Understanding the Behavior of Twitterers
We take a look at a research paper called "A Few Chirps About Twitter", presented at SIGCOMM '08, that reveals a great deal about the behaviors and trends of Twitterers. In the paper, researchers identify 3 major groups of Twitter users and explore how, where, when, and how much Twitterers use the service.
What is Twitter?
If you haven't used or heard of Twitter before, here's a short introduction. Twitter is a rapidly growing Online Social Network (OSN) where 140 characters makes you a "microblogger". Although the paper does a good job of defining Twitter, I find that the CommonCraft video does a great explanation of what Twitter does and can do for you:
Analysis of Twitterers
The research paper uses data gathered over a period of 3 weeks between January 22, 2008 - February 12, 2008. Approximately 100,000 Twitter users were used for the analysis. caveat: The timeframe is particularly poignant because since January 2008, the service has grown significantly and numbers reported in the paper regarding followers/following may appear low if you're a heavy user of Twitter already. (See Twitterholic for top Twitterer stats)
What Kind of Twitterer Are You?
They discovered 3 major groups of Twitters (Figure 2):
- Broadcasters: users that have a many more followers than they themselves are following.
- Acquaintances: users that have near a 1:1 ratio of followers to following
- Miscreants / Evangelists: users that are following many more users than they have followers.
- Miscreants are defined as potential spammers, lurkers, or stalkers
- Evangelists are defined as users who contact everyone they can, and hope that some will follow them
"Many of the users here are online radio stations, who utilize Twitter to broadcast the current song they are playing. Others include the New York Times, BBC, and other media outlets generating headlines."
How Often Do You Tweet?
From the data (Figure 4):
- Broadcasters: tweet (update) quite a bit, most which fall into the top 90 - 99th percentile (964 - 1,727+ tweets) with some broadcasters tweeting over 140,000 updates.
- Acquaintances: with near 1:1 relationships tend to be in the top 99th percentile and a large majority of the 90th+ percentile reside within the acquaintances group.
- Miscreants / Evangelists: groups are not among the most active tweeters. I see this as very typical of lurker behavior.
Also, they found that users with more than 250 followers tend to tweet more than those who follow more than 250 users. They also have an 80% probability that they have more than 6,000 updates (at the time).
What's Do You Use to Tweet?
Twitter makes it increasingly easy to tweet your updates from a multitude of devices and applications. It also allows you to receive tweets from followers with the same diversity of mediums. In the chart below you can see the various formats for input and output. (note: The same chart exists in the paper, I've taken the liberty to add some creative touches to it.)
|Source||Status Updates||% of Updates|
3 PM, Tweet Rush Hour!
As you can see by the locally time adjusted graph (Figure 5), most twitter updates peak around 3 PM and reach a low at 4 AM. The bulk of the activity occurs between 10 AM and 4 PM, which coincides with most normal work hours. This is very typical of many websites, as many people visit websites during the work period.
The workload shows a rise during later morning hours, relatively steady use throughout the day, and drop off during the late night hours. There was no significant information in the patterns with in days of the week.
GeoTweet: Twittering from Around the World
From the graph (Figure 13) you can see that the Top 3 are the USA, Europe, and Japan respectively. It's notable that although Japan started off slow it's had rapid growth over time. In addition, although Twitter is mostly popular in the US, the Top 10 countries after the US account for 50% of users in the test sample are: Japan, Germany, U.K., Brazil, Holland, France, Spain, Belgium, Canada, and Italy.
Takeaways for the Social Marketer
Broadcasters and Acquaintances have significantly more followers than the Miscreants and Evangelists. In addition, those with more than 250 followers tend to send frequent updates. So if you want to be effective on Twitter and you want to increase the number of followers, you'll need to be an active participant in the community.
In addition, if you want maximize the reach of your message, you'll want to tweet between 10 AM and 4 PM, when users are most active.
Lastly, even though the majority of the users are in the US, the growth rate and percentage of over all impact is still significant in other countries. So for a the non-US social marketer, getting established now may prove fruitful not too far down the line.
Other Related Online Social Networks
Although the paper takes a look at twitter, they also cite a few other small microblogging networks:
- qik: streaming videos from mobile phones
- Jaiku: similar to twitter where users share their "activity stream"
- dodgeball: lets users update their status along with geographical info + allowing the system to locate friends nearby
- GyPSii: a Dutch social network for mobile users, combining geo-location of users with image uploading (think brightekite)
- Bliin: like dodgeball and Rabble, GPS + mobile microblogging
Source & Citations
|Balachander Krishnamurthy||AT&T Labs - Research, Florham Park, NJ, USA|
|Phillipa Gill||University of Calgary, Calgary, AB, Canada|
|Martin Arlitt||HP Labs, Palo Alto, CA, USA and University of Calgary, Calgary, AB, Canada|
Web 2.0 has brought about several new applications that have enabled arbitrary subsets of users to communicate with each other on a social basis. Such communication increasingly happens not just on Facebook and MySpace but on several smaller network applications such as Twitter and Dodgeball. We present a detailed characterization of Twitter, an application that allows users to send short messages. We gathered three datasets (covering nearly 100,000 users) including constrained crawls of the Twitter network using two different methodologies, and a sampled collection from the publicly available timeline. We identify distinct classes of Twitter users and their behaviors, geographic growth patterns and current size of the network, and compare crawl results obtained under rate limiting constraints.