To enhance point #1, you don't even need to have churn rates that change over time. Consider a very simple model where the probability of "disengaging" in a gi

To enhance point #1, you don't even need to have churn rates that change over time. Consider a very simple model where the probability of "disengaging" in a given year is 50%, at which point the follower doesn't engage any more with your content (but they don't unfollow you). It doesn't matter what churn rate you pick, you still witness the same effect. And then assume that your follower count grows linearly by 1000 each year. Time 0 1 2 Engage 1000 1500 1750 Follow 1000 2000 3000 Ratio 1 0.750 0.583 In fact, if you have a flat disengagement rate, as long as even a single person disengages without unfollowing, you will see this downward trend in engagement rates. Same assumptions as before, but this time, of the people that churn, 99% unfollow. Time 0 1 2 Engage 1000 1500 1750 Follow 1000 1505 1762.5 Ratio 1 0.997 0.992