Eugene Yan, a data scientist, writes that he initially disliked using Scrum for data science work. Over time, however, the framework grew on him, and he now sees clear benefits in applying it to machine learning projects.
Yan's reflection is personal and specific rather than a broad industry survey. He does not claim Scrum is the only way to run data science, but he does argue that its structure can be valuable in a field where work is often open-ended and hard to estimate.
The piece is a single viewpoint, so there is no opposing source to compare against. Readers should treat it as one practitioner's experience rather than a definitive verdict on Scrum for all data teams.