Companies need to add experimental business intelligence to their curated and federated BI planning, as it’s a high-risk, high-reward arena in the data and analytics space.
According to Gartner, business intelligence (BI) platforms enable enterprises to build BI applications by providing capabilities in three categories: analysis, such as online analytical processing; information delivery, such as reports and dashboards; and platform integration, such as BI metadata management and a development environment.
I have a slightly different take on the three categories of BI, as follows:
1. Curated BI: The formal dashboards and reporting every organisation requires − the home of statutory reporting. Behind this BI delivery are professional IT staff running governed procedures over formally managed data.
2. Federated BI: Dashboards and reports vary from formal to ad hoc. Federated BI teams are usually supported by the curated BI team and encouraged to conform to governance and data policy. The teams report directly into business rather than IT, staffed by professional IT or tech-savvy non-IT. Procedures are borrowed from curated BI. Federated BI solutions can be extended formally by moving them to curated BI, while others remain ad hoc.
3. Experimental BI: Here the results range from semi-formal to ad hoc. These are tech-savvy individuals with an idea. These are seldom supported by curated BI − that is until the results that emerge are reported to management as being important. The formulated idea is what is used to drive experimental BI, and its momentum uncovers necessary data with procedures engineered to get the job done.
Curated and federated BI are well understood. Key benefits to IT of enabling federated BI include moving 'shadow IT' out into the open and influencing the way data is managed and/or applied.
This is because shadow IT is usually unknown until a major failure or breach occurs and then IT must struggle to add or replace the solutions run by shadow IT. Naturally, the organisation benefits from an expanded data and analytics team, with business funding federated BI based on perceived, or measured, benefits.
However, not everyone benefits. It is essential to have a rich source of metadata from which to observe the behaviour of analytics consumers. Many analytics consumers use curated or federated solutions as a data source for further analysis, which translates into only one thing: they are experimenting outside of the BI space.
The work of the experimental category of BI is not new. It is a practice that is prevalent in every business, emanating from the age of the spreadsheet. It is by nature a high-risk, high-reward arena in the data and analytics space.
It is necessary to move from acknowledgement of experimental BI, to enabling it as a valid component of BI.
High-risk because people usually spend more hours fleshing out an idea than they would have expected. The results can be scuttled by lack of data or quality of data, or simply the wrong idea − or by the isolated nature of these exercises. Being outside curated BI, these experimenters are usually unaware of available data and analytics.
High-reward because organisations don't know what isn't analysed; therefore, when someone has an idea that is relevant and with reasonable data, the impact is naturally significant. This does not need defending. Curated and federated BI exist because of this fact. When it happens, the BI units must scramble to catch-up, and incorporate the new analytics subject matter.
It is necessary to move from acknowledgement of experimental BI, to enabling it as a valid component of BI. Expansion of data and analytics teams is crucial as it will mean less data opportunities are missed.
Moreover, it is important that the business understands that investment is required for moving experimental BI to curated BI.
- Data strategies must be expanded to allow experimenters to work with governed data wherever possible.
- Curated BI needs to embrace provisioning imperfect or 'good enough' data sources to allow experimentation.
Experimental BI is very hard to offer without SaaS. It is crucial to acquire SaaS solutions that remove capacity planning because the microservices environment expands and contracts in line with demand.
SaaS offerings that warn and manage individuals who exceed tenant limits are important. What I mean is in a SaaS solution, each client is provisioned into a tenant − these tenants are self-contained and parameterised to a client's needs.
A robust architecture that moves teams from traditional development, test, production environments to spaces that perform the same function within the one tenant is required.
Technical people understand DevOps and promoting solutions through development and user acceptance testing during a development cycle. Experimenters will balk at this, however, as each will know the value of polishing and checking their work before placing it somewhere for others to view and critique.
Silicon Valley's famous mantra, 'fail fast, fail often', is repeated at many tech conferences and disputed as hype by some, but failing fast mitigates the high-risk of experimenters dropping hours of effort without notice.
Rich metadata must allow data and analytics teams to understand where and how work is done across all three categories. Experimenters who embrace this approach are more likely to engage with data providers if the usual data source does not have what they are seeking.
So, what's the bottom line? The advice is clear: you need to add experimental BI to your curated and federated BI planning.