Disrupting BI Part 2: A Solid Data Architecture Empowers Data Visualization
Data visualization provides instant gratification to an analytic-starved user community, much like a kid in a candy store. Beware of the analytic lows to follow, much like the drop in energy after consuming too much candy. Almost any user of a data visualization tool can rapidly gain deep insights into their data by quickly visualizing it. With the desktop tool and some data (either query results or spreadsheets), an end user can start benefiting from some analytics in minutes. Armed with this capability, end users have grown impatient with the expectations that come with traditional business intelligence (BI) projects and implementations.
Understanding BI Complexity
Traditional BI projects take longer because the tools have more moving parts and complexities, such as semantic layers and report creation tools, which require more skills and effort to deliver results. Although the additional moving parts provide tremendous value, they also require careful forethought, planning, and development to effectively deliver. Thus, no end user can deliver traditional BI on his or her own; it must be a coordinated effort with skilled technical resources. I can understand the excitement many end users have now that they can control their own BI destiny.
Establishing Sound Data
Building data visualizations provides tremendous value more rapidly than we’ve seen before with traditional tools, but beware of jumping into it too quickly. A successful long-term data visualization approach still requires a strategy, architecture, and standards. The common issue that I’ve seen with clients who have started visualizing their data is that a great number of visualizations quickly become unmanageable, which I refer to as data visualization proliferation. The symptoms of data visualization proliferation include data disorientation, data redundancy, and data inconsistency, all of which can be solved with a sound data architecture as the foundation to the visualizations. Additionally, a sound data architecture adds value and enables deeper analytic insights, such as solving complex data problems, overcoming data integration limitations, and improving performance.
Using Data Marts
Only a shared, centralized data foundation, such as a data mart, can help control the inevitable data visualization proliferation. With the proliferation of many visualization documents, users may need to search through the many documents and may end up noticing the same information, such as customer information, repeated in many different documents, dashboards, and reports. This results in data redundancy and causes a degree of disorientation, not knowing which document, analysis, or component contains the needed information. And worst of all, the proliferation can also result in data inconsistency, where each document may calculate a metric such as sales, but since it could be calculated several times over, it’s likely each calculation will be slightly different with slightly different answers. Instead of producing data visualization documents from spreadsheets and queries maintained on users’ desktops, there is great benefit to using a shared data layer, such as a data mart, to target the problem. A key concept of data marts is that they are modeled after a business process, and are not just designed to answer the question of the moment, so they will be resilient to the changing needs of data visualizations. Properly modeling a business process in a data mart naturally eliminates data redundancy, inconsistency, and disorientation. The concept of modeling after a business process should also be applied to the data visualization document, which is a natural extension of the data mart, making it a perfect foundation for data visualization.
Visualizing Your Insights
Only a shared, centralized data foundation, such as a data mart, can empower data visualization tools to solve complex data problems. On their own merits, data visualization tools cannot maintain and display historical reporting capabilities, such as year-over-year reporting by today’s sales hierarchy, or how the data may have looked a year ago. Data marts are ideal to handle these types of complex data issues, such as tracking changes as they happen. Although some data visualization tools can integrate small amounts of data from different sources, and usually up to two at a time, they fall flat compared with the capabilities available in a solid data architecture, like traditional ETL tools. A data mart can integrate data from many disparate sources and then present a single view to the visualizations, thus eliminating integration challenges that data visualization tools sometimes face. When the visualization tool no longer needs to concern itself with data integration and data complexities, the visuals will perform at their best while also offering even deeper insights.
Ready to get started? Contact us to learn how you can optimize your data architecture and your BI implementations.