Next generation BI
I was asked recently to consider what would be required for a next generation BI solution. Never one to shirk a challenge, here are some thoughts.
The key is at the user level. First, accessing data and reports needs to be simpler: you and I may be au fait with making selections from hierarchies but lots of people aren’t. Tree hierarchies are not intuitive, which is why you use Outlook as a filing system. What is needed is something more visual. Secondly, better visualisation would also be useful for presenting data.
However, accessing and viewing data is only the beginning. You would also like to be able to look at any relevant combination of information. And by relevant, I mean relevant to you regardless of whether it happens to have been developed that way by your IT department or software vendor. For example, suppose that you want to compare column A in report X with Column B in report Y: you would like a simple, intuitive way of telling the software that that is what you want and then have it go and do it. This raises two issues: the interface that allows you to do this and the technological underpinnings of such a solution. I’ll discuss the latter (which also includes the possibility that the data make come from different sources) in due course.
As far as the user interface is concerned this, again, needs to be visual. If it was just a question of two columns to be compared you could just have a facility to drag one onto the other. However, the data may be being depicted graphically so you would also need the ability to drag and drop an axis from one graph onto an axis in another graph or, conceivably, an axis onto a column (or row).
In fact this is still not enough: what if you want to compare multiple columns, rows or axes? What if you want to add calculations or Boolean expressions? In these sorts of cases, you need something more sophisticated, such as a workflow style palette, though maybe the bright visualisation boys can think of something even simpler for this sort of purpose.
Next, users will want integrated search, not just within the same interface but integrated with conventional queries. For example: “show me my top ten customers together with outstanding proposals related to those customers”. It is also likely, though this is probably further away, that we will want to see the integration of text mining (or, at least, more sophisticated text querying capabilities) with conventional query capabilities. Certainly, query facilities will need to extend to XML documents.
Then, there is a need for operational business intelligence and an understanding of processes. The former does not create any particular issues except when the intelligence you need is needed in real-time, and it is also in the case of real-time operations that you need to understand processes. For example, suppose that you are monitoring a trading desk within an investment bank. First, you need the software to be able to alert you to the fact that you are underperforming (say) given that this is 3.00 pm on a Thursday. That is, you need the software to recognise that current behaviour is anomalous. If it is, then you need to able to find out what part of the trading process is out of kilter, so the operational intelligence software needs to understand processes as well as just being able to compare current conditions with previous ones.
Finally, users will want to get easily from where they are today to where they want to be tomorrow. They will need facilities that enable them to easily migrate from their existing environment and, at least in the short term, the ability to reuse existing BI assets.
These are some of the things that users are likely to want to see in a next generation BI product. I do not suggest that it is a comprehensive list (performance is always an issue, you might want avant garde capabilities such as voice activated query, there are Amazonesque “if you like this then you might like that” facilities, and so on) but I do think that it captures the main causes of concern. I will consider the technical issues underlying these requirements in the next article in this series and then conclude with a third article suggesting how they might be met.