Saturday, July 11, 2009

When Feedback Loops Go Missing

I've been noodling around a couple of thoughts for a while now about Twitter - one having to do with what I'll call a Publisher Rank (similarity to PageRank intentional). The other has to do with the meaninglessness of Twitter social climbers who follow too many users to ever possibly read their tweets. The two thoughts go hand-in-hand and their relationship is interesting. But I avoided the topic because it had nothing to do with design. Plus, my last post was about Twitter.
However, something occurred to me as I was unfollowing someone because he was too chatty. I realized that the user probably had no idea I was unfollowing him - and most certainly would not learn anything from my unfollow. Of course Twitter does not notify for 'unfollow' operations. (Twitter users can employ services like Twitterless or Qwitter to figure out they have been unfollowed. But most don't use these services.) And compounding the problem, users think they have no reason to look for unfollows because their follow count keeps going up. After all, they must be doing something right since their follow count keeps rising. Right?
The fallacy in this is two-fold: 1. The Twitter population is growing so fast that everyone's count is going up just as people try out the service. 2. The social climber doesn't care if you tweet too much. After all who can follow thousands of tweeters anyway - just ignore everything and follow to be followed.
The bottom line for every publishing twitter user is that the feedback loop is quite broken. And this is the loose tie-in to design. Twitter publishers have designed their behavior to draw in followers. But the feedback loop for validating their design is essentially missing.
This can be fixed with some effort.
1. Watch for 'unfollow' operations. You don't have to obsess about every one. Just keep track of how the number correlates to how much you tweet. Or if you're really ambitious, there may be something to learn about your tweeting style.
2. Rank yourself with the Publisher Rank (see below). If all of your followers are social climbers, you'll have a rather pathetic score and may want to do some soul searching about the design of your tweeting plan.

Defining Publsiher Rank:
First, we have to define Follower Rank. This is the rank a user has as a follower - effectively his value to the publisher. A follower who has few tweets in their inbound stream is worth more to the publisher because she has fewer tweets to read and can spend more time on each (perhaps even buying what was tweeted about). A follower who follows many users is worth less to the publisher because he has less time to read each tweet. And in extreme cases may never see a tweet. For practicality, let's call 100 tweets per day for a single user a Follower Rank of 1. This is arbitrary. But it is a constant we can use throughout the calculations.

Using this constant, a precise Follower Rank for a user can be obtained by reading their follow stream and counting the number of tweets in a 24 hour period. Divide 100 by this number to get their Follower Rank. If they get 200 tweets in a day, their Follower Rank is 0.5. If they get 5000 tweets in a day, their Follower Rank is 0.02. Remember, more tweets in a user's stream means less time per tweet to read (and act on) the stream.

Estimating Follower Rank
According to HubSpot the average number of tweets a user publishes is 4.4. Let's use 5 to keep the math simple. Using an average, we can simplify the Follower Rank calculation by multiplying 5 times the number of users a user follows. A user follows 20 users, he gets a Follower Rank of 1. A user follows 1000 users, she gets a rank of 0.02 for that user.
Calculating Publisher Rank
Next we calculate your Publisher Rank by adding up the Follower Rank for each of your followers. You can use the precise method (today's tweets for each user) or the approximate method. Either way, the sum of these is your Publisher Rank. If you have 1000 followers each with a Follower Rank of 0.002 your Publisher Rank is 2. If you have 100 followers each having a Follower Rank of 0.5, your Publisher Rank is 50. In the latter case, you have a much higher rank because your followers have greater capacity to read your tweets.

For those who prefer equations to prose, here is the formula for the precise method of computing Publisher Rank:

P represents our publisher we want to rank
Fj represents an individual follower of P
Gji represents an individual follow of Fj
k = the count of followers of P
nj = the count of follows for an individual follower Fj of P
mji = the count of tweets for an individual follow Gji of Fj

It's left as an exercise to the reader to implement a service that computes Publisher Rank based on this formula. Maybe I'll work on it in my spare time.

Any thoughtful reader will come up with flaws in this ranking system. Not all Twitter users are active. Some people aren't even online much. Others spend their time in other social network contexts, etc. To see it taken to an extreme, check out this post. In fact, I would argue that any system that can track actual follow-through (clicked linked, purchases, etc.) trumps the guesswork about how to rank a Twitter user.

Still, contemplate this: If we all used Publisher Rank instead of Follower Count to rank Twitter users, a bunch of social climbers would simply vanish from visibility. I could live with that. Could you?

No comments:

Post a Comment