Katie Bauer!

What were the outputs of those teams like? Did the ICs directly collaborate on producing them?

In the first team the contributions vary, some are analytical in nature and thus inform decisions, or provide monitoring reports, but other are pipelines the ICs design and implement to support particular application features that require heavy (or subtle) data operations.

The second team is building an entire product, so that's what they do, they implement (in collaboration with designers and product managers, and paying close attention to the needs of users). They live in the conjunction of statistics and software development.

If I'm interpreting your tweets correctly, the first of those teams seems more like an analytics team (delivering analysis/insights and the pipelines used to produce them) and the second like more of a traditional software engineering team. Do you manage those groups differently?

Yeah, very differently. (The first also provides feeds some of their pipelines directly into application (say, recommendations, or similarity scores) and thus have user-facing consequences.)

Veeery interesting! That sounds like a huge scope for Team 1. What do you do to keep that team on track and prioritize to make sure they're delivering as much value as possible, especially when there's so many ways to do it?

Yep, that team (that is now becoming two) is very hard to manage. One of many things we do is keeping a very strong relationship with the relevant product managers through regular meetings (with me) and shared prioritization/planning.

Another one is having an always evolving (but still quite minimal) system for collecting potential work and determining which deserves execution. (Working on requests gotten via DM, for instance, is not okay.)