Design Tip #153 Three Critical Components for Successful Self-Service BI
March 4, 2013
© Kimball Group. All rights reserved.
The business intelligence industry has been using the phrase self-service business intelligence (BI) for several years. Self-service BI means enabling the business user community to create their own reports and analyses from scratch. Self-service business intelligence is nothing new. The Kimball Method has focused on delivering ad hoc access as an integrated – even integral – component of the DW/BI system for over 20 years. But self-service BI is a good name for an old concept, and one that may capture the attention of your business user community. Old name or new, let’s review what you need to have in place to support your business users’ information self-service. These three components are the foundation of a self-service BI environment: a solid dimensional model, a good user support system, and an effective BI tool.
Solid Dimensional Data Model
The first key to success is a dimensional data model. The benefits of dimensional models are widely known, and documented in all the Kimball Group books and throughout our website. In brief, the dimensional model delivers:
- Simplified structure that business users have some hope of understanding
- Simplified structure that works seamlessly with the best self-service BI tools
- Consistent attribute change management
- Excellent query performance, which comes from several directions:
- Fewer table joins because decodes and hierarchical structures are collapsed into a simple flat dimension table
- Simple table joins because an efficient single column surrogate key always joins facts and dimensions
- Database engine optimizations that recognize dimensional structures
- Where possible, pre-calculation of metrics or attributes during the ETL process rather than at query time
Good User Support System
Many of the organizations I’ve spoken with expect that self-service BI means they can reduce the size of IT staff who develop reports. That may be true. You may have fewer report writers in IT, but you’ll almost certainly replace them (or change their duties) with folks tasked with a broader set of service bureau responsibilities.
Effective self-service or ad hoc use of your organization’s data requires some services. The provision of these services should be coordinated centrally, even if the BI “front room service bureau” is staffed and funded locally. The services that are absolutely necessary for successful self-service BI include:
- Documentation and metadata. In order for business users to successfully use the tools and data you provide, they need to know what data elements mean, where they came from, how they’re organized, and what to watch out for. In the Kimball group, we are fans of having the business users actually see a simple representation of the dimensional model, ideally in the BI tool. After all, many of them participated in the collaborative design sessions, right? In this way, all three components of self-service BI come together.
- Metadata delivery. Developing and maintaining the metadata and descriptions are only part of the battle. You also need to make them available to the user community. The descriptive metadata is the most important, and if possible, should be exposed to users from within their BI tool.
- Training. It would be nice if your data model, documentation, and tools were so fabulous that your smart users could walk up to them and be effective right away. It just doesn’t work that way; you need to teach people how to use the environment safely and effectively. Users need to learn how to use the data as much as how to use the tool.
- Assistance. The dream of self-service BI is no assistance required. Realistically, there are always going to be questions, queries, and analyses that your users will need help with. Users can and will help each other, but it’s most effective to have a few people on the BI team to help with the really tough problems.
- Rich standard reports. The ability to “lightly customize” a standard report by choosing parameters is enough self-service for many users.
Effective Self-Service BI Query/Analysis Tool
Of course, you need a tool or tools for self-service BI. These are different tools than the standard report development and delivery software, although they may be provided by the same vendor. There are two main kinds of self-service BI tools, which for lack of better names, I’ll call conventional and visualization.
Conventional tools include the flagship offerings by many vendors, including from the venerable SAP (Business Objects), IBM (Cognos), Microstrategy, and many others. These tools are venerable because they work. They provide a semantic layer that enables users to construct queries by dragging and dropping. Once the data set is defined, users have a variety of options to construct a report or export the data to Excel. Most of the vendors provide a version of their tool with a simplified user interface for report layout, aimed at business users who want to rapidly prototype rather than report developers who aim to be pixel-perfect. IT organizations appreciate how easy it is to turn a business user’s analysis built in a conventional tool into a standard report.
Visualization tools are newer, but they’ve still been around for nearly a decade. Examples in this category include IBM (Cognos Insight), Tableau and QlikView. These tools tend to have the words visual or insight (or both) in their names. The visualization tools provide a more bounded or guided user experience. This guided query experience makes it easier to color inside the lines, but harder to break out of the box. Of course, the visualization tools really shine when it comes to analysis, particularly graphical representation of data. Visualization tools retain a close link between display and the underlying query, so it’s really easy for users to drill up, down, and across, and experiment with different presentation types such as spatial maps, animated time series, and heatmaps.
In my experience, the visualization tools do a better job of keeping the users inside the BI tool environment. With the conventional tools, there’s a strong tendency to construct the query in the tool, then export to Excel for analysis and display. I love Excel as much as the next person, but it can be very difficult to reverse engineer what the analyst has done, if we want to productionalize that analysis.
Self-Service BI: It’s Not Just a Tool
The BI tool vendors imply that if you buy their product, your business community will reap the benefits of self-service BI. There are many great BI tools out there; some of them are very fun to use, but don’t kid yourself (or let the vendors kid you). There’s a bunch of work that has to get done before your business users can leverage this fabulous capability. Don’t let the effort stand in your way, but be realistic about how much work there is, and how much time and resources it’ll take to deliver great self-service BI to your user community.