Nonanalytic NonapplicationsMost analytic applications lack the interactivity and procedural structure to be called suchContinued from Page 1 In well-organized analytic applications, users start with a summary view, which may be implemented as a dashboard, scorecard, or personalized portal. From here, they can drill down to analyze details. The user crosses a fuzzy line, moving from static reporting to interactive analysis without fretting over the difference. When an analytic application — of any type — supports this additional depth, it is truly interactive and analytic. Without drill down into an analytic layer, however, an application belongs to the realm of nonanalytic reporting, so the rubric "analytic application" is inappropriate. ANALYSIS CAN BE PROCEDURALNow that I've split hairs to separate reporting from analysis, it's time to vent my prejudices learned from exposure to procedural applications. My definition of application — analytic or otherwise — is that it encapsulates the expertise and automates the processes of a specific business domain. With operational and transactional applications, the business process is often reduced to a procedure, which software automates via a linear user interface. For instance, order-entry applications I've used have all been ruthlessly linear, presenting a series of forms in a fixed sequence, sometimes forcing you to fill out fields in a particular order. At the other extreme, many of us think of data analysis as a nonlinear discovery mission. Imagine a business analyst who relies on ad hoc queries and interactive manipulations of data sets to enable a free-form quest for insight. That's a good description of using a generic analytic tool, perhaps established by long-standing best practices for online analytic processing (OLAP). But a true analytic application is not that open-ended. An analytic application focuses on a specific business domain. The domain can be horizontal (such as sales, operations, and financials) or vertical (such as retail, manufacturing, and healthcare). An analyst can attain valuable insights from a nonlinear romp through the data of these domains. But — and here's my point — most business domains include best-practice processes that can be captured and automated as linear analytic procedures. I call such a procedure a guided analysis. It's my opinion that — for an analytic application to be an application — it should include guided analyses for the analytic processes of its business domain that can be automated as linear procedures. Otherwise it's not an analytic application. It's merely an analytic nonapplication or just another generic data analysis tool. But let's be clear on one point: There's a need for both guided analysis and ad hoc analytic discovery. After all, these capabilities are for different users and purposes. Guided analysis leads the majority of users to answers for the majority of their analytic questions. Yet, an analytic application should also accommodate power users who need answers to ad hoc questions and are able to find answers themselves. I've watched guided analysis evolve in small niches over the last few years. And I've gotten sneak previews of upcoming analytic applications that implement it. This evolution suggests to me that guided analysis will be a differentiating factor for a new generation of analytic applications. Hence, guided analysis shines some light into the future of analytic applications. It also demonstrates a way of making analytic applications more usable and palatable to the ever-broadening range of analytic users, many of whom — like me — bring with them their expectations and biases learned from procedural applications. Philip Russom, Ph.D. [www.PhilipRussom.com] is a Giga Research Director at Forrester Research Inc., where he provides advice to user organizations about business intelligence, data warehousing, and data integration. RESOURCESRelated Article at IntelligentEnterprise.com: "Beyond the Bucket of Reports," Oct. 16, 2001
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| |||||||||||||||||||||||||||||||
