User retention and turnover is a much discussed issue on Hive and Steem historically. Retention and turnover are very close conceptually, with turnover being one metric which reflects how well a project is succeeding at user retention. High turnover generally means you are succeeding in bringing in new users, but that those users are not staying for long, which constrains your overall growth. For a project to grow, it must bring in more users every month than it loses. High turnover may be especially bad in a space where there is a significant learning curve, where it not only limits growth but means lost expertise as well.
Cohorts
One way of presenting user retention is to divide our users into cohorts, generally based on when they joined the network. I will define each cohort based on the year of account creation, although I think there is a better way to it this is the simplest and the easiest to grasp at a glance.
The first two charts are the sizes of the cohorts overall, the first based on users who have made a post (aka authors) and the second based on users who have made a vote transaction (aka voters or curators). All further charts will be based on making a post.
The oldest cohort is the original from 2016. Not the largest, but with the longest history, it shows the typical journey of a cohort - they typically get smaller every year as fewer users continue to post. A mere 670 accounts from the 2016 cohort have made a post this year.
Big Picture
Since there have been 7 calendar years since the start of Steem/Hive, there are 7 cohorts overall. Their activity can be seen in the stacked bar chart below.
As is clear above, the largest cohorts of 2017 and 2018 dominate the chart, despite being mostly gone by 2020, the year of the Hive exodus fork. The cohorts of the Steem era were vastly larger than after the fork. That is despite the success Hive has had in the last year. Of course, it could look very different if custom_json transactions were the basis of this post.
For the sake of readability, here is the chart again starting in 2020 (the year of the Hive exodus fork).
Some notes again: We're only half way through 2022, the chart will look different when the year ends.
Although the decline in active authors has slowed, lack of user retention and high turnover absolutely still seem to be a problem in Hive.
Stakeholding by Cohort
Before I finish, I think it's also worth briefly looking at the holdings of stake between the different Hive cohorts. The first is based on the entirety of the cohorts regardless of posting or voting activity.
When we only include those who made a post this year things shift somewhat, with the 2019 cohort gaining significantly in relevance. I don't know why the 2018 cohort diminishes so much here. 2022 cohort is barely visible.
However when active voters are considered instead, the 2017 cohort gains, although still is behind the really old guard.
Although a tiny fraction of the original cohorts have remained, those who did have a massive disproportionate stake in the network overall.