What’s in a Retweet? The Data Behind Viral Messaging on Twitter







I started collecting ReTweets a few weeks ago and have collected just over 84,000. I’m working on a system that will allow for mapping and analysis of ReTweet streams (sneak peak below), but in building that, I’ve already uncovered some interesting data.

Contrary to what I initially thought, “RT” is used more than 4 times more often than the full word “retweet”.

ReTweets occur at an average rate of around 258 per hour, and show a distinct increase during the work day and early evening.

Retweets contain the word please over 5 times more often than most tweets.

Retweets are generally longer than other tweets.

Almost 70% of ReTweets contain a link.

Tinyurl is overwhelmingly preferred as the URL shortener to use in ReTweets.

Let me know what other data points you’d like to see and I’ll see what I can do.

And here’s a very simple, very rough preview of the mapping tool:

If you liked this post, don't forget to subscribe to my RSS feed or my email newsletter so you never miss the science.

{ 45 comments }

Stuart Foster December 5, 2008 at 10:44 am

Simply awesome Dan. I was looking forward to this post when you first told me about it. I’m really fascinated by the best method to retweet and promote other’s content on twitter and this answers a lot of the questions that I had. So first off RT looks to be far superior to Retweet (in terms of both saving length and effectiveness). I am interested in the use of the word “please” in the retweets…looks like I will no longer be using it. I’d rather be more effective with my promotion (by separating myself from the crowd). Thanks Dan…damn useful stuff.

Liz December 5, 2008 at 10:56 am

Whose Tweets did you look at, the ones you follow or the public timeline? Just Tweets in English? I’m curious about how representative your sample is.

The findings are interesting whether it is a random sample or not.

Dan Zarrella December 5, 2008 at 11:11 am

The retweets are all the retweets I could find in the public timeline.

Dan Zarrella December 5, 2008 at 11:11 am

and yes, just english tweets.

Simon Owens December 5, 2008 at 11:25 am

Maybe some kind of ratio to show the number of retweets one gets when he begs for them? Is this similar to reciprocal digging?

Taylor Host December 5, 2008 at 11:38 am

This is fantastic. What is your analysis on the frequency of “please?” I’m curious to the implied significance.

Keep up the good work – Thank you!

Dan Zarrella December 5, 2008 at 11:40 am

@taylor People often say things (or at least i do/have) like “please retweet.” I think at least on some level occurrence of the word “please” implies that the user is asking for the retweet (or some other action).

Melanie Phung December 5, 2008 at 11:42 am

Great sneak peek, Dan. Here are some additional questions:

Chart 1: In addition to “retweet” and “RT”, what about the variations “R/T” and possibly “re-tweet”?

Chart 2: Would like to see retweets by hour graphed against regular tweets. Would be more meaningful if we could see if the pattern differed from non-RT activity.

Chart 4: What are the units of the X- and Y-axis? (#characters and #total tweets, respectively?) Maybe make averages and percentages clearer too?

Chart 6: There’s no “other” category? Instances of cli.gs, zi.ma and the like are statistically non-existent?

What about an analysis of tweets/retweets ratio per user? Eg., Avg. user’s tweetstream consists of 5% retweets. X% of users retweet Y number of times per day. Etc.

Dana December 5, 2008 at 11:43 am

I’m guessing “please” is there because of the number of people who end their tweet with “please retweet?”

So people are more likely to respond if asked than just because they really wanted to share?

Dan Zarrella December 5, 2008 at 11:48 am

@melanie thanks for the input.
1), 4) and 6) I can do, look forward to some data/graphs on that.
2) and 7) (where’s 5? ;) ) to some degree require more general data than I can pull from Twitter, though in some cases I can kind of get close.

Beth Kanter December 5, 2008 at 12:08 pm

Wondering if there is any way to get some word analysis or content analysis of what tends to get retweeted?

Terri December 5, 2008 at 12:09 pm

Great analysis. Are there days of the week that Twitter posts are most often retweeted? I’d imagine M-F but is there any one day that stands out more from your data?

Beth Kanter December 5, 2008 at 12:10 pm

also who is retweeting? How much does the person who retweets play in this? Number of followers?

Bruno December 5, 2008 at 12:18 pm

Hi Dan,

Great work. How many tweets are you seeing per hour? What fraction of that are RTs?
Thanks!

Dan Zarrella December 5, 2008 at 12:20 pm

@beth, yes word & content analysis is coming, I need to map the tweet streams top do that effectively, so its taking a bit longer. Same for user to retweeting stats.

@terrir i need to be collecting data for a longer time to be able to do day of week, but i will when i can.

Warren Sukernek December 5, 2008 at 12:26 pm

Very interesting analysis! In the spirit of the analysis, I retweeted your post and someone retweeted mine! A few questions: If the please is the request for the retweet, then wouldn’t a truer picture of RT’s remove those with please in them? It would be great to know, the average number of retweets a post gets and how viral on average a post is (in other words, how many people RT a retweet).

Great job!

Melanie Phung December 5, 2008 at 12:58 pm

Dan said: “(where’s 5? ;) ) ”

I didn’t have any questions about Chart 5. It told the story of the link-v-no-link data perfectly exactly as it was – nothing further needed ;)

Dan Zarrella December 5, 2008 at 1:03 pm

oh, i get it, you were listing notes by chart, sorry.

Monica Wright December 5, 2008 at 1:05 pm

Wow, this is really a great analysis. Something to keep in mind generally a RT includes an @twittername (sometimes 2) which may impact the tweet length. Not that it matters too much. (Wow, I feel really dorky suddenly). Still, great post.

Dan Zarrella December 5, 2008 at 1:10 pm

@monica thats a great point, my mapping system is able to remove most of the “trappings” of retweets, like rt/retweet, via and @s related to those.

Ed December 9, 2008 at 5:41 pm

Thanks for collecting and graphing.
The need for a kidney donor, and the 2nd to last Amber alert were 2 of the biggest retweets in the last half of 08.

Also, “RT” wasn’t even close to “Retweet” when Twhirl first launched. (We were already using it, but the one click was huge) Then folks figured saving the characters was cooler, and typing 2 char’s. was easy.

Marc Brooks December 10, 2008 at 3:06 am

All my retweets are done as (via @foo). Are you picking those up, and are you going to do some stats on them?

Brian Carter December 12, 2008 at 11:54 am

Dan Zarrella, someone needs to have your baby. You freakin rock the Social Media science world.

Anton December 12, 2008 at 12:17 pm

Nice article! Interesting charts!

I wrote a blog post about this phenomenon for a couple of weeks ago, with quite interesting discussion. Interested? Check it out: http://blog.twingly.com/2008/11/24/the-microblogging-phenomenon-retweet/

… which also give us a good idea: could we use retweets as a parameter of influence or ranking critera to our blog search engine? Probably! Now it’s a suggestion in our open tech plan (need to log in to see it, I think): http://techplan.twingly.com/feature/?id=123 .

Jim Wolff December 12, 2008 at 1:26 pm

Interesting post.

Is there any way you can tell how far a Retweet travels. As in, how many people a single tweet is passed to?

Dan Zarrella December 12, 2008 at 1:58 pm

@jim yes, I’m working on that (the mapping system thing)

@marc I’m finding tweets that use via in r/t streams i know about, but not using it as a keyword in discovery. I was having problems with it before but its something I’ll need to address.

@ed great data, where’s it from?

Peggy Dolane December 13, 2008 at 7:57 pm

Great material, Dan. Thanks! Wondering why budurl or tweetburner aren’t showing up in the link shortening data. Tweetburner’s got to account for a significant number (or at least if you believe their link shortening hype they post on their site.)

Along those lines, would love data about link tracking tools. Are they worth it and what do they really tell you?

Andrea Hill December 14, 2008 at 12:07 pm

Very interesting! I too think that RT makes sense from a “character-saving” standpoint. I know twhirl uses “retweeting” and that’s just too many characters!

One thing I’d be wondering about recently (and may post on it myself yet) is retweeting etiquette. If something gets retweeted 3 or 4 times, do you “source” the originator, or the person you found it from? Doing both may be prohibitive due to the character limit.

PRJack December 15, 2008 at 2:36 pm

Interesting, but not wholly unexpected. The additional variable is ‘level of user experience’. Many folks who start on Twitter will try to spell things out – I know I did! But quickly we start to adopt the shortcuts we see in use. The more one becomes aware of the 140 character limit, the more one recognizes these shortcuts.

Also, I’d expect more occurrences of ‘ReTweet’ (or variations thereof) when someone makes the initial request for followers to perform that action, and for the use of RT to be favoured by those doing the retweeting.

Jesse Liebman December 16, 2008 at 11:08 am

RT is an essential piece of having an effective Twitter strategy. Here’s what I’ve applied so far, including the infamous RT:

http://www.jesseliebman.com/2008/12/15/captain-planet-meets-twitter-strategy

bansi December 16, 2008 at 9:57 pm

As a web analyst and Twitter lover, this post is just awesome!

Molly Gordon December 20, 2008 at 2:38 am

Awesome. And I’m interested in data about the number of iterations of a given Tweet/RT. How many Tweets are RTd twice? Three times? Ten times?

Thanks!

Gerald Weber December 23, 2008 at 8:07 pm

hmmmm. A lot of thought going into retweeting. he he. ;)

realmatrix December 30, 2008 at 12:51 pm

I stumbled on this through a tweet or retweet by @niceprof. There is some very interesting and salient information provided for those of us who use twitter on a regular basis. Thanks

Santosh Puthran January 29, 2009 at 6:34 am

Dan,

Very informative article. Did you design the ReTweet This widget ? that’s really cool. I would also like to have such widget on my blog.

Regards,

Santosh Puthran

fleurbrown January 31, 2009 at 9:13 am

How can you tell if someone has retweeted a tweet (other than staying online to check). Is there a widget for that?

Nicole Simon July 12, 2009 at 6:21 am

Nice Charts Dan, though I am missing something:
An additional point that Twitter is a world wide used tool and that your graph on time is most likely spoiled by that. :) )

Monika Miklos August 17, 2009 at 3:36 pm

Very informative! Nice job Dan! :)

resume template October 1, 2009 at 6:33 pm

test

scottfasser October 13, 2009 at 4:48 pm

Hi Dan, interesting data. I'd be interested to know the avg number of RT's per tweet or how many tweets generate a RT.

media buying October 14, 2009 at 2:24 pm

Interesting info, I wonder if advertisers take this into account when doing media buying on social networks.

media buying October 14, 2009 at 9:24 pm

Interesting info, I wonder if advertisers take this into account when doing media buying on social networks.

college essay January 5, 2010 at 10:45 am

RT is simpler; that explains its popularity. this really shows how people take control of one part of technology. social media is a great driver of people's participation

Turon at Lugaw March 17, 2010 at 2:03 am

This is very interesting, Dan. I'm very interested to know how and where you got your data.

Ford Spares September 24, 2010 at 8:43 pm

Nice set data on retweets you have there Dan.. This is the proof if your tweet is interesting it might get retweeted. :)

{ 17 trackbacks }