Even as it flourishes, self-service business intelligence continues to draw abundant criticism. Though much of it revolves around how the software is marketed and improperly implemented, some implicates the technology itself.
Critics dismiss self-service business intelligence (BI) as a blunt instrument incapable of surfacing the detail necessary for informed decision making, and some worry that users will make business decisions based on such insufficiently detailed data insights. They also say business users are simply not skilled enough to be trusted with it. In Five Drawbacks to Self-Service BI, Matthew Gierc says non-data scientists are prone to confirmation bias and therefore likely to spread misinformation. Even Nimrod Avissar, who admits to liking self-service BI, depicts a scene of “data anarchy” with unsupervised self-service BI users churning out disparate data models (and taking twice as long as IT to do it).
These failings have little to do with business users. Rather, they reflect shortcomings either in the self-service BI software or how it’s been institutionalized. As Gartner research director Carlie J. Idoine concedes, “If data and analytics leaders simply provide access to data and tools alone, self-service initiatives often don’t work out well.”
Though companies certainly need to take a centralized and intentional approach to data governance, these efforts will amount to little if the self-service BI solutions they implement are inadequate. There’s a fundamental misunderstanding among consumers about what it is and what it must be capable of in order to provide that technological foundation data-driven enterprises need to be successful.
There’s a fundamental misunderstanding among consumers about what self-service business intelligence is and what it must be capable of.
The Business Application Research Center (BARC) defines self-service BI tasks as “those that business users carry out themselves instead of passing them on to IT for fulfillment.” By this logic, it should include not just ad hoc report creation tools but any and all tools that enable business users to independently meet their data needs when they arise.
The following list of essential self-service BI capabilities, therefore, goes beyond report creation tools to include the more traditional BI practices of supplying and distributing premade assets.
1. Operational reports. To provide business users with actionably granular detail, a self-service BI solution must offer operational reporting in addition to data visualization and dashboard tools. Visual summaries, even when paired with drill-downs, are insufficiently granular for day-to-day operations. Robust tabular reports that can be manipulated at the record level are a must.
2. Parameterized reports. These canned or prebuilt reports provide customizable settings such as sorts, filters, parameters, conditional formatting, and styling options. Business users cannot edit these reports’ definitions, but they can manipulate their settings to affect the output.
Parameterized reports are critical to self-service BI, even though business users are unlikely to build them on their own. By including parameters in canned reports, IT can provide business users with more flexible report libraries. Users don’t need report-writing or editing skills to select a value from a drop-down menu or fill in a field, nor do they need help from IT. In this way, parameterized reports broaden the means by which business users may help themselves in the moment.
3. Duplicable canned reports. Business users with some BI training will appreciate being able to duplicate canned reports (normally read-only) and edit the copy to their specifications. Having a foundation to work from is easier than starting a report from scratch, and the process familiarizes users with the ad hoc reporting tools available.
4. Scheduling and bursting. Frequently used reports may be set to execute at regular intervals, the output either saved to a repository or sent directly to stakeholders. Scheduling is a convenient way for IT analysts, data analysts, and business users to meet teams’ recurring data needs. Bursting saves even more time by splitting or filtering a single report according to the recipient’s credentials. Like canned and parameterized reports, scheduled reports reduce the number of reports business users must build themselves.
5. Dynamic data modeling. Data modeling is one of the more complex aspects of authoring a BI report, and vendors would do well to offload the bulk of this process onto IT to set up ahead of time. Some self-service BI solutions supply users with a library of premade models, but this, too, presents challenges to novice users. Dynamic data modeling, on the other hand, uses administrative settings to build models automatically as users select data fields, allowing power users to make tweaks.
6. Report authorship and setup signatures. Critics of self-service business intelligence are concerned that those consuming a business-user-generated report will mistake it for an “official” report authored by IT. It makes sense that BI reports, like any other publication, should include authorship information as well as information regarding the report’s design (e.g., what filters have been applied). Vendors should make it easy to add such information to a report so that enterprises may adopt this accountability standard.
7. Ad hoc reporting. Business users should not only be able to build their own reports — they should also have access to a range of tools designed to accommodate a spectrum of skill levels. If there is only one report designer available, it should offer a range of “modes” from novice to expert. This helps prevent users from becoming overwhelmed by UI controls and allows them to learn at their own pace.
A Final Word
With this set of capabilities, enterprise IT departments can, as TDWI research director David Stodder advocated in a 2017 webinar, successfully “enable” self-service BI platforms by facilitating business user independence.
Of course, critics will point out that IT’s very involvement renders business users dependent on IT, and that’s technically true. To borrow Avissar’s analogy, enterprise self-service BI should operate like a buffet. You should have your fully prepared chicken cordon bleu (canned reports), your customizable pasta with choice of sauce (parameterized and duplicable reports), and, of course, a salad bar that allows you to do whatever you like with the ingredients provided (ad hoc reports). Everything is carefully prepared by the kitchen (IT), whose main objective is to ensure that customers can help themselves without having to wait.
Originally published with TDWI.