Radial

Steven Sermarini is a Senior Director of Data and Advanced Analytics at Radial. He provides thought leadership on Big Data Analytics, Data Science, Cloud Architecture, Business Intelligence, Reporting and Microservices. With over 25+ years of industry experience including tech start-ups and fortune 50 companies, he has helped several organizations transition to cloudnative architectures and has worked on patents related to Big Data Analytics projects. Steven holds a Master of Software Engineering from Carnegie Mellon University, CPM-Project Manager from Boston University, and a B.S. from Wester Chester University of Pennsylvania.

As we see the mass proliferation of Power BI across organizations, the tug-of-war has intensified between BI platforms and portals run by IT vs. localized BI tools run by business groups. In general, the recent popularity of this tool has greatly accelerated the drive towards de-centralized BI development, down to the team level. There are philosophical battle lines that typically get drawn between centralized IT organizations and other business units regarding control of BI development.

What the business wants:

• Vision and control – They want to have control at the departmental or team level over the BI product, to align it with their subject matter experts’ vision of the target state and solve their most important business problems.

• Velocity – They want it now. IT delivery times can be long, as most IT organizations are chronically understaffed in relation to project workloads, tech roadmaps, and tech debt work.

• Agility – The business wants to be able to try something and pivot quickly when it’s not quite what they need or when the business environment is changing.

What you get from a centralized, IT managed solution:

• Single version of truth - Master data is always changing (i.e. – codes, flags, filter fields and conditions) continually evolve within the source data ecosystem, leading to data built on a specific logic becoming stale quickly. Creating many copies of data proliferates misalignments of the data.

When IT centrally manages BI, they have governance and controls in place to ensure the calculations are being done correctly by keeping the logic updates when changes happen.

• Data quality - Continuously compiling data and cross-checking localized BI and reports with colleagues takes a lot of time and can lead to distortions, and incorrect aggregation logic. When IT centrally manages BI, they have governance and controls in place to ensure the calculations are being done correctly by keeping the logic updates when changes happen. Nothing is worse than sharing an incorrect report with an important client, customer, or partner.

• Standard working best practices – If each department builds their own BI, how do you share data between departments when there are no standards?

• Support – Who is responsible for answering questions, fixing bugs in reports, making sure report updates are run, handling system outage or data inconsistencies when each team runs their own BI? If consultants are brought in to write macros or Power BI logic, what happens when they leave? Who understands what they built? Who will maintain it?

• Consistent Performance and Results – When you need to report on and aggregate large amounts of data, the architecture for localized BI tools can quickly hit limitations. The added load that distributed BI can add to a data warehouse can be significant once the number of users grows large. All of those connections may be hitting some of the same data sources with no way for IT to consolidate or optimize the queries.

• Data access control – Unlike a centralized BI portal with access controls, tools like Power BI allow users to share information with no access controls, thereby allowing for the possibility of sensitive company information to be exposed to unauthorized people.

So, what is the best path forward for an IT organization dealing with decentralized BI sprawl? A mix of both solutions can be useful in an organization. Some potential steps that help mitigate the pain points of letting each department or team run their own BI can include:

• Centralized licensing/procurement – Centralize all procurement so that all licenses, including Power BI, require IT approvals. This allows IT to have a chance to collect information from the business on what they are trying to solve and a chance to align strategies with corporate goals.

• Centralized access to data – Understand who is accessing what data directly and ensure that proper business justifications are provided before providing access.

• Create a Center of Enablement - The previous 2 items are essentially gatekeeping that provides IT with the opportunity to apply some governance principals and controls that can help avoid many of the pitfalls of distributed BI listed above. Here you can create governance standards for how to source and aggregate data, best practices, and an open line of communication with users to notify them of master data changes and other impacts to the data they are depending on. IT may be able to provide resources with some limited availability for help and training as well as centralized catalogs of what data and resources are available to pull from.

• Training - Have a mandatory introductory session and routine training/ information sessions (and record them). As business groups request new tools and access, require them to have a bootcamp session to get them started in the right direction, applying all governance and best practices.

For centralized BI practices and portals run by IT, here are some key considerations to enhance the profile of these platforms:

• Have strong product owners for the BI platform. They need to be connected to the lines of business, understand the biggest pain points requiring BI and who the subject matter experts are.

• Create a backlog of projects, communicate to all stakeholders the prioritizations and get buy-in.

• Have regular user group sessions and internal newsletters that highlight the business value and what reports and insights are available. Many times, business teams re-invent the wheel because they simple don’t know what reports IT currently can offer.

• Track usage statistics to manage user adoption and ensure you are building useful visualizations.

• Offer training sessions on a BI portal to encourage usage.

Final thoughts:

Depending on the size and type of organization you run, you may lean more heavily towards de-centralized BI or vice versa. Small shops may forgo centralized BI all together, while larger organizations that require a client and/ or internal facing BI portal may decide to run mostly centralized. Either way, it is inevitable that proliferation of tools like Power BI will happen at some point. It is important to have a strategy on how to best benefit from those tools without sacrificing quality and reliability. Having oversight and governance is the key to making sure that grassroots work does not run counter to the corporate values.