1. They are just trying to ride the ML hype train and they sell effectively polished turds to board member buddies. This is the vast majority of the alt data and analytics firms. They have flashy dashboards and allow you to drag&drop together models and data. The whole thing looks very fancy during the sales pitch. It's all animated and shows charts that impress CEOs. Then when your quants get hold of the data they discover that it's real world predictive power is zero.
2. They actually discover valuable data or build models that genuinely produce an edge on the market. These either turn into hedge funds or are bought by an existing hedge fund to run the model in house. You rarely hear of these because their offering is public only briefly or not at all.
Hedge funds also fly cesnas over refinery fields to look at the levels of those big tanks you see off the highway... most of them have roofs that move up or down with capacity, and those that don't you look at with an infrared camera to look at relative temperature differences.
https://www.economist.com/finance-and-economics/2019/02/28/n...