5 Comments
Mar 6, 2023Liked by Ryan Dolley

Build the strongest roots, allow the most flexibility towards the ends of the branches to grow towards the most nutrient-rich light (insights) as possible. Find the best trunk (metrics) that allows for the most types of branches to grow. Make sure that the ends of the branches have a way to transfer those increasingly nutritious insights back to the roots. Easy enough?

Expand full comment
author

Haha, you nailed it!

Expand full comment

When is part 3 coming out?

Expand full comment

Can you think of an example of discovering a new insight/metric during exploration/discovery and then what the current as-is state would be to export out or push that insight back down to ML/AI apps, or even a new dashboard?

Expand full comment
author

Yeah, imagine I'm using Tableau to do what they used to call 'visual data discovery' and I do a market segmentation that looks promising. Now we want to use it for a 'next best offer' ML model to integrate with our online storefront.

What you'd have to do today is have a meeting with the data engineering team and explain the logic to them, then have them recreate it in ETL and land it in a database of some sort, then they can pull it into the ML model. There's a ton of waiting.

In an ideal world the BI tool would support pushing to a headless semantic/metric layer that could be directly accessed by your data science team. No engineering intermediary.

Expand full comment