The usage of self-service business intelligence (BI) has grown significantly in recent years. And while many self-service BI platforms provide significant benefits, there is a downfall. Without proper governance, the same data silos from before the days of comprehensive BI proliferate once again throughout companies. Users are experimenting with their own data discovery, but those findings live only within each user, or at best, each department.
What is BI Governance?
The success of any self-service BI project depends on the execution of end-to-end BI governance that must be carried out seamlessly and unobtrusively. Otherwise, BI will be dismissed as yet another approach to optimization that only clutters the workflow, and the reign of the data dictatorship will live on.
We’ve written about the importance of sandbox analytics as a BI best practice for data governance, but here, I’d like to talk about the overall BI governance. That is, the all-encompassing quality assurance of not just data, but the entire system.
This includes monitoring the usage patterns of all BI applications and taking any necessary corrective actions, such as eliminating duplicates and identifying candidate applications to promote to the final production environment.
Two BI governance best practices should be a part of every company’s BI routine:
- Standardize the BI lifecycle
- Promote constant collaboration
Best Practice One: Standardize the BI Lifecycle
At TARGIT, we see BI governance best practices in the context of the following closed-loop BI lifecycle to ensure consistency and quality:
Insight discovery
The insight discovery stage is where self-service BI takes off (and can easily get out of hand). Here, business analysts (and other self-service BI content providers) create BI content for their own use. This can be based on authorized (BI production) data sources and/or combined with any data source that could be useful for business analysts.
Skilled Business Analysts can determine whether a measure in the given context makes sense and is accurate or accurate enough for the specific context. This is BI governance, as opposed to data governance, which relies on IT test mechanisms, data protection, and accessibility.
Insight testing
These new insights must be tested on the potential users before being released for general use throughout the organization to determine viability and usefulness. In other words, break up small, isolated groups to produce, experiment with, and share data before considering wider adoption. These groups should be intra-department to help better shape how this data would affect the greater company.
This is the "insight testing" phase of the lifecycle. The BI governance is performed by users participating in this testing. In other words, sandbox analytics. In this phase, Business Analysts must also weed out unnecessary data and clean up any duplicates and overlaps before promoting data to the next phase of the lifecycle.
Execution
Finally, tested solutions and verified data sources should be moved to production, that is execution, by the BI CEO/CC team based on the proof-of-concept or near production-ready solutions created by the Business Analysts.
By applying a lifecycle approach to self-service BI, the organization can better optimize resources as BI solutions are co-created between the BI COE/CC and the business user community. BI Governance ensures that the BI COE/CC only implements tested solutions.
Best Practice Two: Promote Constant Collaboration
One key to a successful (controlled) self-service BI environment is collaboration between the BI COE/CC and the business community. Therefore, it is paramount that the BI platform itself provides features to facilitate this collaboration.
In general, self-service content creation must happen in connected mode. That is, access to all data sources — even local Excel spreadsheets — should happen through the platform where relevant metadata about the data sources are created. This will enable the BI COE/CC to perform the necessary housekeeping on data sources and will provide the business users with a central repository for data — a single place to search for data.
Secondly, BI assets — data sources, data mashups, data models, and visualizations — must be tagged in accordance with their lifecycle state to enable a flexible model for governing the promotion from “Insight Discovery” to “Insight Testing” and from “Insight Testing” to “Execution.”
Furthermore, end-users must be able to distinguish “Insight Testing” from “Execution.” For data visualizations, this can be done by applying different style sheets. Promotion from “Insight Testing” to “Execution” can be done in a place where the BI COE/CC strengthens BI assets and silently promotes them. It can also be done by building the production solution from scratch, followed by the retirement of the “Discovery Testing” version of the solution.
As described above, this will further enable the BI COE/CC to execute proper maintenance of the BI assets, with analysis revealing any duplicates or overlapping assets and their usage.
