In-Depth
How to Avoid the Common Traps in BI Implementations
Industry experts identify the pratfalls which can sink even the most solid BI project
How much do you know about successful BI implementations? Just as important -- do you know what doesn't work?
Industry watcher Gartner Inc. recently compiled a list of BI flaws it says organizations should safeguard against. Although Gartner's list includes a requisite straw man or two -- it's rare for an IT shop to deploy BI tools without first assessing competitive offerings -- it's also a useful primer for what not to do in business intelligence.
According to Gartner, unsuccessful BI projects typically display one or more of the common flaws. For the most part, Gartner argues, these shortcomings take the form of the familiar people and process bugaboos that bedevil all technology implementation efforts. "Despite years of investing in BI, many IT organizations have difficulty connecting BI with the business, and to get business users fully involved and out of the 'Excel culture,'" said Bill Hostmann, vice president and distinguished analyst at Gartner, in a release.
"Just one common mistake can destroy a BI program, and there is far more risk in nontechnology issues -- [i.e.,] sponsorship, politics, data quality, and so on -- than in deploying the infrastructure, tools, and applications that support BI. Forewarned is forearmed, so organizations must understand the common flaws that undermine BI projects and prepare an approach to avoid or minimize them."
The first and most obvious flaw, according to Gartner, is what might be called a Field of Dreams-like expectation among BI architects: just because you build it doesn't necessarily mean users will adopt it. In some cases, Gartner maintains, organizations can have more BI pros supporting an implementation than consumers actually using it. Gartner's recommendation is a familiar one: ensure adequate business buy-in, chiefly by giving business stakeholders meaningful representation in the project planning and management process. Gartner also prescribes establishing a business intelligence competency center (BICC) to champion BI adoption on the business side.
Gartner also takes aim at the proliferation of multiple, insecure, and frequently contradictory spreadsheets, thanks to a built-in bias in favor of spreadsheets among managers, analysts, and other power users. If not properly addressed, "Excel Bias" can undermine any well-meaning BI implementation.
There's a sense, too, in which post-implementation Excel Bias actually derives from the build-it-and-they-will-come illusion: BI practitioners assume that just because they've given spreadsheet users suitable alternatives -- including better-than-spreadsheet solutions that maintain an Excel-like look and feel (and in many cases actually support the use of Excel) even as they guarantee governance, auditability, and a single version of the truth -- users will adopt them.
The in-the-trenches bias in favor of Excel is a strong one, Gartner counsels, and BI pros will need to tackle it directly: once again, by soliciting sponsorships from business stakeholders who champion transparency and have the necessary clout to drive adoption.
Intractable Data Quality Issues
Data quality is that most stubborn of data warehousing (DW) and BI problems: most (if not all shops) know they have potential data quality issues, and many have taken steps to mitigate these, but data quality problems persist. Earlier this year, for example, Daniel Teachey, director of media relations for data quality provider (and SAS Institute Inc. subsidiary) DataFlux, told BI This Week that his company still sees a lot of reactivity among customers.
"We're called in to put out a lot of fires," Teachey confirmed. Things have improved, Teachey allowed: customers -- spurred by governance requirements -- are getting more proactive (after five years or more of temporization) and beginning to adopt best-of-breed DQ technology. Mainstream vendors -- i.e., IBM Corp. (which picked up DQ with its acquisition of Ascential Software Corp.), Microsoft Corp., and Oracle Corp. -- are also pushing into the DQ segment.
Not surprisingly, Gartner, too, highlights poor or insufficient data quality as one of its BI fatal flaws. It couches the issue in a highly tendentious manner -- suggesting that organizations are oblivious to potential data quality issues -- but correctly identifies the bulwarks to adoption posed by irrelevant, incomplete, or questionable (and questionably-sourced) data.
The solution? Shops should establish either process controls or deploy automated tools to ensure the accuracy of their data.
No large shop would deploy a BI tool without first comparing it against several competitors. Typically, shops make their selections on the basis of both feature/functionality comparisons and RFPs. However, the BI platform play -- which promises everything-and-the-kitchen-sink functionality in a single BI toolset -- complicates that proposition: where once an organization could select from competing best-of-breed OLAP, reporting, ETL, dashboarding, and other tools, they're increasingly turning to one-stop platform shops that offer perceived TCO advantages.
Gartner doesn't necessarily counsel against such an approach; it merely cautions shops against blindly adopting a platform on the bases of all-in-one-convenience or perceived TCO savings.
It's good counsel, as far as it goes, but Gartner also suggests that organizations are buying BI platforms without first shopping around -- i.e., without comparing them (on the basis of features/functionality, TCO, and integration/interoperability with existing assets) against competitive platforms from both mainstream vendors (IBM/Cognos, Oracle/Hyperion, SAP/Business Objects) and best-of-breed platform providers (SAS, MicroStrategy). Gartner says that's a bad idea. "Integration between the application vendor's ERP/data warehouse and BI offerings is not a compelling reason for ignoring alternatives, especially as many third-party BI platforms are as well integrated," it counsels.
There's a similar straw man lurking in Gartner's admonition against BI complacency -- that just because something is running smoothly doesn't mean it can't be improved. Successful BI implementations aren't fixed or static entities, according to Gartner: they are, instead, moving targets, dictated as much by technological improvement (e.g., new features, functionality, integration, or interoperability scenarios) as by feedback from users. In the latter case, Gartner points out, users are especially vocal in the first year after implementation, requesting changes to either improve their experiences or improve business processes.
Gartner's solution: "Organizations should therefore define a review process that manages obsolescence and replacement within the BI portfolio."
Outsourcing
Shops concerned about the high cost of IT operations -- or with concerns about their own implementation or management expertise -- often turn their attention to outsourcing an entire BI project -- which Gartner lists as its sixth flaw.
A monomaniacal focus on TCO or development staffing concerns usually produces inflexible, shoddily architected systems. Gartner advises shops to outsource only in the case of non-core competencies or business operations. In scenarios in which shops embrace outsourcing as a means to redress their own IT shortcomings, Gartner suggests an enterprise outsource temporarily while fleshing out one's internal IT know-how.
A similar monomania attends the use of dashboards, which (according to Gartner) companies increasingly view as ends in themselves. It's a flawed perspective, if only because a focus on dashboarding-for-the-sake-of-dashboarding results in -- surprise! -- dashboards that ultimately deliver very little value. They're data source connectivity is limited. In some cases, Gartner says, they're effectively silo-ed -- consuming data from one or only a handful of data sources.
Nor are they aligned with corporate objectives. Gartner's prescription is a surprisingly stopgap one: focus on gentrified reporting -- complete with advanced charting or visualization capabilities -- as you build out more robust (i.e., feature and connectivity-rich) dashboard implementations.
Finally, Gartner cites poor or non-existent master data management (MDM) practices and the lack of an overarching BI strategy as its last two flaws, respectively. In the latter case, Gartner says, shops should avoid tactical thinking -- focusing on BI projects that deliver largely immediate benefits -- and focus on the long (or strategic) view.
To help ensure a commitment to strategic BI, shops should create teams -- populated with representatives from both the business and the IT sides of the aisle -- to codify (and periodically revise) BI strategy documents. Such a team should nominally fall under the oversight of a BICC, it suggests.
"Simple departmental BI projects that pay an immediate return on investment can mean narrow projects that don't adapt to changing requirements and that hinder the creation of companywide BI strategies," said James Richardson, research director at Gartner, in a statement. "Business users must take a leadership role in the BI initiative -- only with their full engagement will investment in BI ever realize its potential."