Engagement on steemit is a key component to its success. Someone writes a posts and others comment and engage. Discussion are created. But how well are people really engaging on posts?
Repository
https://github.com/steemit/condenser
Aim of Analysis
The aim of this analysis is to establish a trend for the number of comments per post over time and also to get an indication on what level of comments are human comments and human engagement. ( It has been requested a number of time to estimate the bot activity)
The Findings
The pie chart below is a simple representation of the number of comments left on all posts since January 2017. Originally I plotted a histogram to plot the distribution of comments but as 80% of posts get less than 5 comment, the chart was skewed too far to the right to be even visible.
Take a look at the image below. It shows in the blue bars the average comments per post per month since the start of 2017. The black line shows the number of posts, the red line shows the number of comments(children) and the yellow link shows the unique authors.
The data looks interesting and in terms of engagement rather positive really. It looks like the average number of comments per post bottomed in November with an average of only 3 comments per post. Since then the number of comments per month has increased massively and the average comment per post was up to 3.3. Looks like good news so far for engagement.
Have a look now at the GIF above. It shows the same data, this time removing the posts with no comments, the posts with less than 2 comments, then less than 3 and then less than 4 and 5 as these make up 80% of the posts.
The GIF shows that when we exclude posts that have less than 2 comments (which is 60% of the posts) the average number of comments per post increase to 8 which is rather stagnant since Nov 17.
By the time we remove 80% of the posts, the average comments per post increases to 14.5.
Overall however the trend remains very close. There was a decrease in the number of comments per post from June 17 to Nov 17 and since then it has gained ever so slightly. Nothing to write home about.
There are many problems with comments or lack of good comments left on post, but there is also a problem of bots. Bots leave comments for many reasons, we have welcome bots and bidbots and flagging bots all leaving comments.
Adjusting for Bots
Lets take a zoom in to March and April this year and we are going to filter the data so that we are only looking at posts that have at least 1 comment.
How different would this data be if we removed the comments from the bots? To get an indication of this I took a random sample of well-known bots.
@minnowsupport @cheetah @steemcleaners @randomwhale @booster @buildawhale @welcomebot @randowhale @randowhale0 @minnowhelper @randowhale1, @drotto.
This sample represents bidbots, support projects and spam filters
We all know 12 is a very small sample, but lets take a look at the data now, excluding comments from these particular bots.
Can you see what happened there? Let’s look at the same information in table format
Looking at all posts with 1+ comments, if we remove the comments from the bots listed we remove 73% of the posts and 85% of the comments in March. On what is left, the average comment per post reduces from 5.73 to 0.84
Conclusion:
Looking at Steemit in general the number of comments per post has generally decreased since January 2017 and the % of posts with 1 or more comments averages at 58% and with 5+ or comments is only 20%
However as everyone is aware since 2017 there has been a substantial increase in the number of bots leaving comments. If we remove all of the comments from the bots listed, the results are not good.
This estimating is showing that the number of posts with 1+ comments reduced by 73% if we remove comments from bots. Therefore 73% of posts have received a comment from a bot. The total comments has reduced by 85%. That is massive. That suggests 85% of comments are left by bots. Suggesting only 15% of comments are human
Data and Query
Although I was expecting a high level of activity from bots, I was not expecting it to be that high, so below are my workings and queries. Please have a look at this carefully and tell me if you spot anything wrong. I would love to be proved wrong on this.
To get all the Comments data the following query was used
SELECT author, root_title, children, created, depthFROM comments (NOLOCK)where created >= CONVERT(DATE,'2017-01-01') I then used the following calcuations
No of Posts = COUNTROWS(FILTER(comments,comments[depth]=0))No of Comments = COUNTROWS(FILTER(comments,comments[depth]>0))average comments per post = [Comments on posts]/[No of Posts]these calculations were further filtered to exclude posts where the children was = 0 to get the values of posts with 1+ comment.
To get data excluding bots I carried out the following query:
SELECT author, root_title, children, created, depthFROM comments (NOLOCK)where created >= CONVERT(DATE,'2018-03-01') and author not in ('minnowsupport' , 'cheetah' , 'steemcleaners', 'randomwhale', 'booster', 'buildawhale', 'welcomebot','randowhale','randowhale0','minnowhelper','randowhale1','drotto')Calculations
Posts = CALCULATE(DISTINCTCOUNT(comments[root_title]),comments[depth]<>0)No of Comments = COUNTROWS(FILTER(comments,comments[depth]>0))A comment per post = [No of Comments]/[No of Posts]I then added a filter to each of these calculations to exclude posts where the children was = 0
At first sight your analysis looks fine. I have one detail you could have a look at.
According to your data the average number of comments per post is always above 3. According to @penguinpablo data this number is in 2018 always below 3. Or I'm missing something and probably need a pair of glass? Or this is an inconsistency in one of the data queries?
I think there's at least one more bid bot that really spams a huge number of comments: @speedvoter. I wouldn't be surprised if this bot created 30k+ comments per month.
I might have an idea for a similar analysis. Based on data that is already available. I will look into that later this week.
I was not aware of @speedvoter. looking forward to seeing what you produce, hope it is a post with better news than this
In that case I better don't write anything about this subject 😋
Hey @crypto-econom1st
Here's a tip for your valuable feedback! @Utopian-io loves and incentivises informative comments.
Contributing on Utopian
Learn how to contribute on our website.
Want to chat? Join us on Discord https://discord.gg/h52nFrV.
Vote for Utopian Witness!
Thanks I really appreciate it!
I didn't know you also upvote comments. Is that something you do regularly?
I could not find information about it on your website?
Utopian recently started picking some comments giving valuable feedback to contributors, and giving them a 1% upvote. You actually gave good feedback, and a community manager saw your comment and there the upvote is :D
See this post for more information (under the "New Tipping Bot" headline)
Thanks for this information.
This sounds like fun. Since I'm no real IT guy. But so now and then I am able to review an Utopian-io post. If only 1 out of 10 or even 25 of these comments gets an upvote, that would just make my Steemit fun a bit bigger. And most likely that goes for many more Steemonians.
Two thoughts:
From one point of view, I see bots leaving their comments a good sign in the case that perhaps it is this comments that could be the lifeline for a beginner to continue posting (in consideration of the use retention)
This may require a deeper (?) look on why some posts don't get enough comments. One thing that come to mind is perhaps most posts are mostly pictures and it doesn't really trigger engagement?
And lastly, although of course this is hoping for something not yet implemented, is hivemind again. I do hope it becomes a game changer for the community. Most of the communities I believe are connected through discord. How this community will become or grow in hivemind is still something to see.
Your contribution has been evaluated according to Utopian rules and guidelines, as well as a predefined set of questions pertaining to the category.
To view those questions and the relevant answers related to your post,Click here
Need help? Write a ticket on https://support.utopian.io/.
Chat with us on Discord.
[utopian-moderator]
a lot of posts are pictures this is true. Dlive has also just taken comments off the block, so this will make an impact on the negative side.....
Just noticed the utopian site.......are we back??????
Hey @eastmael
Here's a tip for your valuable feedback! @Utopian-io loves and incentivises informative comments.
Contributing on Utopian
Learn how to contribute on our website.
Want to chat? Join us on Discord https://discord.gg/h52nFrV.
Vote for Utopian Witness!
I don’t think this is as bad of news as it might appear.
Steemit like all other social sites relies on the network effect. Currently there are just barely enough people on the blockchain to make things work in a small number of niches. As more sign ups come in there will be more people posting in a wide variety of niches that often don’t have an audience yet.
I think if you segmented the data by excluding bots, especially ones like tts that leave lots of comments, and by tag this might be more apparent.
There's also the calm after the storm. Obviously, with cryptomania that occurred end of last year into early this year, there were all the late comers to the game and they saw the mess that's on steem it. It's a bad thing if steem is trying to position itself as something that it wants to be, all the while it looks like something else entirely. The drop in organic comments certainly looks like a reduction of human interaction (for the worst), but that might just mean that the platform is going through troublesome times....enough trouble to need change. Better to hit rock bottom and rise than stay mundane.
"Steemit like all other social sites relies on the network effect. Currently there are just barely enough people on the blockchain to make things work in a small number of niches. As more sign ups come in there will be more people posting in a wide variety of niches that often don’t have an audience yet."
this is very important for people to understand.....
Hi @paulag I find your analysis on steem blockchain extremely useful for app developer. I see your analysis might avoid some fail startups if we can know where the demands truly are. I wonder if you can publish more about those kind of data.. Like usage of dapps among steemians.
I was doing them, but I stopped for a while cos I am very restricted on time (real world work commitments)
Thank you for the thorough analysis of comments per post as a measure of engagement on SteemIt.
Have a great day!
Steem on,
Mike
you are most welcome
I admit this does not look good in terms of engagement, however at least this data is visible; I would LOVE to see just how many comments and views other websites (such as YouTube) have that are made by bots, but they will never release that data
we so need something to compare too...
I'm not sure how we should expect organic comments to scale with posts, and I suspect some of your disappointment is from an assumption that it ought to be linear. I don't have data, but my impression of every other platform I've been on is that that isn't true: comments per post have been much higher in early adoption than in platform maturity. It's too bad we're the first people to be doing this completely out in the open, and don't have anyone else's data to look at.
it would be awesome to have something to compare too
I think part of the problem is that the current setup makes it hard to establish communities within Steemit. Sure, you can frequent certain tags and see some familiar people posting, but its hard to create a sense of community without some method of creating formal communities to bond it all together.
Hopefully a future release will enable tags to be secondary to communities, where people can gather and build passionate sections of Steemit that encourage better comments, more discussions, more teamwork and even starting some new friendships.
I cant wait for the communities feature, it think it will be a game changer
I believe as long as there are bots Steemit will lack credibility as a professional blog/vlog platform.
It will be a great day when a new Steem-based app comes along that is bot-free. Since bots are synonymous with money earners, let's not hold our breath on that idea. ;+)
Peace.
I feel it is time to leave a human (non-bot) comment 😉
lol thanks!
How could you tell if a comment came from a bot?
How do you remove greed from civilization? Maybe we are just going to have to settle with constantly changing and evolving. If there is one thing this data shows, its that the best at changing and adapting are the scammers. Work is never done. Full steem ahead!
thats really depressing....the best at changing and adapting are the scammers
I didn't mean it to be depressing. The constitution was a great idea in theory, until people began to exploit it and find loopholes for personal gain. Same with the federal reserve. In theory a good idea but it has been manipulated to the point where there is way more currency being created to keep up with inflation. That's the bubble no one in mainstream wants to talk about.
Depressing but not surprising in my opinion @paulag.
I agree with @abh12345. It would be interesting to see the figures if all the bots were removed (or as many as possible) including those that leave messages in our wallets for resteems.
Then I'd like to see all the comments made by people in yours and Asher's commenting Leagues and then see what we're left with. 😊
ah the gold ol message in a wallets. they are soooooo annoying......
I actually don't mind those @paulag. At least I get paid for them. 😁
What really annoys me is all the spam accounts that automatically follow me every time I make a post. I'd love to be able to stop them!
do you get a lot of them? I dont get to much, do it need a name and shame post?
I don't think a name and shame it post would work @paulag as they're either automated or the accounts don't care anyway.
Until today I was getting 4 or 5 follows with every post I made. All of them were spam accounts with a stolen video from youtube and no text. They also looked like they were related to each in that the accounts were created very close together and they seemed to get upvotes from the same accounts.
One day last week I got so fed up with it I reported about 20 of them to Steemcleaners but they contacted me on Discord to tell me they don't deal with videos.
At that point I just gritted my teeth and tried to ignore it.
So far today it hasn't happened. Maybe they've moved onto someone else.
When it happens it happens immediately a post is made.
Ha, ha, ha. Well that didn't last long @paulag.
These four accounts just followed me immediately I made my most recent post. Look at the number of accounts they are following.
The number of followers they have concerns me too. Since they are blatantly spam accounts then who are these followers? More spam accounts in the same ring?
lol...the rank and the follow/followers are waaaaaay askew!....lol...
Thanks to both you and paulag for raising the awareness on interactions...lol...I agree it's a small few in reply mode...
Now, let Steem/SBD increase, and the replies will do the same.
Btw, this explains with great clarity why the rank isn't keeping up with the followers. No voting power results in "dust" up votes which have zero affect in reality. Let's hope they are people who really appreciate our work...lol...
Eh, as with life, it is what it is. Make the best of it all always.
Peace.
i get these too!! probably 500 of my followers are these types of immediate follow accounts- i wonder who they are!!??
I do't know who they are @mountainjewel but they're a pain in the preverbials!
My followers tally is slowing creeping towards 1000 but it's pretty meaningless. I have no idea how many of those are people.
Of the ones who've followed in the last couple of months I would guess about one in ten.
This, for me, is the down side of decentralisation.
It's blatant spam but it appears that there's nothing we can do about it. 😢
It would be great if there was a spam button that we could hit and then the code could check certain things like number of followers, length of comments etc. before flagging it.
At the moment these accounts are really easy to detect. they follow very obvious patterns.
Wow. I knew bots are kind of taking over, but these numbers are depressing. By the way, I get a lot of good, well-thought comments under my posts. Makes me wonder why. Maybe because I don't make it to the trending section?
Thanks for the analysis!
hahah I don't make it to trending and I get loads of awesome comments
Thanks for another great analysis @paulag! I agree that there should be more engagement from human members of Steemit. I think some people are just to shy to comment. I try to make the effort to comment each post I upvote. And as for the Bots, when I use them for promotion I try to select only the ones that don't leave comments, but some other sometimes just appear.
hi and thank you for the comment, lol this post contradicts the analysis, look at all the awesome comments and they are all from people!
True! I guess you have inspired us :)
Nice work and data.
Your analysis feels about right. Steemit feels only about 15% human.
Steemit, the first decentralized blogging application for bots.
lol seem that way at the moment!
Interesting I do often wonder what impact the different apps have on increasing the number of posts per month. If you took a photo and just want post that there an app for that, you just want a tweet like blog hey there app for that. Go a meme—app it!
Shame it’s hard to quantify what is a decent to high quality post is and see if those are increasing or decreasing in number of comments they have received.
by the looks of the traffic on Dtube and Dlive, there is a lot of stuff from the apps. I need to revisit the data
I’ve always said that bots make this ecosystem less social. And the less sustainable in the long term.
Unless you befriend the bots, become one with the bots, assimilate.
Greetings fellow humans.
I will not assimilate.
Resistance is the default human state.
Sky net is real.
John Connor was right.
Calling all survivors to Join the resistance now.
OK, the most common topic on Steemit that is not machine manipulated data must be the Steemit is going down post... You really did a great job at documenting it :)
I do get a lot of good comments myself, but I have to admit that the botification is rather suffocating.
I have been doing another experiment that inveigled me review 500+ accounts manually. it was from there i noticed the level of bot comments which lead me to doing this analysis. I get a lot of awesome comments, sure just look at this post. I am disappointed that i am in the minority
I just looked it up. This platform has some amazing people (which means I find you amazing). I have actually started auto-voting for the profiles I know make good content. That way I am able to give small donations to some of the profiles I can't check every day. But the workings of this platform is not necessarily the best for quality posts and human interaction.
I get very few comments from bots, so maybe I'm an exception. I also get a very low rate of spam comments, so I'm fairly optimistic about the state of steemit. I wonder how it will change as the next million sign up
I get very few comments too from bots, but we are in a small few. Not feeling as optimistic today....this was only a small sample of bots.
I will not try to address the number of comments, but I've always been pretty lucky that way!
However, regarding the quality of comments I'm noticing a drastic change. I am no longer getting broken English comments from those who obviously aren't reading. I am getting thoughtful and real engagement and that feels great. I wish there was some way to track that! :)
One thing I noticed early on here - I don't think I'd found you yet, even - is that authors who consistently reward quality comments with upvotes have much better-quality comment sections than those who don't.
Obviously this is a large minnow/dolphin privilege, but as we get more posters into those groups (partly by upvoting their comments) I think the overall quality of comments on the platform should improve. That may already be part of what you're seeing.
oh I wish I could track that too, because I would rather post good news data than this!
Great comment!
Just kidding. 😋
The serious part of my comment is, that when you receive quite some comments, you must also receive some of these, don't you?
I still wasn't clear on your meaning. 3 months ago I would get 15-20 comments and a small fraction of them would be "Engagement" The rest were so difficult to read, I couldn't tell if they were real attempts or not. Currently, I am getting more comments and the vast majority are real, well spoken and make sense. The conversations continue!
I meant that there are a lot of humans that write the following comment:
Great post!
If you also deduct these nonsense comments you come to the conclusion that the engagement on Steemit is even lower 😥
I get awesome comments, I am one of the lucky ones that have real people, like you, engaging with me
About engagement/comments I have the following hypothesis:
The Steemonians that receive the upvotes are also the ones receiving the comments.
So basically 1-2% of all Steemonians receive 90% of the upvotes as well as the comments.
Most likely the percentage for comments is a bit lower, because some red fish and minnows receive more comments then upvotes.
Quality is harder to track via a script.
I'm sure if we looked at the histogram, there would be a small number of posts that are highly engaged with. That may give a hint of quality of both the post and the comments.
Just look at the ammount of upvotes you've given. (that would fall in the inciteful category) If you are someone that usually replies to well made comments then look at the ammount of replies you gave.
stats done.
I would not fully agree, I don't give to many upvotes because I bagatelle out a lot of power. lol last week i ran out of bandwidth and couldnt vote because all my power was delegated (for free by the way, not for profit)
Well. If you were someone that upvotes comments you think of as quality and respond to comments that bring induce such feelings. Then it would be accurate would it not?
It's a sad evolution. People can say and think about the bots all they want. If you add up all analysis you have done it is hard to deny that a big part of this blockchain is fake somehow.
till today i was okay (ish) with the stats, okay the retention and churn bug me, but this is crazy, and I only took a few bots.
Thank you @paulag, for organizing a league for the Indonesian community, it is very helpful to improve the engagement, now I see more comments and long discussions in a post, it is good and fun :)
fantastic, thank you for sharing that, its the sort of news I like to hear
That would be bad enough if those were the only bots, but there are quite a lot more. It may only be 10% human commentary at this point.
:-(
Data analysis is very complete, as long as there is a curation project that you run, it is very useful for the development of steemit.
Looking at Steemit in general the number of comments per post has generally decreased since January 2017 and the % of posts with 1 or more comments averages at 58% and with 5+ or comments is only 20%.
heheh if you get 5 or more comments your post is in the top 20%
@randomwhale is not a bid bot, it's a dead sockpuppet/scam account. Is that just a typo in the post or was that part of the query? The correct account is @randowhale
feck i pulled it in, guess really then were are looking at a sample of 11
It looks to me like the correct account was included after all, the query doesn't have 'randomwhale' included.
if you have a look at the queries she has included all of them: randowhale, randowhale0 ,minnowhelper, randowhale1. Randomwhale is dead now, but not in the past. Though I think it made no comments.
You are right, it was only in the post text it seems.
As far as I know @randomwhale was never a vote bot. It was purely a scam, seeking to make money from users making typos.
Which fits nicely with my guess of between 100-200k real unique signups.
Would be cool if you could add more bots and put the last graphs out again. What would it look like with 500 more bots added - oh dear.
Very cool analysis boss 😁
cool but not good @abh12345. Don't know why this has disappointed me so much!
because we are all here on steem and we like it and dive deeper into that rabbithole. We all see the potential of the whole thing but then there are these stats that prove that we might be living in our own little bubble and that when it pops there might be no more steem.
I think that the only thing that is covering this up slowly is the rising popularity of Dtube and co and hopefully it will not be to late
Dlive is rocking too. today i'm feeling like the bubble may have burst for me. I only took a few bots into this, so the reality is a lot worse.
It really depends upon your personal network on Steemit. In our circle of artists, we have a high engagement level. The human comments outnumber the bots.
Interesting.
I bet if someone got deep into the data, they could train a Neural Net to discriminate between human and bot posts, just due to the variability in human responses versus the usually-static nature of bot comments.
In any case, it confirms partially the suspicions I've had about true engagement here.
I appreciate your analysis.
This is insanely interesting.
As a matter of fact, I am seeing this place to become something else entirely, because of bot activity.
At first sight, this could be awful to look at, but in the last few days, I am starting to notice a certain pattern here. Like a change of dynamics. Like a website with a different kind of movement. Not bad. Different.
I think Steemit is slowly but firmly losing its human factor, and it is replaced by automatic behavior. Upvote selling, bot activity, commercial delegation.... So, maybe this would be the new parading around here. And maybe it is not all that bad.
Imagine this:
The new user would have to be ready to engage in certain activities that give him tokens he can use to buy or bid for upvotes that only bidbots can give. Or getting SP for a fee. These tokens would be granted by posting content, sharing it, getting views or clicks, bringing new users and commenting, in an automatic way (Not by human upvoting). The only way he can get SBD by upvoting is in the case he upvotes his own comments and content.
The other users can upvote (sort of) of flag your content, but that does not give you or take from you any SBD. It affects your reputation. The bigger your reputation is, the more you can earn for your activities on the site.
The bidbots will be regulated, limited. Maybe they would pay a fee to the site for the right to upvote with monetary value. A place on the trending page would be bought. You pay for exposure, and everybody knows it. No more hidden stuff.
Just some wild toughts. Maybe it is not that bad.
could you imagine telling the bidbot owners they will now be regulated lol....that would be funny
Steemit all the way
Very telling how useless much of the activity and so-called "engagement" is from comments overall :/
Thanks brother for sharing a good information for us.
Also hope to see those whales redistribute their votings to promising minnows instead of autoving those with established reps already. Comments may increase in this matter but just my speculation.
Hey @paulag
Thanks for contributing on Utopian.
We’re already looking forward to your next contribution!
Contributing on Utopian
Learn how to contribute on our website or by watching this tutorial on Youtube.
Want to chat? Join us on Discord https://discord.gg/h52nFrV.
Vote for Utopian Witness!
Fascinating, thank you.
Great analysis for the bot activity. It validates my feelings if i look into my posts. Recently it felt i had a whole bot army with always the same annoying, senseless, stereotype comments. Often very short, not content related, sometimes longer to work like a smoke grenade but after several times clear that's a deceptive approach to earn money out of the reward pool. As i wrote in my suggestions in @lukestokes post about long term thinking for Steemit we need something in the code with the HF20 to reduce or minimize these activities because we cannot hope that people stop with that behaviour what is offered by the system.
Thank you for the insights!
yep we need to write something into the code, but we have to get it right...
im rather lucky I do not have a bot army commenting on my posts but I do know some that have