Saturday, April 20, 2013

Guy Kawasaki Deletes his Tweets while You Sleep

I had to learn the hard way that Guy Kawasaki deletes some of his tweets.  The data I use for this blog is diligently collected by a little script I wrote.  I am not a great coder, so when Twitter started telling me that some of the tweets I had tracked the day before no longer existed, I blamed my coding skills.  A few nights later, I got lucky and left a browser tab open with one the tweets that would eventually disappear that night.  So, when I woke up the next morning, I could prove to myself that I wasn't making up these tweets.  Kawasaki was deleting his tweets!

Now, I was curious why Kawasaki would delete some of his tweets.  The chart below shows the percentage of tweets deleted by the hour when the tweet was posted.  Based on the time of day when Kawasaki is more or less likely to delete his tweets, my first hypothesis was that he uses these deleted tweets for testing purposes.

One of the most surprising insights of the data above is that Kawasaki was more likely to delete a tweet than he was to keep it.  In our sample, 66% of the tweets were erased.  In addition, you can see that the hour when a tweet was posted tells us a lot about whether a tweet will be deleted or not.  Between midnight and 5 a.m. Pacific Time, all tweets will be erased.  Then, we see a big decrease during the next two hours in the probability that tweet will be erased.  Across the day, the probability of being deleted increases slowly, with the exception of a few sharp decreases like the one at 6 p.m.  Finally, between 10 p.m. and mid-night, the probability of Kawasaki deleting a tweet increases significantly to 92% on average.

We will discuss other reasons why I think Kawasaki deletes his tweets in future posts, but for now, I will talk about my testing hypothesis.  Tweeters can predict how popular a tweet will ultimately be, as measured by number of retweets, in the first couple of hours of the tweet’s life.  So, if you want to make sure that you tweet the information that will be most popular, you could test your tweets on a subset of your audience before sending  them out to your whole audience.  A good way to segment in twitter, is by time of day.  The night-owls that are checking twitter at 2 a.m. won’t be doing so as frequently at 2 p.m. and vice versa.  Therefore, I believe that Kawasaki tests his tweets on his audience of night-owls in the western hemisphere and if the tweets are successful, he deletes them and tweets them again for his daylight audience.

To finish, I have something easy for you to do if you want to test whether Kawasaki really deletes his tweets.  Check Kawasaki’s twitter account on any day between midnight and 5 a.m. Pacific Time.  Click on the details of any two tweets in different windows.  Then, reload both windows 12 hours later.  I am very sure that at least one, if not both of the tweets, will be gone when you reload.  (I ask that you use two tweets instead of one because I want to hedge against a tiny probability that Kawasaki doesn't delete his tweets between these times).  Let me know what you find!

UPDATE: April 22, 2013: In a recent Google+ thread, Kawasaki was asked whether he deleted any of his tweets during a particular date.  He responded that he "...did not delete any tweets yesterday or today--though Twitter may not display them all. [He] can't remember the last time [he] deleted a tweet.".  Therefore, my statement that he deletes his tweets might be premature.  The missing tweets might have another explanation. More analysis to come in future posts.


  1. Here's a hypothesis for you to check out if you're inclined: Perhaps when Guy puts out two identical tweets, Twitter only keeps the latest one. I put it that way because the only tweets I've been able to show were missing were replaced later by identical tweets.

    Is it possible that twitter is acting up in that way? 

  2. Max,

    Your hypothesis seems to be correct. Guy Kawasaki's reposting behavior seems to explain to a large degree whether a tweet goes missing or not.

    You can find all the details about the relationship between reposting and missing tweets in my latest post here: