BI vendors talk about how businesses can improve performance by analyzing data from multiple sources, but when you talk to IT people they say it's hard to get even two clean, reliable sources of data for real-time analytics. Why is BI still so hard to do?

The fundamental problem rests in the following: There are on average, in any sizeable organization, six systems that deal with customers -- you've got tools for selling, order processing, billing, maintenance, marketing. These systems were traditionally built independently and each has their own way of describing a customer. So this data is to some degree incorrect, and reconciling which of these systems is the primary, accurate representation of the business is very hard. So the idea of master data management, which in essence is not a tool issue but a political issue -- figuring out who is the real owner of the information -- is where the real problem arises, and it's different for every customer. So it's hard to come up with a solution which says, out of the box, here's what you do. You need someone to do the analysis to understand what the sources of data are, how good the systems are that produced that data, which of the systems holds the real canonical version of that information, and then align your systems to that one. Because you can clean up your data once, but if the underlying systems that created the data remain the same, a nanosecond later you need to clean it all again.

Oracle CEO Larry Ellison has said you can overcome this problem by putting all your data in an Oracle database, but obviously that won't work for all customers. So are we just left to struggle with this problem?

I think we are. It's good for our business by the way.

The interesting thing is that many customers can't agree among themselves. I had dinner last night with four CIOs that represented four divisions at a large customer. They are to some degree different businesses, but at the end of the day they deal with the same end consumer, even though the consumer has four different relationships with the company. They would like to get a single view of the consumer, and there isn't a power in their organization who has the right to decide which is the right view. So they struggle and they compromise and they build subsets of those views that they try to reconcile, and even that's hard to do.

Having good tools is only one part of doing BI. There is also governance and how a company should architect its systems. How much of a role should BI vendors play in telling customers how to design their IT?

A very large role, because BI is wholly dependent on access to information. If the information is not available or architected poorly, or incorrect for that matter, which is 90 per cent of the problem, the BI doesn't deliver any value. So the ability to create a trusted broad universe of information is a fundamental success factor for the customer, which is why we went after the EAI market ourselves, because we didn't want to depend on other players to do it for us.

So what advice do you give them about the ways to architect their system?

We try to deliver reference examples of other customers who have done this successfully, we work with our partners who make their living designing data environments, partners like BackOffice Associates, a premium MDM solutions provider, to make sure their tools are mapped to ours in a way that makes the solution easy. There are some fairly intellectually intensive techniques that are brought to this problem because it's not a tools issue, it's a design issue.