Followers, following, and false fellows

I've been running my Music Suggestron twitter bot for more than a month now. That didn't take much effort as bots tend to do most of the work on their own. This bot doesn't require much user input, either: it just spits out randomized tracks and genres available on Spotify. It was born out of my desire to listen to something new every day. Spotify claims to have 30 million tracks in its database, or enough for more than 171 years of continuous listening (assuming an average track length of 3 minutes, a value based on Conventional Wisdom but increasingly untrue), so there are numerous options.

The project became a kind of metagame for me (at least, if writing the bot itself was already a game and the Twitter component provides the meta). I've never had to promote anything besides myself on a social network. My bot didn't have any identity or purpose outside Twitter, so how could I increase its visibility? It's not designed to spread marketing buzz or sell products so, again, promotion was just for my own entertainment.

I set my initial goal at attaining 1,000 followers. I started having the bot follow other accounts, starting with anyone in the music industry but later focusing on musical artists, DJs, producers, songwriters, vocalists, and instrumentalists. In theory, these would be the people most likely to see the bot's tweets anyway should it post a link to their work. Here's what I learned.

It's not easy! Twitter limits the number of "followings" for any one account to 2,000, give or take a bit depending on your following/follower ratio. I hit that limit with a follower count of ~850. Without the ability to follow new accounts, my follower count slowly atrophied. 

A noticeable proportion of Twitter users exist for promotion only. They could be called fake; about 10 percent of users may be fake. "Fake" is an ambiguous term when we consider that a fake Twitter user could be masquerading as a real person, could be a real person solely posting links to counterfeit sunglass stores, or could just be a bot. Musicians actively promote themselves and others on Twitter so it can be difficult to distinguish a real human from an automated marketing machine. Two non-musical examples:

I'm going to assume that the fictional Sheldon Cooper isn't posting context-free quotes (are they even quotes? I don't watch Big Bang Theory) several times a day, but this account isn't posting horse_ebooks-style spam links either. It's benign.

This account may have a real person in there somewhere and the associated tweets include some authentic-looking text (e.g., quotes from Winston Churchill). Recently, it's just promotional retweets. The activity isn't harmful but it's pure advertising. 

So why am I focusing on fake Twitter accounts? Isn't every social network profile an exercise in brand building? Does it even matter if the "self" in self-promotion isn't a single human being?

It's not difficult to automate social network activity. It's certainly much easier than automating other social interactions. This is exactly why I have difficulty taking Twitter seriously as a venue for useful, productive communication. Too much of the noise in the system is automated, devoid of content, and impossible to trace back to a message other than "your voice was heard".

Promotion lacking human input may not be worrisome when the product is music or art, but other fields require more personal communication. Science is one such field and Twitter is increasingly gaining acceptance among scientists as a medium for professional communication. This is a new phenomenon. Should I start setting up bots to re-tweet my new publications if I want to widen their audience? All the open access policies in the world won't help broaden my work's reach if the loudest voice wins anyway. Or, perhaps louder isn't always better.

Twitter analytics could be more helpful. I was excited to find out that Twitter offers a full buffet of metrics for how users interact with your account. They're what turned it all into a game for me. So many numbers to track!

It's difficult to complain about the service as it's free and I haven't paid Twitter a fee to promote anything for me, but I'll complain anyway. The Analytics interface lacks obvious options like accessing individual tweets or even seeing who has retweeted a particular message. It's all presented as averages and summaries, potentially to reduce database stress, but I'm just speculating. 

Here's my impression trend. Twitter defines an impression as any single view of a tweet, even if it's just scrolled past.

The blue values on top are impressions and the lower values are total tweets. New impressions from tweets made on previous days count toward the day they're made. My bot usually tweets about 140 times a day. Twitter insists on double-counting that total sometimes so there's a spike near the far right. More importantly, the bot's total impressions tapered off immediately after it hit the follow limit of 2,000 users. This suggests that most impressions are from new followers, as following users encourages them to follow back.

I won't bother showing the graph of followers as it's either massively inaccurate or includes unexplained values. It may not properly account for lost followers.

This bot is a fun programming project so I'm going to try taking it in some new directions. Perhaps it should have moods. Perhaps it should provide lyric samples. Perhaps it should violate copyright and generate mashups on demand. We'll see what happens.

In the meantime, check out some of these bots: