A holistic view of second generation BI
Houston, we have a problem
Comment I wrote a series of articles about second generation BI (which I refuse to refer to as Business Intelligence twodotoh) about six months ago. Why then I am returning to it? Because I think it is time to take a step back from the BI market and view it more holistically before we even consider what a next generation environment might look like.
The simple problem is this: there are too many BI tools (taking BI to have the broadest possible meaning) to get a clear picture of the market. There is OLAP, reporting, data mining, text mining, ad hoc query, report mining (a la DataWatch), report reuse (see my recent article about 90 Degree Software) and web mining, plus there are dashboards, analytical applications and the sort of complex analytics for which data warehouse appliances are suitable.
Then there is operational intelligence, process intelligence, BAM (business activity monitoring) and operational BI (if that is different from operational intelligence and/or process intelligence), which may or may not be event driven and may be in real-time, near real-time or right time.
And then there is the added complication of performance management and whether that is within the domain of BI or not; there is a spreadsheet management, which is a fraught enough subject in its own right; and I haven't even touched on the ways that you can implement OLAP or the whole subject of data warehousing and data marts, or the need to support new types of devices, technologies such as mash-ups, the requirement to support dynamic chart updating and so on and so forth.
Given that this is the mess that BI is in, I think that there needs to be some clear holistic thinking about how all of this can be structured and put together in some reasonably simple way, and that this is as much a part of second generation BI as broader and wider functionality that appeals to more people throughout the organisation. Indeed, arguably the provision of a platform that underpins these various technologies is what will enable what HP refers to as ubiquitous BI.
However, what we have today is, in effect, a siloed approach to BI: companies like SAS, Business Objects and Cognos do some of the things mentioned but not many others; the application vendors like Oracle and SAP do others; IBM has a different subset; and then there are various pure plays that are active in various sectors of the market. But the different solutions from different companies often do not work well with each other or only to a limited extent.
We all know that silos are not a good idea. It is, for example, gratifying that all, more or less, all of the vendors of MDM (master data management) recognise the dangers of siloism but this doesn't seem to apply in the BI space and I think it should.
To conclude, I don't think I have an answer for this: I am merely pointing out the problem and suggesting that it needs resolving. It needs an architecture if you like. I would welcome readers' comments and suggestions as to what this might look like.
Copyright © 2007, IT-Analysis.com
Sponsored: Hyper-scale data management