To What Degree Should You Let Data Free?

Self-service analytics, citizen data platforms, or even data democratization is a north star in many organizations. It promises to both remove engineering bottlenecks and improve the depth and speed of data insights. As a result, this will satisfy data stakeholders and ultimately create business value.

Data wants to be free, but to what degree?

In this article, I propose an assessment framework to support what ‘data-decentralization’ practice would best fit your organization.

Between a Rock and a Hard Place

Great data comes with great responsibility. I’ve seen my share of failed initiatives because the organizational data maturity and understanding didn’t match the available data’s level of responsibility.

This creates noise, frustration, and, ultimately, a mistrust of data.

This year’s State of Analytics Engineering report from dbt¹ reflects the same narrative. On the one hand, data teams’ primary measure of success is enabling other teams (fig 1).

¹2024 State of Analytics Engineering

… and on the other hand, the main challenge is the stakeholders’ data literacy [my emphasis] (fig 2).

Data Literacy

For a data practitioner to try to enable others and to be, yet again, unsuccessful is, of course, not sustainable. Therefore, one point of departure is to assess the level of data literacy in the organization.

Create a pragmatic way to measure data literacy by designing a questionnaire with point scales and open-ended questions:

On a scale from 1 – 10, how comfortable are you working with data in your current role? How often (daily, weekly, monthly, etc.) do you use data to make decisions in your role? Describe a situation where you used data to solve a problem or make a decision in your role.

The above are examples taken from an excellent article published by Atlan². However, other questions that are more specific to your organization would take priority.

²Measure Data Literacy: How to Assess Your Organization’s Understanding and Usage of Data

Analyzing, segmenting, and visualizing the results, e.g., by department, seniority, teams, etc., provides valuable insights into understanding the data literacy level. And where to start and prioritize initiatives.

Furthermore, plotting the insights into a learning or maturity framework, such as the Dreyfus model of skill acquisition³, provides a good visualization and overview of creating a baseline to work from (Fig 3).

³Dreyfus Model of Skill Acquisition

For example, let’s say the marketing department has been assessed as having an advanced beginner level of data literacy and the finance department a proficient level. Different data processes and practices are in place in both departments, documented or not. And the best way to support these from a management point of view is to meet them where they are. The marketing department would need a more rule-based or centralized data governance practice. Moreover, the finance department would get the most value from a more contextual or federated practice (Fig 4).

A Framework for a Data Democratization Practice

Now that we have assessed and documented data literacy, we can examine current data management models and practices. Centre of Excellence (CoE), Data Stewardship (EA), Centre of Enablement (C4E), and the Community of Practice (CoP).

1. Centre of Excellence (CoE)

Centre of Excellence (CoE) is often defined as a place of knowledge, concentrating existing expertise and  resources, usually having a defined list of definitions, processes, and working methods.

2. Data Stewardship (EA)

Data Stewardship (EA) is defined by key people (Data Stewards) who enforce set data governance policies in an organization. A point-of-departure from a central department, like an Enterprise Architecture (EA) function. 

3. Centre of Enablement (C4E)

Centre of Enablement (C4E) is a “platform” that promotes self-service by offering tools, documentation, and resources. Founded in distributed technology and development practices.

4. Community of Practice/Guild (CoP)

Community of Practice/Guild (CoP) is a group with a common interest in a professional field (e.g. data science) that shares experiences, ways of working, documentation, etc., with each other. Learning as a collaborative process.

Inspired by a framework coined by Winfried Adalbert Etzel⁴, we can align these models and practices in a matrix. A matrix with one axis scale going from Centralize to Federate and the other from Control to Trust (Fig 5).

⁴Winfried Adalbert Etzel

With the assessment and framework in place, we now have an opportunity to map which practice(s) best fit the organizational maturity and need. If one has a high level of data literacy, taking inspiration from a Community of Practice model could boost the data initiatives. In contrast, if one has a lower level of data literacy, aspects from a Centre of Excellence and a more centralized approach would be a better fit. 

Life and the day-to-day are not this black and white. It’s important to note that many organizations will be somewhere in the middle. And they should take the best bits and pieces from each model. 

To what degree should data be free? It depends on the specific level of data literacy in the team, department and organization. And also what the best model and practice, presented above, brings forth the best balance between control-trust and centralization-federation.

About the Author

Robert Børlum-Bach has his day-to-day leading the Data & Audience Management practice at TV 2 Denmark – one of Denmark’s largest news and media organizations. This includes both managing and directing a team of software and data engineering professionals responsible for the collection, infrastructure, and activation of customer and behavior data across internal and external stakeholders. As a data consultant by heart, Robert has worked with international companies in industries ranging from manufacturing finance to fast-moving consumer goods. Knowledge sharing and community building are paramount, as made apparent by external teaching, conference organization, and public speaking.

As a speaker at the ninth annual Data Innovation Summit, Roberts speaks on (Re)introducing the Data Citizen to a Decentralized World! With data citizens’ right to data also come responsibilities, even more so in a decentralized or domain-driven context. His talk explores good, bad, and ugly ways of enforcing data governance with an increased autonomy of consuming data in an organization and can hopefully provide some takeaways on how (not) to do it.