Cohort analysis like you're doing there can be super useful for building communities, according to my own research and study.
Especially you want to find ways to expose trends in disengagement: at what point do most folks drop out? # of contributions? # of days? Is there a threshold of responses from other members below which they do out?
Knowing these trends gives the community useful information to tune new member engagement efforts.
It would be helpful to know something about what kinds of content keeps new members around, both posting and following. There's a lot of Bitcoin content here obv. Are newcomers starved for other kinds, and going back to greener pastures?
It'd be interesting to see if interaction with this new #grownostr content effort provides any lift in long term engagement.