Folding Data #28
An Interesting Read: How Data Teams Evolve
The last decade has shown that to become data-driven, any organization needs to establish a dedicated data team. How to build, run and scale a data team is, however, still an evolving field of thought with multiple dimensions: for example, what infrastructure, tools, and structure do data teams need at different sizes? And, importantly, how does the culture and the role of the data team evolve as the company matures? Emily Thompson suggests a natural progression from reactive to proactive to influential stages in terms of how the data team manages its work.
I wonder if a similar progression is happening in how teams approach data quality: the default state is fighting fires and fixing what blows up. Then you put some monitoring in place to at least learn about the issues before your stakeholders do. But it seems most exciting (and logical) to deal with data quality issues before they make it to production. More on that soon, but in the meantime let's make sure our team cultures are sustainable!
Tool of the week: Datawrapper
Because sometimes you don't need a BI tool. You just need a simple and beautiful chart.
Learning of the week: Graph Neural Networks
My introduction to Math and ML was – in the Soviet tradition – quite far from gentle. While I am proud that I intellectually survived, I think it's way cooler and better for the world to teach people complex topics in a more empathetic and enjoyable way.
That's why I am a big fan of Distill – an organization that gives prizes for works such as "clarifying things we formally know but don't understand" or "communicating novel results extremely well."
So what does all of that have to do with Graph Neural Networks? Deep learning is quite complex, and doing it on graph data structures can be even more intimidating. And Distill's Gentle Introduction to Graph Neural Networks does precisely what its title promises.
Before You Go