Twitter statistics for City Council’s vote on the financial framework for the downtown arena

By Mack D. Male · October 27, 2011 at 11:55pm

After a long public hearing on Tuesday, City Council yesterday debated the proposed financial framework and ultimately whether or not they wanted to proceed with the downtown arena project. They voted 10-3 in favor of the framework, and also voted to spend $30 million to complete the design to 60%. Here’s my analysis of the the arena-related tweets posted by Edmonton users between 9:30am and 9:30pm.

By graphing the tweets per minute, you can very clearly see the time the vote took place (~3:03pm):

I was curious to know if the things people were tweeting before and after that moment were different. Here’s a word cloud of the tweets prior to the vote:

Caterina was mentioned a lot, which makes sense considering he turned out to be the surprise dealmaker of the day. The other Councillors were mentioned quite frequently too, as myself and others tweeted their comments.

Here’s a word cloud of the tweets posted after the vote:

It’s very interesting that “Edmonton” was tweeted so often after the vote passed. There were a lot of tweets similar to “Edmonton will get a new arena” that were retweeted after the vote. You can also see that “Iveson” was fairly prominent after the vote, reflecting the large number of tweets about his final remarks on the deal.

Other stats:

It was another busy day for tweets in Edmonton with more than 42,000 posted by Edmontonians. That works out to an average of about 30 per minute.
More than 880 users posted at least one arena-related tweet.
On average, 5.0 arena-related tweets were posted per minute between 9:30am and 9:30pm. The peak was 43.
Roughly 14% of the tweets were replies to other users.
Roughly 29% of the tweets were retweets.

Here are the top 20 most active local users (most tweets to least):

Here are the top 20 most retweeted local users (by other local users, most retweeted to least):

I gave Paula a run for her money, but she remained the most retweeted user on the arena issue!

UPDATE: I’m always looking for better ways to analyze tweets. Finding a good, reliable way to do sentiment analysis (are tweets positive or negative) is a challenge, partially because tweets are so short and because they usually include weird entities like hashtags (weird from a natural language processing point-of-view). To analyze the arena-related tweets, I used uClassify’s Sentiment Classifier. Here are the results:

I would say this is pretty much as expected. Tweets before the vote probably expressed less emotion one way or the other. Most people tweeting after the vote seemed happy with the decision Council made.

Twitter statistics for today’s public hearing on the proposed arena deal

By Mack D. Male · October 25, 2011 at 11:59pm

I think it’s fair to say that the public hearing on the proposed arena deal was the talk of Twitter in Edmonton today. Here’s my quick analysis of all arena-related tweets posted by Edmonton users today between 9:30am and 9:30pm.

A word cloud of the 4500+ tweets:

Here’s a breakdown of tweets per minute (you can clearly see the lunch and dinner breaks):

Other stats:

Thanks largely to the arena-related tweets, more than 44,000 tweets were posted by Edmontonians today. That works out to an average of about 31 tweets per minute.
More than 800 different users posted at least one arena-related tweet.
On average, 6.3 arena-related tweets were posted per minute between 9:30am and 9:30pm. The peak was 24.
Roughly 17% of the tweets were replies to other users.
Roughly 25% of the tweets were retweets.

Here are the top 20 most active local users (most tweets to least):

Here are the top 20 most retweeted local users (by other local users, most retweeted to least):

Council voted to deal with the issue at 9:30am tomorrow (Wednesday). You can watch or listen live here.

Five Years with Twitter

By Mack D. Male · July 15, 2011 at 3:30pm

It was five years ago today that Twitter officially launched to the public (the very first ever tweet was sent on March 21, 2006). It was also five years ago today that I signed up for the service. It has become my claim to Twitter fame (such as it is) – I was the 985th person in the world to join! More than 600,000 people joined Twitter yesterday, which is pretty amazing when you consider that it took more than 16 months for the first 600,000 people to join!

When it launched, Twitter was actually Twttr (no vowels). At the time I was busy working on Podcast Spot. We were always paying attention to what our competitors were doing, and one of the biggest names in podcasting at the time was Odeo. I remember reading that they had launched a side-project named Twttr, and I remember thinking “this is dumb” after I checked it out. I mean the idea was neat, without a doubt, but I couldn’t fathom why they would be putting resources into Twttr rather than into Odeo. Anyway, as you know Odeo died and Twitter took off, so obviously they made the right decision!

I’ve written over a hundred Twitter-related blog posts over the last five years. My early entries seemed to be all about Twitter’s infamous fail whale and how the service struggled to stay operational, though I did immediately pick up on the ability to track topics. It was well into 2008 that they were still experiencing issues with reliability. That was also the year that I organized our first ever tweetup here in Edmonton (with help from Melanie and others). In June of 2008, I was down in Calgary for BarCamp and did a presentation on Twitter. After chatting with Wil at the bar afterward, I decided we should borrow the city hashtag idea from Calgary (they were using #yyc). The first #yeg tweet went out on June 20, 2008 (I wrote a bit more about that here). Exactly two years after Twitter launched, it purchased Summize, the search engine that now powers Twitter Search. That was a big deal, as it made the service much more useful. It also made it possible for me to start tracking the Edmonton Twittersphere, and I posted my first look at those statistics in February 2009. That seemed to give the local scene some momentum, and a month later I was at CTV talking to their newsroom about Twitter. That was the turning point in Edmonton, in my opinion. A lot of people joined after they ran the Twitter story, and I think the fact that CTV embraced the service gave it some legitimacy. The local Twittersphere has been growing in size and influence ever since.

I have always been a web user of Twitter. Over the years I have used apps on my mobile phones, text messaging, and I’ve dabbled with apps like TweetDeck and HootSuite, but my primary interface remains the Twitter website. I was particularly happy about #newtwitter, though I know a lot of you didn’t like the redesign (at least initially). It’s kind of incredible to think back to the time when Twitter didn’t have retweets, mentions were just replies, and hashtags were rare. The addition of lists was another thing that changed the way I use Twitter. I’m often asked how I can possibly follow nearly 6000 people and the answer is always “I don’t.” I use a combination of lists and search to pay attention to certain people and/or topics! I rarely, if ever, look at the timeline. It look me a long time to get over that – early on I definitely felt like I didn’t want to miss anything! Twitter is still largely the same as it was in 2006 (at least conceptually), but the changes that have been made have really had an impact.

I don’t know what Twitter will look like five years from now, but it certainly shows no signs of going away. I look forward to its continued evolution, and I hope Twitter continues to have a positive impact here in Edmonton!

Special thanks to Jeff and Sally for the Twitter birthday post today! And yes, I need to get on with updating stats!

State of the Edmonton Twittersphere – 2010 Year in Review

By Mack D. Male · February 4, 2011 at 12:30pm

Welcome to the State of the Edmonton Twittersphere: 2010 Year in Review, my look at the intersection of Twitter and Edmonton in 2010! You can see my recap of 2009 here.

I’ve done my best to ensure all of the data in this post is accurate, but I make no guarantees – use it at your own risk. The data comes from the Twitter API, and has been collected over the past year. If a user has his or her location set to Edmonton, St. Albert, Sherwood Park, Leduc, Nisku, Stony Plain, Fort Saskatchewan, Beaumont, Spruce Grove, or matching lat/long coordinates, they are considered an Edmontonian, and thus a “local user”.

If you compare the monthly statistics here to my monthly State of the Edmonton Twittersphere posts, you’ll notice they are slightly different. The monthly posts represent a snapshot – this post reflects the most up-to-date information I have been able to gather as of the end of 2010.

Summary

Here are the highlights for 2010:

There were 22,000+ local users.
Those users posted more than 4.9 million tweets. That works out to 9.4 per minute.
Of those 4,948,409 tweets:
- 381,752 contained #yeg or one of the #yeg-related hashtags (like #yegfood) (7.7%)
- 357,206 were retweets (7.2%)
- 1,715,507 were replies (34.7%)
- 668,368 were replies to other local users (13.5%)
- 1,331,306 contained links (26.9%)
- 191,060 were twooshes (exactly 140 characters) (3.9%)

Let’s look at users. While more than 22,000 local users were on Twitter last year, only 10,200 of them were active during the month of December (active means they posted at least one tweet). But that was up from just 5601 who were active during the month of January.

When 2010 started, local users were posting about 260,000 tweets per month. By the end of the year, that number had grown to more than 525,000 tweets per month.

This chart gives you a sense of the trends over the year. I think it is interesting that the lines for #yeg-related tweets and retweets are almost identical (red and green).

Roughly 50.8% of all tweets in 2010 were posted between the hours of 9 AM and 6 PM. Not surprisingly, the number of tweets posted between midnight and 7 AM is quite a bit lower than the number posted during the day and early evening. There seems to be an early morning (9-10 AM) and late night (9-10 PM) spike.

If we look at days of the week, we see that more tweets are posted on Tuesday and Wednesday than on any other day.

In the chart below I have plotted the number of tweets posted per day for each day of the year. The trend is clearly up, and the spikes and troughs reveal some interesting events. Election day, October 18, saw the most tweets posted at 23,234. On average, 13,558 tweets were posted each day in 2010.

Users

Here are the top 25 most followed users:

The average local user has 120 followers. Just 420 users have more than 1000 followers.

Here are the top 25 most listed users:

The average local user has been listed 5 times.

Here are the top 25 most active users:

The top 100 most active users accounted for 18.5% of all local tweets.

Here are the top 25 most active users using #yeg (and #yeg-related hashtags):

The top 100 most active users using #yeg and its subtags accounted for 51.8% of all #yeg-tagged tweets.

Here are the top 25 most replied to users (by other local users):

The top 100 most replied to users accounted for 32.8% of all local replies (replies from one Edmontonian to another).

And here is what I think is the most significant list, the top 25 most retweeted users (by other local users):

A total of 103 users were retweeted by other local users 100 times or more. Just 18 users were retweeted by other local users 1000 times or more.

Hashtags

The most commonly used hashtag was #yeg. Local users used #yeg roughly 6.5 times more than the next most popular hashtag, which was #FF. Here’s a word cloud of the top 1000 hashtags, including #yeg:

And here are the top 1000 without #yeg:

The average length of a hashtag was 13.7 characters (including the #). There were hashtags that were just two characters, and hashtags that were 140 characters. Here are the top 10 hashtags:

Here are the top 10 #yeg-related hashtags:

Clients

There were more than 2000 different clients used by local users to post tweets in 2010. Here are the top ten clients:

Text messaging was the next most popular client. Here are the top ten in graphic form (percentages are of the total number of tweets, 4.9 million):

The top ten clients accounted for 76.2% of all local tweets in 2010.

Final Thoughts

Twitter continued its impressive growth all around the world in 2010, and Edmonton was no exception. Though the number of people with Twitter accounts in Edmonton pales in comparison to the number of people with Facebook accounts, I don’t think that is necessarily the best comparison to make. You need a Facebook account to access most things on Facebook, you don’t on Twitter. Twitter reaches far beyond the 22,000+ local users with accounts.

There were lots of tweetups in 2010, but fewer and fewer focused just on Twitter. Because so many more people have joined, even non-Twitter events seem like tweetups! I thought that geolocation might play a bigger role in 2010, but it didn’t really. Just 3124 users have enabled geolocation (up from 270 in 2009). Perhaps 2011 will be the year that geotagged tweets take off? You need to enable it in your settings.

I hope you’ve found this look at the Edmonton Twittersphere in 2010 interesting and informative. Thanks for reading!

State of the Edmonton Twittersphere – December 2010

By Mack D. Male · February 2, 2011 at 11:00am

Welcome to the twelfth State of the Edmonton Twittersphere for 2010, my look at the intersection of Twitter and Edmonton, AB. You can see the stats for November here.

For information on the data, definitions, and other background, click here.

For December 2010:

# of local users: 10198 (an increase of 259 from November)
# of tweets by local users: 529387
# of tweets by local users containing #yeg: 37063 (7.0%)
# of tweets by local users that were replies: 191871 (36.2%)
# of tweets by local users containing links: 141953 (26.8%)
# of tweets by local users that were retweets: 34255 (6.5%)
# of tweets by local users that were twooshes: 22703 (4.3%)

Here are the numbers above in graphic form:

Here are the top clients used by local users for posting updates:

Some other interesting stats for the month:

Just under 50% of all local tweets were posted between 9 AM and 5 PM.
Local users posted roughly 11.6 tweets per minute in December (compared to 12.2 tweets per minute in November).
The day with the most local tweets posted was December 14 at 21297. On average, 17077 local tweets were posted each day (compared to 17527 in November).
Of the 191871 replies posted by local users this month, 78509 or 40.9% were to other local users.
A total of 1923 users posted 50 times or more in December. In comparison, 1667 users posted just once.

Here are the top ten most active local users (not including bots):

Here are the top ten most active local users using #yeg (not including bots):

Here are the top ten most replied to local users:

Here are the top ten most retweeted local users (by other local users):

Final Thoughts

Edmonton finally broke the 10,000 active user mark! Aside from that, however, it was definitely a holiday month. You can see a noticeable drop in the number of tweets posted around Christmas. Also fewer tagged tweets and more replies in December, which suggests to me less news being posted and more conversation taking place.

I should have my 2010 year-in-review up later this week. If you’re interested, stop by Twitter Talks on Thursday evening for even more twittersphere stats!

As always, keep in mind that the stats above rely upon users setting the location field of their profile to something like “Edmonton”. Users who leave that field blank or who put something like “Canada” are not reflected in the above stats.

State of the Edmonton Twittersphere – November 2010

By Mack D. Male · January 31, 2011 at 11:00am

Welcome to the eleventh State of the Edmonton Twittersphere for 2010, my look at the intersection of Twitter and Edmonton, AB. You can see the stats for October here.

For information on the data, definitions, and other background, click here.

For November 2010:

# of local users: 9939 (an increase of 318 from October)
# of tweets by local users: 525804
# of tweets by local users containing #yeg: 40905 (7.8%)
# of tweets by local users that were replies: 184033 (35.0%)
# of tweets by local users containing links: 145767 (27.7%)
# of tweets by local users that were retweets: 38958 (7.4%)
# of tweets by local users that were twooshes: 22722 (4.3%)

Here are the numbers above in graphic form:

Here are the top clients used by local users for posting updates:

Some other interesting stats for the month:

Just over 50% of all local tweets were posted between 9 AM and 5 PM.
Local users posted roughly 12.2 tweets per minute in November (compared to 12.0 tweets per minute in October).
The day with the most local tweets posted was November 23 at 20311. On average, 17527 local tweets were posted each day (compared to 17252 in October).
Of the 184033 replies posted by local users this month, 73837 or 40.1% were to other local users.
A total of 2030 users posted 50 times or more in November. In comparison, 1394 users posted just once.

Here are the top ten most active local users (not including bots):

Here are the top ten most active local users using #yeg (not including bots):

Here are the top ten most replied to local users:

Here are the top ten most retweeted local users (by other local users):

Final Thoughts

We got really close to 10,000 users in November! The number of tweets overall was slightly lower than in October, which is partially explained by the month having one less day and partially by the election that took place in October. The average number of tweets per day was almost exactly the same as in October, however.

I should have December 2010 and my 2010 year-in-review up later this week. Also, stop by Twitter Talks on Thursday evening for even more twittersphere stats!

State of the Edmonton Twittersphere – October 2010

By Mack D. Male · January 19, 2011 at 10:00am

Welcome to the tenth State of the Edmonton Twittersphere for 2010, my look at the intersection of Twitter and Edmonton, AB. You can see the stats for September here.

For information on the data, definitions, and other background, click here.

For October 2010:

# of local users: 9621 (an increase of 433 from September)
# of tweets by local users: 534809
# of tweets by local users containing #yeg: 47851 (8.9%)
# of tweets by local users that were replies: 187975 (35.1%)
# of tweets by local users containing links: 137406 (25.7%)
# of tweets by local users that were retweets: 41205 (7.7%)
# of tweets by local users that were twooshes: 19444 (3.6%)

Here are the numbers above in graphic form:

Here are the top clients used by local users for posting updates:

Some other interesting stats for the month:

Just over 49% of all local tweets were posted between 9 AM and 5 PM.
Local users posted roughly 12.0 tweets per minute in October (compared to 11.7 tweets per minute in September).
The day with the most local tweets posted was October 18 at 23232. On average, 17252 local tweets were posted each day (compared to 16783 in September).
Of the 187975 replies posted by local users this month, 74095 or 39.4% were to other local users.
A total of 2035 users posted 50 times or more in September. In comparison, 1323 users posted just once.

Here are the top ten most active local users (not including bots):

Here are the top ten most active local users using #yeg (not including bots):

Here are the top ten most replied to local users:

Here are the top ten most retweeted local users (by other local users):

Final Thoughts

October was a big month for Twitter here in Edmonton, thanks primarily to the municipal election. More users, more tweets, more tagged tweets, and more retweets. The busiest day of the month was election day (there’s a noticeable spike the in graph). Twitter was used to discuss the forums and other election-related events, to learn about the candidates, and to debate the issues. And on the big night, the candidates themselves used Twitter to capture the moment. You can see my election-related entries here.

Slowly catching up on my stats – I should have the rest of 2010 up soon.

State of the Edmonton Twittersphere – September 2010

By Mack D. Male · January 18, 2011 at 10:00am

Welcome to the ninth State of the Edmonton Twittersphere for 2010, my look at the intersection of Twitter and Edmonton, AB. You can see the stats for August here.

For information on the data, definitions, and other background, click here.

For September 2010:

# of local users: 9188 (an increase of 700 from August)
# of tweets by local users: 503471
# of tweets by local users containing #yeg: 39862 (7.9%)
# of tweets by local users that were replies: 178519 (35.5%)
# of tweets by local users containing links: 131229 (26.1%)
# of tweets by local users that were retweets: 34503 (6.9%)
# of tweets by local users that were twooshes: 16028 (3.2%)

Here are the numbers above in graphic form:

Here are the top clients used by local users for posting updates:

Some other interesting stats for the month:

Just under 51% of all local tweets were posted between 9 AM and 5 PM.
Local users posted roughly 11.7 tweets per minute in September (compared to 10.8 tweets per minute in August).
The day with the most local tweets posted was September 23 at 21096. On average, 16783 local tweets were posted each day (compared to 15536 in August).
Of the 178519 replies posted by local users this month, 69081 or 38.7% were to other local users.
A total of 1904 users posted 50 times or more in September. In comparison, 1233 users posted just once.

Here are the top ten most active local users (not including bots):

Here are the top ten most active local users using #yeg (not including bots):

Here are the top ten most replied to local users:

Here are the top ten most retweeted local users (by other local users):

Final Thoughts

I’m not sure why there was a bit of a jump in users and in the number of tweets posted in September, except perhaps that people go from “summer” mode to “get down to business” mode. Interest in the upcoming election made have played a role as well.

As with the previous month, the number of retweets includes both new and old style, both with and without extra comments.

Slowly catching up on my stats – I should have the rest of 2010 up soon.

How Friday’s AMBER Alert unfolded on Twitter

By Mack D. Male · November 1, 2010 at 10:30am

On Friday afternoon, RCMP issued an AMBER Alert for 12-year-old Jacob Telford. Thankfully, he was found “in good condition” after just a few hours. For a while that evening, it seemed like every tweet in Edmonton was related to the incident. At one point during the evening, @tkoriordan said “So I’ve been away from Twitter all day. Somebody want to fill me in on what I missed?” to which @Kiri_W replied “About a billion amber alert retweets.” Curiosity got the better of me, so I looked at the data to see how the AMBER Alert unfolded on Twitter.

Here are the first four tweets that appeared:

Just heard about amber alert, young boy from #shpk, check out alert and key an eye outMatthew Ross
Oct 29 at 2:11 PM via ÜberTwitter

1st time I have hear the emerg broadcast have a real emerg. Its ano amber alert for a little boy missing from Sherwood ParkLesley S
Oct 29 at 2:11 PM via Twitter for BlackBerry®

Amber Alert in #YEG child abducted in #sherwoodpark. #Edmonton #AMBERALERT!trevortye
Oct 29 at 2:15 PM via Echofon

AMBER ALERT: 12-year-old Jacob Telford taken from Sherwood Park last night – White 2002 Ford Taurus – BC Plate: V2P4J3CTV Edmonton
Oct 29 at 2:17 PM via Seesmic Desktop

Actually, Rob McAnally had one up before the main CTV Edmonton account did, but it seems that tweet has since been deleted. CTV posted the first link to a new story at 2:29 PM (that link has since been clicked over 1000 times).

From Matthew’s first tweet at 2:11:17 PM until 7:23:43 PM when Jeremy Lye’s tweet appeared declaring that the AMBER Alert had been cancelled (the first to do so), a total of 958 tweets were posted (mentioning either the AMBER Alert or Jacob Telford). That works out to just over 3 tweets per minute (in the first hour, it was nearly 7 tweets per minute). By 5 PM the next day, the total number of related tweets had grown slightly to 1126.

Of those, 983 were retweets. Here’s what the retweets looked like:

@ctvedmonton was retweeted 674 times
@SimonOstler was retweeted 126 times
@sonic1029 was retweeted 70 times
Taken together, everyone else was retweeted 113 times

They weren’t quite the first to tweet the start or end of the AMBER Alert (they were about six minutes behind both times), but CTV Edmonton was definitely the most visible account to do so, and that is reflected in the number of retweets they received.

In total, 710 different Twitter users tweeted (or retweeted) about the incident. Just less than half were located in the Edmonton area, while some were as far away as Brazil, Russia, and Kuwait (based on the location in their profiles). I suppose there’s always a chance that their retweets could have helped (maybe they have followers in Canada) but I still have to wonder why someone who is so far away would tweet about something like this. Is it because it’s easy to click the retweet button? Is it because we all like to feel as though we’re helping? I’m not sure.

If you add up the number of followers those 710 users have, you get a potential reach of more than 288,000 users. That’s not accounting for overlap though, so the actual number is probably quite a bit less. And not all of those users will have been online to see the tweets. Still, there’s no question that the AMBER Alert was seen far and wide on Friday.

Here’s what I found interesting about all of this:

85% of all the tweets posted about the incident appeared before we knew the boy was safe
60% of all the tweets posted about the incident were simply retweets of CTV Edmonton’s tweets
It took less than 20 minutes for the first news story link to appear
After 24 hours, almost no one was tweeting about it anymore

You can learn more about the AMBER Alert Program here.

State of the Edmonton Twittersphere – August 2010

By Mack D. Male · October 29, 2010 at 10:00am

Welcome to the eighth State of the Edmonton Twittersphere of 2010, my look at the intersection of Twitter and Edmonton, AB. You can see last month’s stats here.

For information on the data, definitions, and other background, click here.

For August 2010:

# of local users: 8488 (an increase of 261 from July)
# of tweets by local users: 481605
# of tweets by local users containing #yeg: 35965 (7.5%)
# of tweets by local users that were replies: 167831 (34.8%)
# of tweets by local users containing links: 144532 (30.0%)
# of tweets by local users that were retweets: 34282 (7.1%)
# of tweets by local users that were twooshes: 16012 (3.3%)

Here are the numbers above in graphic form:

Here are the top clients used by local users for posting updates – BlackBerry apps really dropped off this month for some reason:

Some other interesting stats for the month:

Just under 51% of all local tweets were posted between 9 AM and 5 PM.
Local users posted roughly 10.8 tweets per minute in August (compared to 10.0 tweets per minute in July).
The day with the most local tweets posted was August 26 at 18232. On average, 15536 local tweets were posted each day (compared to 14390 in July).
Of the 167831 replies posted by local users this month, 65767 or 39.2% were to other local users.
A total of 1727 users posted 50 times or more in August. In comparison, 1199 users posted just once.

Here are the top ten most active local users (not including bots):

Here are the top ten most active local users using #yeg (not including bots):

Here are the top ten most replied to local users:

Here are the top ten most retweeted local users (by other local users):

Final Thoughts

Better late than never right? It has been a long time since I’ve posted any Twitter stats! A vacation, the election, all kinds of stuff has kept me busy. I think the August stats represent the start of a big increase in election-related tweets. You can see a number of #ecca users in the lists above, and there were 580 tweets posted in August using the #yegvote hashtag.

For August, the number of retweets jumped quite a bit because of changes I made to my analysis code. Previously I only counted tweets that started with “RT @” or tweets that were new-style retweets (using the retweet button). This time, I counted retweets that contained additional text as well. It’s common for users to post a retweet like “LOL! RT @…” and now those are included. The number of retweets therefore jumped from 21528 (old way of calculating) to 34282 (new way of calculating).

Considering it was the end of summer, I’m a little surprised that the average number of tweets posted per day went up!

UPDATE: @burnstand pointed out that the day with the most tweets, August 26, was the day Lady Gaga was in town.