Over the AFL Grand Final weekend we were pretty busy doing a Twitter data visualisation for an article by the Herald Sun but there's so much more in this data set that we decided to write a follow-up post as well.

Have a look at the below video or the interactive Tableau dashboard behind it, showing volume of Twitter mentions for the different AFL teams over a period of time leading up to the Grand Final. Further down are some screenshots of the Ross Lyon (St Kilda coach) incident, in which he was accused of stabbing his predecessor Mark Harvey in the back for the position of coach at the Freemantle club - the whole story was actively discussed on Twitter as you can see.

The video (and dashboard) present what we call 'Twitter Time', all Tweets broken up into chunks of 400 and then analysed. As the chunks of Tweets (blue lines) are visualised, time will slow down at moments when the volume of Tweets was high, particularly at match dates (green circles) or at the Grand Final (dark green circle). The jersey size is proportional to the number of Tweets mentioning a particular team, and you'll notice that they go gray once they are eliminated from the series. The analysis was run on about 35,000 Tweets provided by Alterian's SM2 social media analytics platform.

Below is the breakdown of the Twitter chatter and mentions of key players, and most importantly coaches, from the time Ross Lyon announced that he was resigning and rumours started about his move to Freemantle (15th and 16th September). All Tweets were analysed for instances of player or coach names. Rumours pop up and die as Twitter volume rises and falls.

Above: Hawthorn vs Sydney & West Coast vs Carlton normal match chatter on Twitter.

Above: rumours of Ross Lyon's move to Freemantle and sacking of Mark Harvey explode on Twitter after Ross Lyon officially resigns. His move to Freemantle is confirmed on the 16th of September.

Above: Neil Craig becomes the topic of conversation as well as it is rumoured he might go to St. Kilda.

Above: AFL fans move on and resume normal match chatter ...

Partnered with Alterian using their business intelligence product SM2 to extract all the information on each Tweet, a database of all word pairs was created (over a million!). The word pairs, like 'asylum seekers' or 'boat people', were generated from each Tweet independently and then tallied up over all Tweets. Common words like 'the' and 'it' were removed, and a stemming algorithm was used to group words such as 'Australia', 'Australian', or 'Australia's' together. All Tweets were treated equal and all Retweets were included so that the content of the most popular and followed people on Twitter would emerge via Retweets.

Once the top word pairs (based on a tally) were finalised an open source software called Gephi, which is a powerful tool for visualising and analysing large networks, was used to present the data. See below for our first attempt; each word is connected to the words that were paired with it, taken from the the top word pairs. The size of the words is related to how many other words are connected to it (not how mant times the word pair appeared in all Tweets).

The whole network is below and shows how the different words are associated. The word 'Raquel' is at the centre (and is the largest) because it was associated with the most words. Many interesting word associations come out of the data. For example, there is a sub-network with words 'live', 'exports', and 'corners' (top left) most probably comparing the SBS Go Back series to the Four Corners program that exposed the live exports trade.

One thing that you might notice is that there are pockets of networks that are associated with a particular Tweet that was Retweeted a lot. For example the Tweet below is associated with the sub-network (bottom middle) that contains words such as 'no', 'vote', 'mad', 'point', and 'court'. You can search Google with any combination of associated words along with the word 'gobacksbs' to find the Tweets that made up the data.

Exact Target in combination with CoTweet has released a pretty interesting research series about the interdependencies of email, Facebook and Twitter.

One of the key insights form the study for me is that while email, Facebook and Twitter compete with one another for marketing budgets, consumers really expect brands to interact with them across all three channels.

That said, it's interesting to see that Twitter consistently scores higher than email and Facebook. For example, over 30% of Twitter followers are more likely to purchase or recommending a brand after becoming a follower whereas email and Facebook score more around the 20% mark.

However, I'm not sure I agree with the channel breakdown by reach, retention and acquisition that the study suggests below, the results may support this interpretation but the real world seems a little more difficult.

Download the full research report series from the official Exact Target & CoTweet website.

The guys at Gigya created a great info-graphic showing what accounts and online identities people are using to sign-in online.