Chris Alexander

On Engineering

My Social Media Experiment

19th August, 2009

Now it’s time for me to fess up. The other day I performed a little experiment on you people, or at least my Twitter followers.

If you frequent my blog recently, you may have noticed something slightly unusual on one of my posts. I fairly frequently post pictures (more often than not, lolcats) to spice up the page a bit and make it a bit more exciting. One of my other recent posts, however, went a bit further - I’d be surprised if the page-high, full-width photo of an oft-admired actress wearing a summery dress didn’t catch your attention.

(No alt specified)

So what was the exact nature of my little investigation? Well, I tweeted the link to the post out twice, with different tweet contents and links masked with the help of Bitly, to see if the responses would be different. The first seemed fairly inconspicuous:

(No alt specified)

The second, not so:

(No alt specified)

Predictions

So, like any good GCSE Science student (at least before they removed all the science from the curriculum, but that’s a story for another day), here are my predictions for the experiment. Simple really: I was expecting more clicks on the second link, sooner and faster. I was also expecting to see a higher proportion of the entrances into the page to be from the homepage than usual (given the image, at least).

Results

Well first of all to the click data from Bitly. I created two unique links (as you can see on the tweet pictures above) so that I could track the clicks from individual tweets. Surprisingly, they were very close - 17 for the first tweet and 16 for the second tweet. That I wasn’t expecting! These are the graphs for the two tweet links (in order) for clicks over the first hour. As you can see, they’re pretty even too.

(No alt specified)

(No alt specified)

So on to the page entrances. For this post, 15% of page entrances were from the homepage of the blog. Compared to the top 5 blog posts (over all time) for the site, this is slightly up from the 11% average (with a standard deviation of 9.3%) .

It should also be noted that the blog post is outside the top 20 blog posts by number of page views, and only just inside the top 20 blog posts by unique page views.

Conclusions

So it seems my saucily-worded tweet did little to affect the number of people coming in to the page. However it does seem that the large picture on the front page of the blog slightly increased the number of click throughs to the post itself from the homepage.

This is, of course, omitting many variables that have not been taken account of. The tweets were posted at very different times, and it would be impossible to post both links while completely removing this factor. It could possibly be minimised by posting the links at the same time on the same day of the week, but a week apart. While I tried to aim for the peak times on Twitter (when people are arriving at work and when people are beginning their lunch breaks) that does not give each link the same start.

Another improvement could have been to use more social networks to investigate the differences between them. In all honesty I had meant to post the links on Facebook at the same times I posted them on Twitter, but I forgot until I went to post the second link and by then it was too late.

All in all it was a nice little insight into what was going on, but apparently my thoughts on the effects of messages in tweets clearly needs some work!