Katie Bauer!

I'm on the record as thinking DS-as-Insights-Generator is a weak vision for data science, but another problem with this model is that it feels a bit like science fair. It can be fun and even educational, but who actually does anything with their poster board once they're done?https://twitter.com/imightbemary/status/1360378868511559680 …

The format you present your findings in is an aspect of this because nobody has time to read your 20 page deep dive, no matter how deeply researched or statistically airtight it is. But even if your Insight is in a highly consumable format, it may still struggle to gain traction.

Data needs to be put in context to be useful, so the real way to make Generating Insights a productive activity is to spend time connecting them them to a broader understanding of the product or system they're about.

The hard part about this is that companies already have systems of meaning-making, even if they're informal, and whoever currently drives this system may not be interested in having a DS team come in with a perspective that contradicts their carefully crafted narrative.

If you find yourself in this situation, it's important to remember you're on the same team as this narrative creator. You work for the same company and both want it to be successful, even if you have different ideas about how to make that happen.

So learn about how they see things and be willing to learn to speak their language. Putting your work in a context they already understand will make you more comprehensible and get you more buy-in and trust.

This might feel like giving in, but it's actually making you part of the existing knowledge creation process. If you do your job well, more people will understand where you're coming from and what you do, and you'll be able to play a bigger role in knowledge creation over time