Three years ago I literally googled “Data Governance” for an interview. I had no clue what it was about; all I knew was that it was a quite new discipline several companies were trying to start with and no benchmark was possible because we were all at a starting point, at least in Argentina. I found tons of information about Data Governance, but very little on how to move from theory to practice. I would like to share some acquired wisdom that I would have loved to know when I started. Spoiler alert: establishing a Data Governance program is an uphill struggle!
You can’t buy data governance
Don’t think a tool by itself will solve all your problems, tools don’t have magic powers. Before investing in one, you should first have a Data Strategy and a framework in place and only then, look after the best tool that meets your requirements.
Set an aspirational goal
If your main motivation for Data Governance is regulation and compliance, the best you can ever hope to achieve is just to be “compliant” (C.M. Bradley, 2013). Data Governance is mainly an enabler of a Data Culture and it is absolutely necessary if you really want to democratize access to data in your organization. That mission is much more inspiring than just meeting regulations.
Words Matter
Starting from the name, Data Governance could sound like something boring, or even worst it could sound like more work to do! Be creative and smart when you have to talk about it. Try to customize your speech for each of your interlocutors, starting with the real connection between their Key Results and how Data Governance will help them.
Express data improvements as business outcomes. Empathize with their daily data problems and start from there. Use analogies, recognize, make it fun.
Governance controls =! Data Handling Best Practices
Catalog =! Curation
Assign =! Recognize
Clean some data =! Decrease the number of undeliverable targeted marketing ads
Develop a Taxonomy =! Create a common vocabulary for the organization
Optimize a query =! Shaved 1 second off a task than runs billion times a day
Build Data Governance Through Data Stewardship
People define, produce and use data as part of their everyday job. Formalize accountability rather than assigning people more work. Set rules of cooperation and collaboration, use a Data Catalog to activate and formalize accountability. Data Stewards are going to be your main change agents, keep them engaged.
Agile Data Governance
Trying to govern all data at the time or trying to govern an E2E Data Domain completely detached from direct business impact, it’s simply impossible. Instead, try to define a process of creating and improving data assets by iteratively capturing knowledge as data producers and consumers work together, usually based on a Use Case. Data Governance should be part of an established process, where there is no other way to do it (e.g. All new Data Ingestion into the Data Lake should trigger Data Governance processes).
Address the Human Side of Data Governance
Data Governance is mainly about people. People’s behavior about data is what needs to be governed. We need to shift the old paradigms such as “IT is responsible for Data”, “BI is responsible for Data Quality”…get over it, we are ALL responsible for Data! We need organizational change management, and the first thing you should guarantee is having a good sponsorship of your governance program (the higher in your organization, the better).
Cultural change has been dramatically and continuously underestimated. Change is never a matter of ability, it’s a matter of motivation (avoid pain/ gain pleasure); invest time understanding what motivates people.
You´ll need an organizational alignment because culture eats strategy for breakfast!
Move From Strategy to Measurements
How are you managing your data governance program if you are not measuring it? What are you actually measuring? Is your measurement telling the story of impact?
% Of Governed DataLake
# Active Data Stewards
# of ad-hoc data inquiries received on slack
# of unique logins into Data Catalog
# of interactions between data consumers and Stewards (endorsements, change requests, chats, etc.)
% of employees accessing Data, per Business units
Identify your expected outcomes, measure, and adjust.
Communicate, Communicate, Communicate, and…Communicate
Nothing else to add, I think I made my point.