Late last year I decided to use my quiet time in the holiday season to set up my own URL shortener. I like bit.ly a lot, but the idea of having full control over my own URL shortener as well as a ‘personal brand’ short URL – along the lines of yoa.st – persuaded me to try and get one of my own.
First of all I needed to find a good URL. This was probably the easiest choice: once I discovered the .ms (Montserrat) TLD existed and the bada.ms domain was still available, I didn’t hesitate for a second.
Secondly I needed to find the right software. I could have gone for bit.ly pro, so that the people at bit.ly would be doing all the hard work for me. But instead I opted to set up my own shortener and not rely on an external third party. Somehow that felt more like it would be mine, instead of something I leased from someone else.
There are a few solutions out there but the Yourls.org package appealed to me most: it’s open source and based on PHP and MySQL, and there are plenty of plugins for it. It also easily integrates with Tweetdeck (my preferred Twitter app) and I found a great installation guide for it on Lifehacker.
Once I had it up and running I installed the Random Keywords plugin so that instead of having to choose my own short URL every time, I just let Yourls generate a string for me.
And there it was, my own URL shortener: bada.ms.
I started using the shortener on December 31st but didn’t unleash it in the wild – i.e. on Twitter – until January 1st. I only used it on links I posted on my @badams Twitter account, and then only for links that didn’t use a branded short URL themselves.
It’s been running for a full month now, so let’s see what kind of cool stats Yourls can give me.
In January I shared 74 bada.ms short URLs on Twitter. These URLs achieved an average of 68.44 clicks, and a median of 41.79 clicks.
The top three most clicked URLs were:
- http://bada.ms/rxd – (Jan 06) 286 clicks. This points to a Slideshare deck by Eldad Yogev which explains Mobile SEO in great detail. Not coincidentally this was also my most RT-ed tweet of the month.
- http://bada.ms/4ou – (Jan 21) 282 clicks, points to one of my own blog posts.
- http://bada.ms/vbn – (Jan 12) 215 clicks, points to a blog post about the dangers of free WordPress themes.
The least clicked publicly shared URLs were:
- http://bada.ms/thg – (Jan 03) 26 clicks, points to a story about spam in Google’s SERPs.
- http://bada.ms/9s6 – (Jan 21) 29 clicks, points to one of my columns on the Belfast Telegraph site.
- http://bada.ms/rw6 – (Jan 06) 30 clicks, points to a news article about the MMR vaccine scare.
In an attempt to find a pattern I looked at the time when these URLs were shared. Interestingly the top five best performing URLs were all created between 9:23 am and 12:45 am. This could indicate that URLs shared on Twitter before lunch in the UK get the most clicks.
However, of the five least clicked URLs four were created in the same time frame, and only one fell outside it (4:27 pm). Additionally most clicks don’t originate from Europe (see below), so there doesn’t appear to be any solid correlation between share time and popularity.
Next I looked at the day of the week when the URLs were created and shared. Again I found no correlation – the five most clicked URLs neatly spanned all working days of the week (Thu, Fri, Wed, Tue, Mon) and the five worst performing URLs were also shared across most of the work week.
The only notable issue here is that there are no cases – best nor worst URLs – that were shared on a weekend. This is almost certainly because I rarely tweet on weekends anyway.
Unsurprisingly nearly all clicks on a short URL happen within the first 2 hours after the URL has been shared. Traffic then drops off significantly, usually lingering on with a few clicks here and there for a few days.
It’s quite interesting to see that many URLs continue to receive traffic for weeks after they’ve been created, accumulating the odd click here and there despite no longer being actively shared on Twitter. Sometimes a short URL may seem ‘dead’ for a week and then suddenly it’ll receive a few clicks again.
The traffic sources report in Yourls makes one thing abundantly clear: people in my Twitter network (my followers and their followers) make extensive use of third party Twitter apps. Clicks from these apps register as direct traffic as there is no referral string sent along with the click. At least 90% of all clicks on any given short URL registered as direct traffic, with the remaining clicks mostly composed of twitter.com referrals and a few others (HootSuite, Netvibes, Wikiwix, etc.)
Apparently I have at least one cautious user in my network: Longurl.org consistently pops up as a referrer. This is a security service that allows you to un-shorten a long URL without actually visiting it, thus ensuring you’re not being sent to a hazardous webpage. I’m not sure whether this is an actual user or an automated check.
Surprisingly my Twitter network is apparently very US-dominated, with usually more than half of all clicks on my short URLs originating from US IP addresses.
Other countries that are strongly represented are the UK, the Netherlands, Canada, Germany, Ireland, and France. Russia, China, and Japan occasionally show up, as do nearly all other European countries. Africa and South America are notably absent from most URL location charts.
I installed a new plugin today – Feb 1st – that filters out the clicks generated by automated bots such as the Googlebot, the LongURL api, and Twitter’s own bots. This should help make the numbers more accurately represent actual user behaviour and might also clean up the traffic locations report.
However, I am getting married at the end of February and will be offline for well over a week, so my stats probably won’t be sufficient to dedicate a whole blog post to. I’ll likely wait until March has come and gone before I publicise my next set of short URL stats.