Starting a data initiative – the bazaar approach

"The construction of knowledge graphs is more bazaar than cathedral."

We love this analogy:

"The construction of knowledge graphs is more bazaar than cathedral."

Barrasa, Hodler & Webber in Knowledge Graphs (2021)

The same can be said to apply to data initiatives in general. From the cases we have been privileged enough to follow up on closely, it appears that organizations taking a step-by-step approach achieve success the fastest. Rather than attempting to design and build a cathedral right from the start, focusing on one specific use case at a time leads to faster return on investment, which makes the initiative be perceived as more rewarding by its target audience, which in turn increases your momentum for follow-up initiatives.

Planning on a data initiative right now? Make sure you zoom in sufficiently on the specific use case you wish to start from. As Simon Sinek would say, start with why: which benefits will your initiative bring to your organization and its users? Proceed by outlining how you see the organization reaching this objective. Make sure you attentively listen to your audiences’ feedback to incorporate their feedback and elaborate on potentially confusing points. Finally, once you made sure you’ve got your audience on board, you can spend some time on the what: what exactly is this initiative called?

Some examples:

  • Why we need this initiative? Better visibility of and easier access to the available data assets in our organization. How we could achieve this? By implementing a user-friendly, automatically maintained catalog of available reports and data. What do we call this initiative? A data discovery portal.

  • Why we need this initiative? Reduce confusion when interpreting reports or data, facilitate the onboarding of new employees. How we could achieve this? By creating a business glossary with harmonized terms and data definitions. What do we call this initiative? A business glossary.

  • Why we need this initiative? Increase clarity on which reports or datasets are trustworthy vs which ones are not. How we could achieve this? By introducing an approval workflow facilitating review by the proper stakeholders. What do we call this initiative? Establishing data ownership & governance.

Taking it one step further, breaking down your chosen use case into several steps can increase your momentum. For example, when focusing on data discovery: unlock one source at a time. When establishing a common business language: iterate per domain.

Building a ‘data bazaar’ might not sound like a particulary attractive approach as compared to the outlook of a well thought-through ‘data cathedral’. It will however help you cope with a continuously evolving data reality in your organization, enable you to deliver value faster and in the end lead to a more engaged and satisfied user base.