Steemit Churn Rate and Engagement level - An analysis

in #steemit4 years ago (edited)

After reading only positives about Steem here; I decided to see if this platform has a good user engagement levels as the rewards part seem to be against the newbies. Steem being a open source blockchain, I decided to check the blockchain data at steemsql.com provided by @arcange.

Initial Setup

@arcange has written a detailed post about setting up the connection to his sql server database in excel which is provided below for your reference:
https://steemit.com/steemit/@arcange/steemsql-com-how-to-create-a-steem-analytic-report-with-microsoft-excel
As I primarily use a Macbook I decided not to use the Excel for reporting. Google Sheets is a good and may be even a better alternative to Excel as it is an online spreadsheet and can be accessed across all devices without any license issues and besides this, it is free! So I decided to set up the connection to steemsql server on a Google Sheets. I will post the details about setting up the connection in another post.

Whales, Orcas, Dolphins, Minnows and Planktons

If you check the steemit feed you can find many references to Whales and Minnows. If you are wondering what they are, they denote the level of Steem Power you possess. After scraping through different posts I discovered that Steem Power is derived from Vesting balance in the Accounts table of the database. Below is the categorisation of the users based on the Vests:

Vesting BalanceCoinage
Above 1000 MVestsWhale
Above 100 MVestsOrca
Above 10 MVestsDolphin
Above 1 MVestMinnow
Below 1 MVestPlankton

The more Vesting power you have, more weightage your vote has. If a whale votes your post then it has the potential to appear in the Trending area. So how are these fish distributed in the network?

Distribution of Whales, Orcas, Dolphins, Minnows and Planktons

The Steemit Network has crossed 350,000 users as of today. I queried the Accounts database to get the total number of users classified based on the Vesting Power table above and came up with this distribution:

User LevelPercentage
Whales0.01%
Orcas0.07%
Dolphins0.34%
Minnows1.5%
Planktons98.08%
Total100%

Planktons or Newbies are the overwhelming majority of users, this is mainly due to a large number of account creations starting from May and has been averaging about 1000 new users daily.

Active User Criteria

Steemit is a social network that emphasises upon voting and commenting so that the members of the community engage and encourage each others. So voting is one of the primary method to identify that an user has been active on the network. I am specifically using the timestamp of when a user last voted a post as the indicator of him/her/it being active in the network.

Quarterly Churn Rate

Churn rate is defined as the percentage rate at which users have stopped subscribing to a service. Putting it in another way; how many users have left the service for good. Facebook for instance; measures its churn rate based on the likes by the user. So Votes could be considered as the Steem equivalent to get the churn rate here hence I am checking the last voted date of the users.

I actually measured the Churn rate based on the users' power level so that we can get an idea of how strong/weak each level is. A user who has not voted in the previous 45 days is considered as a 'Stale' user whereas if the user has not voted for the past 90 days then the user is counted as 'Dead'. Using this as the yardstick I queried the Accounts table are was alarmed at the results:

User LevelChurn rateStale percentageActive Users (45 days)
Whales35.71%4.76%59.52%
Orcas33.48%2.17%64.35%
Dolphins25.00%3.46%71.54%
Minnows31.19%6.56%62.26%
Planktons41.72%22.23%36.05%
Overall41.50%21.92%36.59%

Key Takeaways

Steemit's churn rate is at a staggering 41.5%. In simple terms 2 out of 5 users have not voted in the last three months or quarter. Particularly the newbies/planktons are leading the churn rate charts with 41.72%; this is really significant as they form 98% of the user base. The stale percentage shows that things might not improve in the near future as about 1 in 5 users have turned stale so far with newbies/planktons again leading the table with 22.23%. The stale percentage for other categories are below 10% but then the newbies/planktons are the majority users. A social network needs it new users to comeback in order to sustain it's popularity.

Conclusion

  • Steemit has a serious user engagement problem on comparing 41.5% churn rate with Facebook's 3.5% or even Twitter's 28% when they were providing the churn rates a couple of years back.
  • Most of the newbies may have joined the network looking to make a few quick bucks but left disillusioned when their posts were not getting them enough SBDs or are barely being noticed(this post may meet the same fate 😊).
  • The steemit hierarchy looks like an oligarchy which is bound to fail as the majority of voting power lies in the hands of less than 2% of the whole populace.
  • The home page feed is really inadequate when the user base is expanding exponentially. One can see the same post coming up in the 'Trending' tab for days as if no other quality posts are available. The tags sidebar is also basic and could be of real use if it uses Twitter style trending tags.
  • The lack of a full fledged native mobile application might have also contributed to this lack of satisfaction as I always had trouble posting photos through the mobile apps available now.

Steemit is a really good initiative that encourages user to provide good content and get rewarded on attracting mileage. But it has to rethink it's voting mechanism and rejig its home feed to let people discover other's posts easily. If more quality posts get more votes, the authors will be happy and may contribute more to this unique social machine.

Check out my voting pattern analysis at:
https://steemit.com/steemit/@dbdecoy/steemit-voting-pattern-analysis-based-on-hour-and-day-of-the-week

Sort:  

I think that perhaps the unequal distribution of voting powers is the single biggest hurdle, which is probably also the one which is most difficult to overcome.

I would love it if you could take it upon you to make a post about what it would mean if the steem was distributed equally amongst the current users. How much would the account balance be worth? How much would a vote be worth?

Sure I will do that. But first need to take a look at the inequality in the vest distribution. I will post about it soon.

Followed just to not miss that post of yours. Always love to see some data.

Thanks for the follow, hoping not to disappoint you in future 😅 Followed you back.

Don't worry. I'm sure you won't! Just keep doing you :)

  • Most of the newbies may have joined the network looking to make a few quick bucks but left disillusioned when their posts were not getting them enough SBDs or are barely being noticed(this post may meet the same fate 😊).
  • The steemit hierarchy looks like an oligarchy which is bound to fail as the majority of voting power lies in the hands of less than 2% of the whole populace.

We are getting new apps and platforms that use underlying Steem technology which could help fix the first issue. Also communities are coming soon.

But the second issue is something that... is a big issue. I hope the situation gets better in year or two, if not there might be something wrong in the system itself.

Communities will help in increasing the engagement levels. Regarding the second issue - most of the whales/orcas were early adopters/steemit devs/witnesses who get rewards for their mining. So this hierarchy might not go away soon.

Here's a glimpse of the author rewards distribution just prior to HF19:

authorrewardchart.png

I am not a data scientist, and this chart was posted by @aggroed, who stated it came from @arcange. I have repeatedly requested more current data, in order to better ascertain the impact of HF19, to no avail.

I suspect the situation hasn't improved.

Yes, I think so. HF19 states that rewards are proportional to the stake, the distribution of stake is skewed in the first place!

I wonder if inequality of STEEM power matching inequalty of income and capital in real world.

Heya @dbdecoy, this post is only more encouragement to actively participate to build up steempower. In the begining it is always hard to het started in participating on a regular basis. Since my Facebook engagement is high, I thought it would be a fun experiment to take a logged out hiatus for a few months to better establish my connections here! Participation is key, and everyone learns fun things c:

@shello

That is true. The response to this post is truly encouraging.

You have encouraged me to get more involved. I'm sure there will be much more participation once DTube becomes more well known about on this platform. I love analysis and will follow you to see more great posts.

Thanks for the follow!

Thanks man, positivity can change the world.

awesome analysis. Churn, or retention is a problem I spoke about before in an analysis and post. I will be looking at it again soon.

Many accounts that are voting, are bots, so really you also need to look at comments and posts.

Good to see someone using google sheets. It;s a grand tool, although I am an excel and power bi lover myself.

Keep these type of posts coming....

Thanks, good observation it was a quick look so I thought voting should be enough for now. Bot or not; as long as they distribute steem among users, it should be fine.

You have made some good suggestions. I wish that the voting power can be equitably distributed. To make Steemit a success, everyone has to contribute.
Steemians Ask Not What Steemit Can Do For You But What You Can Do For Steemit

Need not be equitable, but definitely not so skewed.

Bots are killing these stats.
Need an insight how to exclude them.

I'm working on a Spam/Bot Classifier. When it's improved, I'll make it available as an API which might help the exclusion in future.

Missed this comment! Nice use of machine learning, I am looking forward for the API. Following you for further interesting posts.

Have no clue about how to do that!

Me too yet.
Just underlining the problem.

You can never solve a problem on the level on which it was created.

- Albert Einstein

Intersting post

The distribution rate looks distinctly like the wealth of the world and how its divided with the top .1% owning as much as the other 98% and we can see where that is leading. I hope thinks can be divided better for minnows and as i've learned us plankton :)
Lets hope HF20 can help

Whenever something claims to be fair or democratic, history always repeats itself. So steemit is sn oligarchic democracy.

This was an incredibly useful analysis. I have resteemed it as well because I think that your views around the churn rate are very useful to the whole community.

Thank you!

Hi, Dbdecoy, In case this has been your first Introdusemyself Post i'm here to welcome you to Steemit. I hope you have a lot of fun here and you may follow me. Have a great time @rightuppercorner

There is ambiguity between the terms 'users' and 'accounts', particularly if a reader understands a 'user' as a unique individual. A single user may have several accounts, indeed there are some with several thousand accounts. Plankton may form 98% of the accounts, it's not clear that they form 98% of users.

That's not to say that there isn't high churn, but some method of determining the proportion of accounts that belong to the same users in each category is required for clarity.

It may not be possible to identify multiple accounts by logic here. We may have to do it manually.

Statistical methods might be possible. Sampling accounts, using a manual process to find multi-account users and extrapolating the numbers based on the sample.

Yes, but we may need more manual power and the result might not be worth the effort. Let the professional organisations do that if steemit becomes the next Facebook and they want to write about this network's bot problem!

Eventually the reward pool should make it financially worthwhile, but not now.

Yes, Agree.

Nice post. A pretty big chunk of the plankton/minnows will of signed up in the last 45 days aswell so the stats will look worse in a few months.

Yes, things will get worse. Need to run this start again in a month to see how bad it is.

Awesome post!

Followed in the hope of more to come. There are some folks that might contribute to trying to sort bots from people, such as @sherlockholmes, @spaminator, and similar folks that are seeking to fight the botnets.

I'd be very interested in seeing an estimate based on data of how many individual users have Steemit accounts, and a churn rate of those people, if you can manage it.

Thanks!

upvote, follow, resteem

Thanks for the follow! Some bots can be identified through database queries but finding spammers/scammers require the investigative skill, manual effort and patience of users like @sherlockholmes and @spaminator.
We cannot really find unique individuals as steemit does not track that data and if people wish they can hide their alternate accounts really well.

Well, @sherlockholmes has identified several thousand bots, at least, and others have also, I think. These can at least be subtracted from the total, providing a more accurate figure. Perhaps those folks might also be able to point to other networks, or provide an estimate for the size of such, that could refine your figures some.

Yes, I will check that. Might not be able to do so till next week as I might will be away.

Very detailed post. I want to add, that voting itself does not necessary mean "active" a lot of users use bots to vote. Another detail, you didn't count accounts that were never active, they are about half of all accounts. And if you count unique account holders that's even less.

I will look at a way to filter some of the bots. Accounts that were never active will be filtered out anyway by the query. Finding unique account holders is not possible with querying the database in the current set up and may require more manual firepower.

Accounts that were never active will be filtered out anyway by the query.

You can find accounts that were never active when "last_vote_time" is equal to "created"

Finding unique account holders is not possible with querying the database in the current set up and may require more manual firepower.

That's right. There are some info in the post, like @noganoo - if I spelled it right - has over 1000 accounts. @berniesanders has at least 55 accounts. There was a spammer here that looks like he owns 2700 accounts: https://steemit.com/steemit-abuse/@patrice/2700-accounts-love-spam-comments-is-this-a-case-of-gaming-the-system-or-abuse
And more... When you filter all these accounts, you will see that there is a small percentage of active Steemit accounts.

How about pooling these account names together into a common Google sheet so that users can update there directly and avoid duplicating their labor? I will try to curate this next week.

Good idea!

I think you'd need to filter out people who have been dead for longer than, say, 180 days. Otherwise the churn rate as you are measuring it is guaranteed to go down over time no matter how well Steem performs in the real world

I picked 90 days so that it represents a quarter and several newbies started joining the network around this time. Old Faithfuls may still be sticking around and are having a rich harvest. If steemit holds the newbies' attention longer then it definitely will grow larger and stronger.

Good Post..
BUT..
wherein you said" The steemit hierarchy looks like an oligarchy which is bound to fail"

From multiple sources it can be noted that:

The iron law of oligarchy states that all forms of organization, regardless of how democratic they may be at the start, will eventually and inevitably develop oligarchic tendencies, thus making true democracy practically and theoretically impossible, especially in large groups and complex organizations.

Is bound to fail a synonym for inevitable?

If it IS inevitable (all of history suggests that it is) then why emulate King Canute?
It would seem to me that we should attempt to find a way to turn lemons into lemonade?

Well... Nothing lasts forever! Even Google might become history. If steemit can prevent this in its early stages it may survive longer.

the 'space'..she is a changing VERY fast. It's natural selection in the InfoSphere. Descent with modification. 'Evolve' to survive in the rapidly changing environment...or die.