Intelligent Enterprise

Better Insight for Business Decisions

RSS
Webcasts
Digital Library
Subscribe
Home

Web Exclusive

April 3, 2002

Paving the Path to Epiphany

An off-road romp through the wilderness of data has its own rewards, yet most users of analytic applications need a paved path that leads straight to an analytic epiphany

by Philip Russom

Data analysis — especially in the long-standing best practices of online analytic processing (OLAP) — can be a somewhat rambling discovery mission that avoids the beaten path. Imagine the business analyst — driving toward a new business insight — who throws the map out the window and steers off the pavement to ford streams of information and climb steep aggregations of data. Negotiating the uncharted trail requires a bazillion ad hoc queries and interactive manipulations of data sets, all necessary in the quest to see what's over the next hill, around the next bend.

Well, that's a colorful description of using a generic analytic tool as if it were a sport utility vehicle (SUV). But the best practices for analytic applications are somewhat different in nature. First, most analytic application users are not business analysts, so they are neither trained nor interested in the SUV approach to data discovery. Second, most want to focus on their well-defined business responsibilities, using an analytic application that presents a short list of predefined data structures representing those responsibilities. Instead of a trail-blazing romp on dusty back roads, these users prefer a smoothly paved path that points (as directly as possible) to the appropriate business epiphany.

The "paved-path-to-epiphany" requirement among business users is being addressed by a fairly new best practice in analytic application design that seeks to record and automate certain business processes as linear analytic procedures.

Emerging Best Practice — Analytic Procedures

Some business intelligence (BI) professionals may consider the term "analytic procedure" oxymoronic. After all, how can analysis be linear like a procedure? Well, many users conduct linear analytic tasks all the time, whether we notice or not.

For instance, if you look over the shoulders of report consumers as they do their jobs on a day-to-day basis, they regularly look at the same values from the same list of reports in the same order. A simple analytic procedure can help these users by recording the succession for efficient playback, akin to the so-called "macro" found in personal productivity software. As another example, some users start a repetitive task by viewing a particular report. One or more values seen in that report determine which reports to view or which analytic steps to take next. Note that the user's recurring task involves a rudimentary form of workflow. This kind of user needs a kind of analytic procedure that encapsulates the decision making done at each branch of the workflow.

Software Support for Analytic Procedures

Recognizing the demand among users, a few BI software vendors have recently introduced support for various types of analytic procedures. For example, Informatica Applications version 5 includes tools and functions for creating both "analytic paths" (straight line, no workflow) and "analytic workflows" (with branching like a decision tree). Similar analytic procedures are seen in the "scenario workflows" of Business Objects SA's BusinessObjects, the "cascading scorecards" of CorVu Corp.'s CorManage, the "guided analysis" of Crystal Decisions' Crystal Reports, the "task flows" of PowerMarket Inc.'s Supplier Performance Management, and the "analysis storyboards" (for root cause analysis) in SageTree Inc.'s Supply Chain Performance Suite.

Vendors aside, some users create their own versions of analytic procedures. For instance, those who receive reports or analytic applications via a portal may simply list reports in the portal's personalized user interface in the order they prefer to read them.

Assessment — The Benefits of Analytic Procedures

There are good reasons why analytic procedures are catching on as a best practice for paving the path to epiphany in analytic applications.

Improves productivity. As a kind of playback mechanism for recurring tasks, an analytic procedure can make the execution of such tasks consistent, predictable, and efficient.

Enables collaboration. By recording the successful work of an experienced knowledge worker, expertise can be captured and shared with other knowledge workers. This is important as more users from lower in the corporate "org chart" come to analytic applications, users who have little or no experience interpreting reports or analytic data sets.

Encapsulates domain expertise. One of the leading challenges to creating an application (whether analytic or otherwise) is the "knowledge transfer" from experienced business users to the application designer. Analytic procedures facilitate knowledge transfer by encapsulating the expertise and linear processes of a business domain.

Provides structure. The most common failing of analytic applications (whether built by a vendor for resale or built in-house by an IT department) is a lack of structure. The linear structure of analytic procedures helps avoid the "bucket of reports" syndrome that plagues many analytic applications. In fact, some leading-edge analytic applications consist mostly of analytic procedures, instead of a body of unorganized reports.

Processes performance metrics. Many analytic applications today implement some form of business performance management (BPM), where the performance of business entities are represented by metrics and key performance indicators (KPIs). BPM methodology requires managers to monitor metrics and KPIs daily, to ensure that the represented business entities are performing according to the business plan. A manager typically looks at one metric after another, deciding which action to take based on the value of individual metrics. In other words, workflow is inherent to BPM methods. With that in mind, you can see that analytic procedures that support workflow can contribute greatly to the automation of BPM analytic best practices.

Philip Russom, Ph.D. [www.philiprussom.com] is an independent industry analyst and consultant based in Waltham, Mass.

Related Articles at IntelligentEnterprise.com:

From Back to Front, March 8, 2002. Informatica expands from ETL back end to dashboard front end.

Nonanalytic Nonapplications, Feb. 1, 2002. Most analytic applications lack the interactivity and procedural structure to be called such.

Beyond the Bucket of Reports, Oct. 16,2001. Too many analytic applications are merely unstructured buckets of reports. But what's of value beyond the bucket, and how do you get there?