Analyse This!

by John 15. November 2011 01:56

We are all aware of how the explosive growth of social networking and publishing tools has transformed the way we interact online. The raw numbers are pretty phenomenal. At time of writing some of the stats are as follows:
Facebook - 800 million active members with over 50% logging on on any given day
Twitter - probably has over 300 million accounts - with over a billion tweets being posted per week
Blogging - there are an estimated 152 million blogs worldwide.

The volumes of data being generated by these tools is staggering and has not yet peaked. However, the rise of these technologies has not just impacted on the way in which we interact and communicate as individuals. They are also re-shaping the relationships between individuals and corporations. This is happening in two ways:
i) they provide a feedback mechanism which corporations can tap into to better understand their customers
ii) they have actually produced a jolt in the balance of power between producers and consumers, with online opinion leaders able to exert massive influence and potentially make or break a brand. You need only look at the trepidation felt towards Tripadvisor in the hospitality industry to see this at first hand.

Obviously, any clued up company is going to want to plug into this feedback loop - both to find out what customers are saying as well as to participate in the conversation. The major challenge is one of scale - with 30 billion items of content being uploaded every month to Facebook alone, you are going to need a very big sieve to home in on the information you need. Once you have found that data you need to present it and analyse it. This is a task which is going to require some heavy lifting and some intelligent sifting.

Step forward the Microsoft Engagement Client. This is a service which runs on Microsoft's cloud-based Azure platform. It gathers together feeds from Facebook, Twitter, the blogosphere, Youtube and other sources and presents the results in a browser-based app (running in Silverlight). At the moment the product is only in the experimental phase but Microsoft are inviting developers sign up to the evaluation program.

When the Engagement Client first loads the interface is minimalistic - perhaps even a little cryptic. You see a set of icons in a panel on the top left hand corner of the screen.

The functionality behind the button with the Twitter logo is not hard to guess - however the others are more enigmatic. Clicking on the filter icon actually brings you to the web site's home page rather than allowing you to refine the data being displayed. As I mentioned - the app is very much at the experimental phase so presumably the code behind this is still in the works.

Anyway, to get started you click on the + button. This gives you the option to add a new Filter - i.e. select a pre-defined search term for filtering the data from the various channels (Facebook, Twitter, blogs etc).

We are going to choose 'Windows 8 and Applications' - and once we do - bazinga! The column populates with all of the most recent tweets, blog posts and Facebook interactions relating to that theme.

That sounds simple but when you think about it is actually quite a feat - every interaction across multiple social networking channels gathered right there at your fingertips in one streamlined space in a matter of seconds.

At first glance you can be forgiven for thinking that the client is only collecting interactions from Twitter. Once you start scrolling down though you will see this is not the case. It is just a reflection of the fact that the number of tweets on any given subject far outweighs the number of blog posts, Facebook comments etc (and by a pretty large factor) - so that visually the tweets seem to drown out the other channels. This is why it would be really useful to be able to filter the results by channel - and it's hard to believe that this feature won't figure in the finished product.

Let The Analytics Begin
Now that we have our filtered stream we can start to dig a littler deeper. Each item appears in its own window which is packed with an array of icons. One of the most intriguing is the small face which appears on the left hand side under the user's avatar. One of the major innovations in the Engagement Client is Microsoft's "sentiment engine". This is a tool which analyses the language of each message and attempts to evaluate its tone. The expression on the face then reflects the results of this analysis. Where there is a thread with multiple posts the sentiment engine will attempt to evaluate the overall tone of the thread.

Conversations
The client doesn't just present you with a data dump of all interactions matching the filter. It actually tries to structure them into conversations. So a re-tweet or a response to a tweet is grouped together with the original tweet:


But that is just for starters - the client also attempts to tie up threads of conversations across multiple channels. In the screen shot below you can see that entries for a blog post and a tweet by a different user linking to that blog post are presented together.

This is a pretty impressive attempt at imposing some order on the teeming, sprawling chaos of the social web.

Some Closing Thoughts
The challenges for Social Analytics can probably be summed up as:
i) inferring meaning
ii) identifying relationships
iii) presentation of data
iv) analysis

So how does the Engagement Client shape up?

Inferring Meaning
The incredible processing power at the disposal of modern PC's means that today's applications can perform some very clever tricks simply through the use of brute force algorithms. However, these algorithms reach their limits in situations where judgement or intuition are required. Even more so when attempting to tease meaning from raw material as sparse as a tweet of 140 characters or less. On the occasions where the sentiment engine makes a judgement call its hit rate is pretty impressive. This in itself would be little more than an amusing parlour trick if it wasn't for the fact that sentiment can also be judged across the entirety of a conversation.

Identifying Relationships
The client performs well in tying together items from various channels which are organically related - i.e. a tweet which relates to a specific blog post - as opposed to just trawling through swathes of data to find text matching a specific search criterion. Documentation on the Engagement Client is sparse and our usage of it has been limited so the level of sophistication involved in intelligently and accuratetly re-constructing 'conversations' (rather than grouping random messages by theme) is a matter of guesswork at the moment.

Presentation of Data
There is an inherent difficulty in attempting to condense the multi-dimensional buzz of cyberspace into the two-dimensional confines of a computer screen. The engagement tool can present individual conversations but at the moment there appears to be no way of placing them in a larger context. On the issue of usability I am not sure exactly how intuitive some of the icons are but that is not a difficult issue to address

Analyis
For social analytics tools to have business value they will need to be able to zoom out from individual conversations and microcosms and identify patterns and trends at the macro level. They will need to identify the tone not only of a conversation or thread but somehow draw overall conclusions from the ceaseless babble that represents the bigger picture. This will require sophisticated visualisation tools with powerful drill-down capabilities. Even this, however, is only the beginning. If we can engage in some future-gazing for a moment we can note that Facebook and Google are not only producers of social content. They are also vast repositories of demographic and behavioural data. The Great Leap Forward in social analytics may lie in integrating these streams of trend, demographic and behavioural data to turn the social web into a living, breathing decision support system.

Add comment




  Country flag
biuquote
  • Comment
  • Preview
Loading


About the authors

This blog is written by staff, partners and associates of Blackstairs Mountains Software Limited - creators of the Context Project and Resource Management System. Any opinions expressed may not necessarily reflect the opinions or policy of the company itself.

Month List

Page List