Bimodal Analytics Defined: Balancing Your Approach

Data discovery tools and techniques represent the largest change to traditional business intelligence in the last decade. Instead of being a replacement for enterprise reporting, however, data discovery is a new approach that augments the old methods. Understanding when to use each and how they work together are key elements in implementing a strategy for bimodal analytics.

The Past: Enterprise Analytics Only

Enterprise reporting driven from a data warehouse has been the focus of business intelligence for a long time. Powerful tools like IBM Cognos provide a simple yet powerful interface for creating dashboards and reports from well-crafted data sources. The capabilities of such reporting engines are immense, but they can take a long time to reflect changes in data sources or business needs. If a new data source arrives, or business users come up with new types of questions to ask, it can take weeks or months to design and implement changes.

Data discovery tools evolved to benefit users who needed more agility than enterprise reporting would allow. Software offerings like Domo, QlikView, and Tableau provide fast reporting and limited means to combine different data sources together. For complex data transformations, data could be processed externally (for example, in Excel) before being loaded into the tool. Most importantly, all of this could be done on a user’s desktop with no need for IT intervention or support. With users frustrated over the rigidity of data warehouses, it was easy to see data discovery as a “warehouse killer.”

The Present: In Search of Balance

Despite their many advantages, data discovery tools have their own issues. Because they are focused on rapid analysis and visualization, they are not designed for repeatable or maintainable data processes. A report that people expect on a regular basis needs to have its data refreshed – which often means manual processing of that data. If an analyst produced a report by taking an export from the warehouse, combining it with a new external data source in spreadsheet format, and then manually reconciling the two where certain categories didn’t line up – that would now be a process to be completed every month. The power of rapid analysis can easily get bogged down in copies of spreadsheets and manual refreshes.

There are other areas where data discovery tools were not as robust in the beginning: authentication, connecting data together, and highly formatted reporting are some examples. Many of the tools have improved in these areas, but their ability to do so is limited by their design. Flexible data discovery and enterprise-grade reporting turn out to be very different in terms of design goals, and it is difficult to create an architecture that achieves both.

Instead of replacing enterprise reporting, data discovery has become its own area of focus alongside enterprise reporting. Analysts and data scientists, along with anyone else looking for quick insights in a world of rapidly changing data, can use data discovery. When more robust and scalable analytics are needed, enterprise reporting is in place. Bimodal analytics means acknowledging that both types of analytics are important and understanding how to maintain a balance between them.

It is important to note that this pair of analytic approaches is not a “business vs. IT” situation. Business users will still be running warehouse reports. IT resources will help maintain and use data discovery tools. In fact, one of the biggest challenges of data discovery tools is that they still require an understanding of basic data manipulation techniques. Business users may be drawn to the ease of use of data discovery tools, but can get into trouble combining data sources together or exploring a complex data warehouse source.

Technical resources are vital at this point to making sure that the right data is being used to gain insight and perform rapid analysis. Just as an understanding of business use cases and needs is critical to designing a data warehouse, an understanding of the technical side of data analysis is critical to performing data discovery. All parts of a company need to be aligned and taking advantage of both types of reporting and analysis.

The Future: Refined Bimodal Analytics

Once bimodal analytics is established, it becomes very important to manage the interaction between the two modes. Managing the migration of insights and data from one to the other is vital to continuing to fully leverage the benefits of each.

Migration from enterprise reporting typically consists of ensuring that established warehouse assets are easy to use and understand in data discovery tools. Many data discovery tools connect at the database layer (bypassing easy to understand semantic layers). It may be necessary to create views or other intermediate structures to facilitate data discovery. Providing data in a smaller number of tables with fewer joins (denormalized) will make it easier to use in discovery. Particular analysts or data scientists may have other specific requests – listen to them!

Migration from data discovery to enterprise analytics presents many more opportunities and challenges. New insights, data sources, and calculations need to be considered for inclusion in the data warehouse. Warehouses should always evolve to meet new business needs, and data discovery is one of the best methods for surfacing those needs. Knowing which items are appropriate for inclusion in a warehouse is a matter for discussion between all involved parties. The more often something is being used, and the more widely applicable it is across the company, the more likely a candidate it becomes for migration.

The cost of processing the data should be considered as well. As a report becomes more important and data refresh requests become more common, the benefits of moving to the warehouse increase. Once this decision is made, output from the data discovery tool should serve as a prototype to accelerate warehouse development.

What makes these migrations work is data governance. Everyone who uses or works with the data needs to be involved in understanding how that data is being used within the enterprise and adjusting analytic assets to match those uses and requirements. When enterprise reporting was the main focus, this type of governance was easier. The addition of data discovery creates many more opportunities to explore and easily understand how data needs are changing. Without frequent communication, however, these opportunities will be lost.

Conclusion

Bimodal analytics represents a fundamental change to approaching business intelligence. Augmenting existing approaches with new ones is just the beginning. With multiple approaches, it’s crucial to understand which approach is best used for a given circumstances, and even more crucial to coordinate use of those approaches over time.

If you want to see how Ironside can help you develop a balanced business analytics strategy, check out our Analytics Advisory team page. Our seasoned and insightful advisors will help you see the vision of what your organization can be and the steps needed to get there.