Analytic applications have gone through a fairly dramatic transformation in recent years, as business intelligence (BI) software vendors have struggled to keep pace with new requirements for business performance management (BPM). In fact, performance-oriented enhancements to analytic apps — and the BI platforms that support them — unquestionably constitute the strongest trend currently wending its way through the world of BI. Instead of focusing on one product in this special product review, this column surveys some of the new or improved features and functions of analytic apps, as a way of charting this trend and revealing its influence on the emerging character of performance-oriented analytic apps.

M Equals "Management"?

BPM has been part of the BI world for several years in the form of the balanced scorecard. Numerous vendors offer analytic apps for the balanced scorecard, including Corvu Corp., PeopleSoft, and SAS Institute. But a successful balanced scorecard project requires hefty doses of process reengineering to achieve a procedural rigor that, incidentally, suits very few corporate cultures. Most need a looser application of business performance concepts that preserves measurement and monitoring tasks (so you know how selected business entities are performing), but forgoes rigid adherence to an overarching management strategy (which is usually so grandiose as to be impossible). Hence, the M in BPM comes closer in practice to measurement or monitoring than management.

This loosening of methodology has made BPM more conducive to automation via analytic apps. Conventional wisdom now defines an analytic app as based on a body of metrics, each metric representing the performance of a business entity. Metrics typically exist in a hierarchy, where metrics roll up into key performance indicators (KPIs), which may roll up into perspectives.

BI vendors have reacted to this new definition by building new analytic apps (or retrofitting old ones) atop prepackaged data marts, where each mart implements a hierarchy of metrics, KPIs, and perspectives. (Obviously, a data mart also includes other data structures, such as time series data and aggregates, which a metrics hierarchy cannot replace.) Each analytic app also includes a body of report objects that query its data mart. Metrics-oriented analytic apps (or BI platforms with explicit BPM support) are available from a broad range of vendors, including Actuate Corp., Brio Software Inc., Business Objects SA, Cognos Inc., Hyperion Solutions Corp., and Informatica Corp.

Extended Visibility

One of the hottest trends in 2002 concerns the new multidomain suites of packaged analytic apps a few vendors have rolled out. Each suite consists of multiple analytic apps, each focused on a different business domain. Most suites cover financials and sales, traditionally the most popular subjects for analytic apps. Other domains typically include operations, customer profitability, and miscellaneous supply chain issues.

Although each analytic app in a suite is built on top of a data mart, the marts integrate via shared dimensions or some equivalent. Hence, queries can access multiple federated marts, providing visibility into cross-functional business processes and their interdependencies. For instance, sales forecasting can be corroborated by information from financials and operations. Single-domain analytic apps can't support this breadth of visibility.

The multidomain suite is a new product type, but already hotly competed. The three leading offerings are Business Objects's BusinessObjects Analytics, Cognos's Analytic Applications, and Informatica's Informatica Applications.

Direct Line to the Source

You can see that a prepackaged, domain-specific data mart is the real meat of a vendor's analytic app. Also in the package are reports and user interface features that query the data mart. Overlooked until recently, however, is that a packaged analytic app should also include a component for extract, transform, and load (ETL). This component requires a tightly integrated ETL server, prebuilt mappings downstream into the data mart, prebuilt mappings upstream to predictable enterprise data sources (such as ERP, CRM, and other popular packaged applications), and a tool for mapping to other enterprise sources.

Cognos and Informatica set the bar by including DecisionStream and PowerCenter, respectively, as embedded ETL tools in their suites of cross-functional analytic apps. Business Objects corroborated the packaged ETL requirement (and succumbed to competitive pressure) by recently acquiring Acta Technology Inc. — its ETL tool ActaWorks will soon be embedded under Business Objects' suite of analytic apps. This trend looks new, but we could look back in time and argue that ETL-centric analytic app vendors (especially DecisionPoint Applications, Sagent Technology Inc., and SAS Institute) addressed this requirement long ago.

Relevance in a Box

Whether an analytic app comes from a vendor or is built in-house by IT, the knowledge transfer from business experts to analytic app designers continues to be the highest hurdle to clear in the race for quality and user relevance. Even so, the infusion of BPM has helped in this area by providing a guiding, metrics-oriented method that complements best practices for modeling business domains.