No one likes spam. So,
when people discovered that Guy Kawasaki had a whole team tweeting for him,
the spammer
accusations started flying. One of
Kawasaki’s tweeting strategies that was labeled as spamming was his practice of
reposting the same tweet multiple times. Kawasaki does this so often that 78% of his tweets are part a series of reposted tweets. I originally thought that with every repost
the number of retweets that a tweet received would decrease. In practice, the
original tweet and subsequent reposts receive the same number of
retweets on average . Reposting tweets is a
great way to target more of your followers. If you only tweet once, you might only target 25% of the audience you could have reached.

Reposting tweets lives in a grey area of Twitter law. For
some time now, Twitter has stated that posting multiple
duplicate tweets on one account is considered spamming, which is not allowed. However, Kawasaki gets away with reposting tweets by using the same content but changing the URLs, which all point to the same location.

To determine whether tweets that were reposted had a decreasing number of retweets, I compared the number of retweets of the first
through fourth tweets in a series of reposted tweets. I chose to look at tweets that were reposted 4
times because Kawasaki repost his tweets 4 times in 98% of cases. The chart below shows the distribution of retweets 15
minutes after posting

^{1}for tweets with a repost count of 1 through 4. The repost count describes a tweet's order in a set of reposted tweets. The tweets with a repost count of 1 were the first in a set of reposted tweets, tweets with a repost count of 2 were the second, and so on.
Looking at a distribution of retweets provides much more information than just looking at the average number
of reposts. For example, tweets with a repost count of 1 have on average 4 retweets, and tweets
with a repost count of 2 have an average of 3 retweets. If we
were only looking at averages, we would conclude that the first tweet in a set
of reposted tweets received more retweets.
However, we can see from the graph that the 25

^{th}, 75^{th}, and 95^{th}percentiles for both types of tweets look the same. Similarly, the distribution of retweets for repost count categories 3, and 4 look a lot like the distribution for categories 1 and 2. So, the number of retweets for all repost categories is very similar. If we wanted to get very rigorous, we could even say that the average values for the four distributions are not significantly different from each other^{2}.
The similar number of retweets across repost
count categories might be due to followers not scrolling back to read tweets
that are earlier in their timeline or that they live in different time zones. Either way, even if you don’t like spam, this might a
case where spam might be good for you.

^{1}I chose the number of retweets 15 minutes after posting because, as discussed in a previous post, 66% of Kawasaki’s tweets are deleted in the first 12 hours after posting. Even though using the number of retweets 12 hours after posting would provide a more representative value of total retweets, this length of time would also remove a large percentage of the sample because the tweets would be deleted. In addition, as we will discuss in a future post, the number of retweets 15 minutes after posting is a good predictor of number of retweets 12 hours after posting.

^{2}A one-way analysis of variance (ANOVA) of the four distributions determines that there is no statistically significant difference between the group means (p = .23 ).

## No comments:

## Post a Comment