The Modern Data Stack Has It Backwards

Most companies trying to be "data driven" are drastically overspending and not getting a positive return on their investment.

There is a term that became popular because of marketing and venture capital called the modern data stack. It said: pull all your data into a warehouse (Fivetran, Snowflake, Databricks, etc.), organize all your data into consumable views and tables (dbt), then give it to analysts and stakeholders to make dashboards (Tableau, Power BI, etc.).

In theory it's nice. The tools themselves aren't bad, per se, but the idea of transforming all of your data into consumable bits is extremely difficult. It turns into a giant game of telephone and guessing what people want. Which leads to more hiring, a bigger game of telephone, and more cloud spending.

The only thing I've seen truly work is to flip everything around:

Find an extremely important problem to solve, gather the data needed to solve it, then solve it with the tools available to you.

Complexity will emerge, but it will happen naturally. If you solve the first problem in a scalable way, you will be able to scale up and become truly "data driven."