Depending on who you talk to, Twitter is either killing news media or saving it. A recent analysis by Hitwise found that less than 0.2% of people who use Twitter wind up going to news and media sites (thanks to Karen for the link). Their analysis looks at Twitter as a whole though, and I’m not sure how well it accounts for local news sites. I believe very strongly that social media has the greatest impact at the local level (more on this in a future post). Given that, I have long wondered how Twitter has impacted local news media here in Edmonton. Last night, I finally did some analysis. I decided to explore how much traffic the Edmonton Journal and iNews880, Edmonton’s two top tweeting media outlets, received from Twitter last year.
First up, the Edmonton Journal. They’ve been tweeting news articles since at least January 2009, so I had lots of data to play with. They used tweetburner to shorten links until September when they switched to bit.ly. Using the APIs available from those services, I added up all the click stats for all the links posted by The Journal. Here’s what I found:
Lots of variation, as you can see. Some of that is down to the use of two services, some of it is because of the number of Twitter users. There are probably dozens of other factors too.
For the period January 30 through December 31:
- A total of 153,968 clicks were recorded on 4737 links.
- That’s an average of 33 clicks per link, and an average of 15 links per day.
- According to the stats on the bit.ly links, 95.4% of clicks come from the Edmonton Journal’s hash*.
- The link with the most clicks (700) was this one, on May 26. It doesn’t work, because annoyingly The Journal doesn’t display old articles for some reason, but it appears it was about Edmonton’s Poet Laureate Roland Pemberton.
- The day with the most clicks, September 14, doesn’t appear to be special…just lots of clicks that day for some reason (any ideas?).
Next up, iNews880, one of the first local media organizations to join Twitter. They used tinyurl.com until July, when Twitter switched the default to bit.ly, so unfortunately I only have data for the latter half of the year:
For the period July 14 through December 31:
- A total of 90,500 clicks were recorded on 3811 links.
- That’s an average of 24 clicks per link, and an average of 22 links per day.
- According to the stats on the bit.ly links, 93.8% of clicks come from iNews880’s hash*.
- The link with the most clicks (1933) was this one, on August 2 (that’s the huge spike in the graph above). The link goes to the report on the Big Valley Jamboree stage collapse, and it was popular because it included before and after photos.
Edmonton Journal vs. iNews880
I wanted to do a quick comparison, so I chose the period September 17 through December 31, because both sites used bit.ly for links during that time. Here’s what it looks like:
During that time:
- The Edmonton Journal posted 2369 links (23 per day) and iNews880 posted 2261 links (22 per day).
- A total of 79,519 clicks were recorded on Edmonton Journal links (an average of 751 per day or 34 per link).
- A total of 53,815 clicks were recorded on iNews880 links (an average of 508 per day or 24 per link).
That’s a lot of clicks! Clearly Twitter and other social networking sites (where most shortlinks are posted) are having an impact. But how much? According to the latest report by the Newspaper Audience Databank (NADbank), weekly online readership at EdmontonJournal.com increased by 35% last year to 115,900 from 85,800 in 2008. That’s an increase of 30,100 readers per week. According to the click stats above, The Journal received 3208 clicks per week in 2009. So what does that mean?
Roughly 10.7% of the Edmonton Journal’s online readership increase in 2009 came as a result of posting links to Twitter.
And if I had to guess, I’d say my analysis probably underestimates things. Apparently the NADbank data is based on surveys, so I’m not sure how accurate it is, but it’s probably within acceptable margins of error. I’m also not sure what exactly a “reader” is – a page view, a visit, etc.
I’ve tried to be as accurate as possible, but I can’t make any guarantees!
- All the click stats are current as of last night.
- I’m suggesting that all the clicks come via Twitter, when that’s probably not entirely true. Links get passed around, displayed on websites, etc. But the shortlinks do originate at Twitter.
- It’s possible that The Journal or iNews880 posted a link to somewhere other than their own sites, but uncommon. I did remove one link from the iNews880 dataset, because it pointed to an Environment Canada site (it was obvious, lots of total clicks as others have linked there too). For the rest, I’m making the assumption that the links point to the news sites.
- I don’t know how reliable the stats from bit.ly and tweetburner are. I suspect they are quite a bit different than server logs or Google Analytics metrics.
- I would assume that both services tweaked the way stats are calculated throughout the year, so 15 clicks on a bit.ly link in May is probably different than 15 clicks on a bit.ly link in December.
* – When you shorten a link using bit.ly, you get a unique hash. If I shorten the same link, I get a different hash. The stats are recorded and made available individually and in aggregate, however.