more holistic approach ensures that your incremental deliveries are aligned with both each other and overall business needs. It also means that with each incremental release there is a related reduction in fragmentation. This is the difference between slowly unveiling a large, coherent edifice and revealing several separate sculptures one at a time.
The link with cultural transformation

In particular if an aim of your BI project is to transform how users behave (of course this should be a central aim of any BI project, what else is BI for?), then this is going to be most easily achieved with a holistic approach where each phase builds on the success of the previous ones. In this scenario, each incremental delivery can be seen more as extending the remit of your BI system to a new area, rather than adding on a new module. Phase N+1 always reinforces the messages from Phases 1 to N. Each step reduces fragmentation, increases consistency and further improves decision-making. This is the best way to make sure that your BI efforts exceed the sum of their parts, rather than falling short of them. Such a rigorous approach is also the best way to ensure that you meet your cultural transformation objectives.