Your Data Modeling Choices Could Haunt You 👻
When guiding organizations on data model choices and structures for deploying martech software, I focus on the current state of the data, ease of use for non-technical users, and the solution's flexibility.
Much like a spider web 🕸️, tugging on one strand affects everything around it.
The Garbage Columns
It’s all about balance. What scares me the most? Decisions like “let’s create columns that could serve multiple purposes,” or even “we’ll define their use later.”
I understand—changing a data model can have serious consequences for some organizations. But even flexibility has its limits.
I’ve done my best to prevent these “garbage columns,” but I know they still lurk in many systems. So, are you doomed with these choices haunting you forever?
Documentation: Data Catalog and Beyond
There’s a remedy: if fixing what’s already done isn’t possible—too much downstream impact—you still have options. Cataloging your data and enriching your metadata and documentation can go a long way.
🔑 We all know data structures are rarely perfect or entirely self-explanatory. That’s why it’s essential for organizations to create meaningful metadata and explicitly document decisions around their data.
That’s the only way to keep your AI agents from wandering around like data-hungry zombies! Without proper data, they’ll stumble aimlessly, causing more confusion than clarity.
Happy Halloween! 🎃