For the last decade (plus some), business intelligence has focused on assembly of information into structured repositories. And now, it may be a well structured meta-data repository with query capabilities on the fly – but regardless it requires a complete definition (and perhaps a separate physical storage component outside the system of record). Call them marts, cubes or warehouses this trend has produced 360 degree views, banded report writers, meta-data management, data appliances and all sorts of beneficial tools for the visualization of information. Path 1 in this case is the robust world of the data warehouse. It has also led to data lineage issues, processing windows, inconsistencies from the system of record and in some cases lengthy wait times for end users to get the access they need to data. Nonetheless, the concepts of shared dimensions, fact tables and the like are at clearly beneficial and will remain at the core of business intelligence for a while.
Path 2 is more along the lines of operational BI (BI 2.0 I suppose). Collecting information from the service stream and injecting metrics into the business process. Or at least providing the user with better visibility control on top of the dimensional and in many cases non-dimensional sources of information. We still have a lot of description going on but things are clearly more fluid as tools advance and far more specific in many cases as BI is pushed closer to or derived from the transaction environments themselves. For sake of argument we can throw the onDemand (model predefined and hosted) providers in this class as well as the attempts are less about defining everything and look to defining the right thing or analysis based on observation. Still, often a significant repository sits behind these tools. (I admit that I likely have not described or know the full essence of this path at this point but look forward to learning more.)
Paths 3 and 4 may be less obvious. They involve the mashup. So, on one side (path 3) you have composition after the fact. The dashboard made of widgets or portal page come to mind. The collection perhaps sharing a set of injected parameters or integration at the glass to provide a view into the information. In this case, you may be asking several core systems to collaborate or integrating from a system of record with a back-end mart view by executing tasks through a shared session element such as the user. These clearly allow more information to become transparent if executed correctly and reduce reliance on back end aggregates. I don’t know if these are slice and dice worthy implementations but it is an option in the quest for reduced time between data capture and visibility.
So what is path 4? Path 4 is composition and service-based as well but prior to visualization not after (or at the glass). This can bring the promise of integrated charts, graphs and visuals typically reserved for more traditional BI architectures forward into the mashup world. Indeed, some glue needs to wind through the data to empower the combination of disparate elements and services but the need for physical storage and full meta-data descriptions may not be required. Views that may be embeddable into business process applications much like BI 2.0 without the aggregations of predefined models.
What path to choose? Well as in most business scenarios it depends (and I will likely comment on what it depends on later). But it is nice to see the conversation becoming more open and services helping everyone redefine visibility as we speak.
