One of my client inquiries Friday repeated a sad refrain I hear all too often. In this case it was a public sector agency. The outline of the call went as follows:
- I was taking to a central state agency who was organizing a data governance initiative (in their words) across three other state agencies.
- All four agencies had reported an independent but identical experience with data governance in the past.
- All had started out as huge initiatives, all with strong executive support and recognition by many that the work was important. But at the end of the day, the initiatives were shut down after a time as there was little known or perceived business impact or value.
- The client had recently engaged with a well-known consulting company that had recommended a large data catalog effort to collect all enterprise metadata to help identify all data and business issues. Through the use of AI and ML, these new catalogs would find all the data and create a new data model much more quickly then before. With this, the recommendation was the formation of a large enterprise data governance board with CEO sponsorship. This sounded to the client just like the past effort, the only difference being the addition of AI/ML.
- Half those that had heard the recommendation thought the whole thing was a silly idea and too reminiscent of the previous efforts that never added any value and never even finished. The other half of the team wanted to try again: maybe AI created catalogs would make things different….
- The question was this- is there a better way to do of this?
This was a great inquiry since it called into question the perceived wisdom peddled by some that cataloging everything was a prerequisite for data (and analytics) governance. It is not, other than in a few use cases.
Modern data (and analytics) governance does not necessarily need:
- Wall-to-wall discovery of your data and metadata
- An enterprise-wide data governance board
- An expensive consulting engagement
I won’t bore you with the details behind these statements but, if you don’t have any effective D&A governance in place, you already have in hand what you need to know and where to start. Don’t focus on data; don’t focus on standards; don’t focus on principles. And don’t start with a focus on domain specific data. Instead, start by thinking of business outcomes. If you need more, give us a call!
- Webinar Effective Data and Analytics Governance – Finally!
- Blog A Little Data Governance Goes a Long Way
The call was my penultimate inquiry of the week. My very last inquiry was much more positive. I spoke with an IT software vendor about an aspect of data and analytics governance. Specifically, we were talking about analytics stewardship. This is a new area of innovation that needs to be understood as an intersection between:
- The work of data and analytics governance
- The use case, and
- The information artifacts in scope
The work of D&A governance should always be explicit:
- Policy setting
- Policy enforcement
- Policy execution
The use case could be:
- Business process integrity and operational data
- Upstream in business applications
- Analytical quality and analytics
- Downstream in the analytics pipeline
Scope could be:
- Data (i.e. raw data)
- Information (processed data)
- Analytic (the analytics itself)
- Records (files, or what you might all unstructured data)
- Images (i.e. digital)
- Events or transactions
- Anything else you can think of
Analytical stewardship is a missing link in analytics, BI and data science. While firms can now be effective with operational data governance in business application (see webinar Effective Data and Analytics Governance – Finally!) they need increasingly to get as good at analytics policy enforcement and execution in their analytic environments and applications: analytic stewardship.
Any solution that is used to develop data warehouses, data lakes, analytic models, build dashboards, analytics or data marts, needs to respect the polices set by D&A governance (which is essentially the same org). The policy enforcement however has to take place in the analytic apps, just like data stewardship takes place in the source business apps. So, with all the excitement with data warehouses and data lakes, and storytelling and AI and catalogs and dashboards, analytic stewardship is work missing in action. Analytic stewardship is a critical capability needed to help assure trust in the analytics used to generate insight and drive effective decisions. We will all see and hear a lot more about analytic stewardship going forward.