To BI or not to BI?

With organisations facing persistent demands to prioritise spending, Wu concedes flagging global economies represent the single biggest barrier to organisational success in building enterprise analytic capability. Those keen to pursue such solutions therefore need first to understand thoroughly the true business value of the analytic applications available. The paucity of case studies makes this particularly challenging, he says.

Wu cites a January 2003 IDC study, The Financial Impact of Business Analytics, Some Key Findings, which surveyed some 34 organisations to find an average five-year ROI of 434 per cent. But he warns enterprise analytics cannot be achieved without executive sponsorship and strong business definitions.

"You have to understand the core business rules and definitions," Wu says. "For example, in one of our [client] companies - a large conglomerate that had 12 different business units - each month the corporate controller would call all the business unit controllers together and they would sit down in a room, and everyone would bring an income statement. The funny thing is if you added up the revenue for each of the business units, it was greater than that of the company as a whole. Now how can that be? You have a bunch of chartered accountants all sitting in one room, yet the total was wrong."

Wu says it took Base Consulting two weeks to understand that everyone had a different definition of revenue. To some revenue meant gross sales. To others it was gross sales minus discounts; or gross sales minus discounts minus commissions; or gross sales minus commissions minus discounts minus returns. In the end his company had to rationalise six or seven different definitions of revenue into one - a new definition called gross revenue, net revenue and commissionable revenue.

"So we were effectively creating a new language or a new definition - metadata - for the company. You have to have that before you can go off and build a solution," Wu says. "So understanding what the business requirements are, the business definition, is really step number one."

Step two is to have a detailed technical understanding of the location and nature of data, and then develop an environment - an operational data centre store, data warehouse or data mart - that allows you to bring together all the data, cleanse it, transform it and put it in a format suitable for analysis. Wu says it is important to deploy the various design techniques of the database to assist with rapid extraction. Without this step organisations are likely to build solutions which might be technically elegant, but will prove absolutely worthless to the end user because of the inability to answer their questions.

Next you need to deploy a business intelligence tool that will allow you to extract the data easily, whether via ad hoc query, data mining or data visualisation. Wu says the market for the off-the-shelf products is now reasonably mature. Of particular interest are analytic applications - effectively packaged data mart solutions - which vendors like PeopleSoft, SAP and Oracle have been selling for the past three or four years.

Wu says these tightly coupled analytic applications are particularly useful because they come with pre-defined business rules, a database schema, data extraction tools and tools to deliver the data to the user.

Organisations can also turn to extraction translation tool vendors, business intelligence tool vendors and a number of third-party vendors selling their own solutions. But he warns CIOs there are a number of issues to be addressed before they opt for a packaged solution.