Why no one reads you on Twitter

I have been on Twitter since 2008. I have also used Facebook, Livejournal, Orkut and others. With Facebook and Livejournal I used to have a wide group of people following me. My posts were read, discussed, and commented on. Nothing of this happened on Twitter, and I think I now know why. I have about 1400 followers on Twitter, more than I ever had on Facebook, or Livejournal. But the Medium is different. It works differently, and not just for me. I have seen other people lamenting this too. I read people with 10 thousand followers saying that Twitter was more alive for them when there were few people on it. Why? I am about to answer you.

By the way, I teach a university course in Big Data at Lumsa University in Rome. And as part of this course, we cover social networks and e-Democratic tools. And the main message I try to convey is: technical decisions have social consequences. We usually observe how people react and interact on social networks, if they can comment, like, dislike, reply-to-comments. And we see how the way social networks let you interact has an effect on the experience of the users, and how this creates different societies. How societies polarize, more or less, depends on the tools you let your users utilise.

There are many reasons why Twitter recently sucks for most users. But it mostly boils down to too few people will read their posts. Why?

Twitter has two main ways to present you the stream of posts. Either time-wise, or popularity-wise. We shall call them Time-Stream and Pop-Stream. The common user does not know that the two types of streams are available. The button to switch from one to the other is carefully hidden in plain sight.

You see the little star on the left of the Search bar, that is a button that let you choose which kind of Twitter Stream you are using.

You see the little star on the left of the Search bar, that is a button that let you choose which kind of Twitter Stream you are using. If you are using the Time-Stream, over your icon it will say Latest Tweets. If you are using the Pop-Stream, over your icon it will say Home.

Even the way they let you know which kind of stream you are using, seems to be designed to hide the dual nature of the stream. But let’s dive into the details because technical decisions have social consequences, or the devil is in the details.

The Time-Stream pretty much works as you would expect it. It will show you the posts of the people you follow, as a time series. You see their tweets and the retweets of the posts they retweeted. And then some advertisements. In my case, this was about 66% of Retweets, 30% of Tweets, and 3% of Advertisements. I do not know if the reasons why the retweets are one-third, the posts three-tenth, and the advertisement the space in between is just a weird coincidence due to the special mix of people I follow (after all this should be the whole set of people I follow) or some engineer decided this mixture was good for me.

Now, let us look at the Pop-Stream. This wants to show you the most important tweets that have been posted recently. This is a really difficult thing to do, engineer-wise. You must juggle between a time element and a popularity element. How many likes, comments and retweets are needed for something to be important enough to appear here? The exact receipt is obviously an industrial secret. And Twitter is not sharing this, probably trying to avoid its competitors from doing the same errors. While we cannot know the exact formula, we can look at the result and compare it with the Time-Stream.

In the short statistics, I could gather (which I think is pretty representative) the number of tweets from people I follow now drop to around 15.4%;
the number of retweets to about 1%.

This was 96% of the material that was available before, that now is only 16.4%. This is what I have signed to follow, the people I declared I am interested in by consciously following the, and it forms 16.4%. And the rest? The rest are posts, not from my friends, but from people followed… by the people I follow. The people at distance 2 from me, which I have no control over.

In particular:
34.3% are posts liked by someone I follow;
43.3% are posts of people followed by someone I follow;
And then 5.5% are people I follow who are replying to a tweet.
In this modus operandi, no advertisement seems to appear.

Let’s make an example to understand those numbers better:

I follow Adam. I think Adam is a cool guy, and I want to know what he posts. Adam writes a post, then reads a post from Betty, a post from Claire and a post from Danny. Now Adam decides to retweet Betty and likes the post from Claire. He also ignores the post from Danny.

So we have 4 posts:
Tweet A (from Adam),
Tweet B (from Betty + Adam who retweeted it),
Tweet C (from Claire, liked by Adam, but not retweeted),
Tweet D (from Danny, ignored by Adam).

What do we see?

15.4% are posts of type A,
1% are posts of type B,
34% are posts of type C,
43% are posts of type D.

But A and B are the only posts I should see. C and D are introducing material inside my stream I never wanted or signed to see. And that the people I follow never took the decision to share with me (by retweeting it).

Of course, if in my stream I only see 16% of posts from people I follow, those are the posts more popular. This means many posts from people I follow are never shown to me. And, for symmetric reasons, many of the posts I write are never shown, even to the people that follow me. To see them they would have to click on the notification bell (but it gets overwhelming pretty soon) or watch the Time-Stream. Which the great majority of people do not even know exists.

So what are the posts that are shown to you? Those are the most popular posts. And this makes those posts even more popular. De facto this system makes it harder for a post to go viral for the average person, but when a post goes viral it spreads it much wider. Instead for the VIP people, the ones with hundreds of thousands of people following them, this makes sure that each of their tweets goes viral.

Have you noticed how on television we always see the same people discussing? In Italy, a nation with 60 million people, the number of people invited to the talk-show is maybe 200. We know them all. They cannot be much more than the Dunbar number. A social network should not be like this, it should permit people to discuss among themselves. Not just be an arena where VIP people are overwhelmed by comments from people, and normal people are unable to speak between them because their friends are only hearing the loud noise of the viral posts around them.

Can we fix this? Surely you can choose to follow the Time-Stream instead of the Popular-Stream, but this will fix what you see, but not what the people near you see. You will still be immersed in the experience of a society which cannot hear you because your posts never make it to their ears.

Technical decisions have social consequences. I hope Elon will fix this, but good intentions are not enough.

I started by saying that I teach Big Data. And in this course, I teach both social networks and e-democratic systems. Which are very similar, except for one reason, the first one tends to polarise society, and the second tends to make it converge (when it is done well). I hope Twitter becomes more of the second type and less of the first one. It’s not impossible, just not trivial.