IT Week: Some critics have argued that business intelligence [BI] tools have been over-hyped. Is this criticism justified?

Andy Honess: There has been a history over the past 10 years of companies being persuaded to undertake big-bang technology investments without realising it will take a lot of time to get the value out of that investment. You can see that being replicated in BI at the moment, with the market growing quickly but people not really getting delivery on the software companies' original promises.

How significant is the problem?

If you look at the DM Review study from the end of 2004, it stated that typical BI projects took about 17 months to implement, cost £10m and had a 35 percent success rate. Those are appalling statistics and it means if you are in the mid-market and have a small IT budget you are not going to invest in that technology, which is why most BI is found at large corporates and if you look at their experiences on the whole they are a mixed bag.

Why is BI proving so difficult for firms?

You have to look at the 35 percent success rate and ask "why is that happening?" Well, the infrastructure the reporting tools are trying to act on is very complex, with different technologies with multiple languages and multiple platforms that have not been integrated as well as they should have been. Then there is the cost that arises from the traditional way of doing BI and the pain-chain that results from hooking up an ETL [extract, transform and load] tool to a high-maintenance data warehouse, to a cube or universe technology, and then to a dashboard or reporting tool at the front-end. This integration results in a lot of cost and complexity and also means you can only analyse a snippet of the whole corporate world.

How so?

Your ETL tool pulls data from different sources into a warehouse and the [Olap, online analytical processing] cube takes a snippet from that warehouse to answer the users' question. But even when you use larger cubes inevitably you get to the edge of the cube. For example, you may ask "how many widgets did we sell in December by geography?" But the next question may be "how many sales people did we have?" Because that dimension wasn't included in the original cube you have to go all the way back down the chain to identify that data and rebuild the cube. That can take weeks. As a result the BI tool inevitably ends up on the shelf and the frustrated executive bypasses the BI technology and goes back to making the decision based on an Excel spreadsheet.

continued on next page...