Official Library Twitter accounts- what factors are correlated with number of followers?

After posting about what Library Twitter accounts are doing here and here, I stumbled upon a more efficient way of getting the statistics and this resulted in the following batch of statistics (Number of followers, Number of following, Followers/following ratio, Age of account (based on first tweet) and total updates for all 456 accounts on the Library Twitter League.

And as luck would have it, at the same time the Twitter League was updated and you can now, get the very same statistics as well from there! You can now order by not just followers, but also following , follower/following ratio, signup date and alphabetical. In any case, the data will be used in this post.

Is the list representative of Library Twitter accounts?

No doubt, many people will have questions about how representative the list is. As mentioned here ,the list began from self-reported cases from here , here and here . I was somewhat concerned that this list has an inherent bias in listing mature accounts, so I searched using Twitter search using “Lib”, “library”, “ref” etc as well as directories such as wefollow and tried as best as I could to add obvious library accounts.

Nevertheless, I can’t rule out the fact that the sample here isn’t representative, though I submit that 456 is a large sample and is probably a good starting place as any.

Descriptive statistics

You will notice from the results above that the mean and median number of followers, 212 and 113 respectively is a bit different compared to what was reported in an earlier post (the mean was lower, but median was higher in earlier smaller sample), mainly because the composition of accounts in the earlier batch was quite different (earlier calculations had left out a few big accounts, while excluding numerous smaller accounts).

Age of account

A new statistic we have is the age of the account based on first tweet in days. The oldest account is 984 days old, or over 2.5 years old, making it one of the pioneer accounts. But 50% of accounts are younger than 132 days old (roughly 4 months), showing a surge in many new accounts.

Above shows the cumulative distribution graph of Library Twitter accounts by age (days).

Total updates

95% of library Twitter accounts have updated less than 1,000 times but this does not take into account the age of accounts.

Updates per day

It was obvious also to calculate updates per day, and generally library Twitter accounts don’t update much with a mean update of 0.98 updates per day.

The twitter library accounts generally don’t update a lot. 95% of accounts update less than 3 times daily. Note that update per day is calculated by taking total updates divided by the age of the account. Some accounts are dormant for a while before they become really active, so this statistic probably understates the current updates per day.

TFF ratio

One statistic, I calculated in an earlier blog post was Twitter follower/following ratio or more formally named Twitter Follower-Friend Ratio (TFF)

I was curious, what the typical TFF ratio of library accounts was. Did Library twitter accounts follow everyone who followed them as recommended here? Did they embark on a strategy to proactively follow identified users and hence had TFF ratios below 1?

As noted in earlier post, proactively identifying people who are your users allows and following them, allows you to proactively respond to them even if they are not directed to you, and even if they just say “library” without mentioning the name of your library.

For all 458 library twitter accounts the following pie chart shows the distribution below

It seems that a majority (57%) of Library Twitter accounts have less followers then those following them! 5% had no followers and the rest had more followers than following.

As we have seen in an earlier blog post, this result differs quite a bit when we only consider the top 100 Library Twitter accounts (based on following size) as shown below. In particularly, only 15% of the top 100, have TFF ratios of below 1 (see below).

In a sense this is not surprising. The sample of all 458 Twitter accounts, includes many accounts just starting out, and they might be still in a “growth” stage or some may be initially employing a strategy of following many accounts to build up reciprocal relationships?

Bio of Twitter accounts followed

What type of accounts are library Twitter accounts following? The following word cloud of the bios of the followers of a Library in Oklahma (via Twittersheep) is quite typical.

Libraries Twitter account seems to be following Libraries, Librarians, writers and probably “book lovers”, and of course people in Oklahma.

Another typical account showing that this Library like to follow authors.

Correlation analysis

Finally did a little bit of correlation analysis

Results

1) As expected age of account is highly positive correlated with number of followers. Number of those following is not correlated with age.

2) Number of followers is highly and positively correlated with TFF ratios.

It seems to me that, at least for the library twitter accounts, larger follower accounts are correlated with high TFF ratios (low number of following relative to followers). Probably the biggest most famous accounts such as Library of Congress, New York Public Library by virtue of their reputation gain huge number of followers automatically and do not need to build up their following by reciprocal following.

3) As expected age of account is correlated with total number of updates.

4) Updates per day is correlated with Followers and followings.

Obviously there is quite a bit of co-linearity.

Multi-linear regression – what explains number of followers

Ran a quick regression analysis

Multi-linear regression (stepwise) was carried out with number of followers as the dependent variable, and the other statistics as independent variable.

The final model, has a adjusted R-square of 0.567.

It indicates that number of followers is explained by increased number of followings, follower/following ratio, age (in days) and updates per day.

Is this model useful? Just putting it out there. Probably requires a larger range of independent variables.

Library twitter league – official library twitter accounts compared

Recently, I was pondering about the need to set targets or KPI (key performance indicators) for Libraries carrying out Social media activities. Take Libraries with Twitter accounts for example, what would be a suitable target?

The first thing that comes to mind is number of followers, which would be a measure of how many people you are reaching, but what would be a creditable target? One very inaccurate and rough way would be to look at the followers that other libraries have achieved (while taking into account relevant differences such as size of population served etc).

Rather than creating a static list I used the TwitterLeague service , and populated it with entries from here , here and here.

I had a suspicion that the lists above were biased towards successful, established accounts, as such I further supplemented them by adding more accounts. I searched for the word “Library”, “Lib”, “Reference” using Twitter’s find people service and added the accounts I found there.The full list as of May 15, 2009 stands at over 400 entries, but here’s just the top 20.

The full results (which update dynamically) are available here .

You can also download a xml version of the list, and view it in Excel etc.

A rough calculation (this will change as I add more twitter accounts) shows that the median number of followers is 147, while the average number of followers is 260.

Not surprisingly, Library of Congress is way ahead in the follower counts, more than 4 times of the next highest library account.

Of the top 20 library accounts, Public and National libraries dominate compared to Academic libraries with only the 3 universities (Yale has 2 accounts in the top 20) making the top 20. They include second placed Open University Library from the UK (a distance learning university), Yale Library accounts (6th and 7h place) and College of DuPAGE library (8th place)

It will probably be interesting to revisit the statistics say 6 months later to see if there are any major changes.

Of course, follower counts alone probably don’t tell everything. Some accounts are just starting out, others follow a large number of accounts, which typically follow back etc. Other Twitter analytic services such as Twitter Grader , Twinfluence , Twitterholic, Twitalyzer, Tweetstats provide more sophisticated measures of influence, by taking into account follower/following ratios, % of retwits etc. The JCPR Twitter index is probably the most advanced metric available used to measure online influence.

Hopefully, TwitterLeague will add more statistics, please help vote for additional features here.

Given that libraries are using twitter for different purposes, I can see possible value in studying statistics such as rate of updates, signal to noise ratio (percentage of twits with links or hash tags) of accounts, Name pointing (percentage of times @ is sent to account) and more. For example, a twitter acccount used mainly for pushing service announcements would probably have a different tweeting profile compared to one that was used for reference service.

I also created a new Twitter account leaguelibrary and populated it with the twitter accounts from the library league twitter list (using Twitterator to quickly add multiple accounts).

I was hoping to find a way to export all the twits from this account and then carry out text analysis, using Many eyes‘s to create Word Trees, Wordle , Tag Cloud and Phrase Net.

Unfortunately while I found many ways to export my own public timeline of tweets, there doesn’t seem to be a way to export my tweets PLUS tweets of accounts I’m following? Does anyone know a way to do this? I did manage to get a RSS feed of what I wanted though, and I might use this for analysis.

For now, I use My Twitterflock to generate a word cloud from the bios of the library accounts I’m following.

As well as a word cloud of what they are posting about


Google Profiles of libraries

Introduction

In this post, I will talk about a little known service called Google Profiles, why it is becoming important, and I will describe how libraries have being using it, the web 2.0 accounts they are listing on it, and compare it to the accounts listed by libraries on Friendfeed.

Google Profiles

For years, Google had a little known feature/service called Google Profiles, which allowed users of Google accounts to setup profiles of themselves. In many ways they were similar to the lifestreaming accounts like Friendfeed, in that you listed other associated accounts such as Flickr, Twitter, Facebook, which you owned. Despite Google’s brand recognition, they have never being particularly skilled at building social networks and Google Profiles was pretty much ignored.

This was until April 2009, where they placed their trump card. Two changes occured, one minor change, one major. The first change was that they allowed users to use a custom profile url, instead of a long string of numbers. For sure http://www.google.com/profiles/aarontay looks better than http://www.google.com/profiles/104559151215707191902. Below shows an example of a Google Profile set up by E.H Butler Library.

More significantly, these Google Profiles appear at the bottom of Google searches! when you do a search.

You can find out more about Google profiles here.

Libraries on Google profiles- Methodology

In a previous post, I looked at Library accounts on Friendfeed (a popular Lifestreaming service) and studied the web 2.0 accounts that Libraries linked to. A very obvious idea now is to do the same for libraries on Google Profiles as well.

I did a Google Profile search, by searching for the word library in the title. This results in about 162 entries. Unfortunately the vast majority of entries were blank, or had at best a link to their homepage. These I ignored on the admittedly dubious assumption that most libraries had at least a blog and accounts without blogs were Google Profiles that were not properly maintained. Moreover I noticed that one of the libraries, the Unquiet Library had both a Google Profile and Friendfeed with the later having more accounts listed. Of the remaining, I took note of the accounts they listed. Again a Google doc version of the data is available.

Results are as below.

The disclaimers I made for previous post applies here, the data above is not representative of what libraries in general are doing in the web2.0/socialmedia arena . Rather they show what libraries who have custom Google Profiles are doing.

Comparing Libraries on Google profiles and Libraries on Friendfeed


I’ve reproduced the chart showing accounts linked to in Friendfeed above.

The main difference between the two charts is that compared to Libraries on Friendfeed, fewer Libraries on Google Profiles list Twitter accounts.

A minor difference perhaps is that Libraries on Google Profiles tend to list more Picasa accounts for photo sharing compred to Flickr. This can be explained by the fact that Picasa is a Google service can will be autoamatically added.

We also see libraries listing Google Books (example)and more Google Reader (example) accounts. They are quite a few innovative uses of these 2 accounts, see more from data.

Unlike Friendfeed where you can only add specific accounts, or accounts with RSS feeds, with Google Profiles you are free to add any URL, and libraries have exploited this by adding links to subject guides (example), internet archive (example), Yahoo pipes (example) etc.

Another difference seems to be that none of the libraries on Google Profiles list links to delicious . It’s unclear if this is a result of libraries not listing them, or the libraries not having an account in the first place.

So does your library have a Google Profile ? Are you maintaining it?

Aaron Tay