Selfies, Sentiment and Social Media Stats

I answered this question on Quora and happened to see Indian news channels flaunting social media stats in National elections. That’s what inspired me to start writing this post. Why was the #Oscar selfie such a big deal and why is social media analytics so hot today?

History was created on twitter when a tweet by Ellen DeGeneres was retweeted over 3M times. As the MoC, she clicked a selfie with some of the nominees and superstars at The Academy awards. This act was seemingly spontaneous and the picture was clicked using a Samsung Note3. Samsung was a sponsor for the show. The host was using the Samsung device on stage. The product placement technique used was standard, the picture clicked was average, perhaps that moment was magical.

This tweet was a big deal for twitter, for Ellen and perhaps for the academy awards. I did not understand how this was a big deal for and by Samsung, especially looking at the original tweet.

Samsung deservingly piggybacked on the wave and got some traction. So did many others in their own sweet ways.

Most news channels and popular blogs have been tracking the voice of the Indian citizens on social media as a run up to the general elections 2014. What is commonly shown across these trackers is a series of charts that compare the “share of voice” of certain politicians v/s that of other contenders. They further go on and publish the sentiment: X% of the mentions are +ve and Y% -ve. Some even show word clouds, trend lines of mentions, gender and geographic split etc

Social Media Statistics
Most of these trackers and news stories are backed by some very solid social media monitoring tools. The last time I checked there were 350+ such tools helping you solve different types of listening and monitoring problems on the web.
What disheartens me the most on reading these news articles and ‘analysis’ is the fact that most of these tools are being used to perform shallow reporting. Let’s get one thing straight: reports, analysis and insights are not the same. Crawling data points and visualizing them as charts is perhaps one small feature most evolved listening tools provide you with. What they also offer the user is the ability to mine insights by analysing the data in a much deeper sense.

While its interesting to be a part of the historic tweet by Ellen DeGeneres, its sort of funny to link the success of the tweet as a return on investment for Samsung’s sponsorship efforts. Sure, it did get Samsung a large number of mentions, but it did that for Apple as well ( Ellen’s actual phone was an iPhone5s). The interesting insight would be how many RTs appeared before the first news/blog article was published that educated people about the device being a Samsung Note3 ( those not watching the show). It would be worthy finding the number of RTs from the USA timezone and ones nearby that were posted +/-20 min from the time the selfie was shot/posted: perhaps get an indication on viewership and causative influencers.

News channels on the other hand just show a split of sentiment and share of voice without explaining the universe they are measuring. Some of these listening tools can actually be deployed to mine insights that will indicate the key issues that concern citizens in various constituencies and how they relate these with prospective candidates from those constituencies. Perhaps a real time variant of exit polls may emerge. Sentiment % without reference is baseless. A higher share of voice may not always be an indicator of popularity. Publishing such stats without context on national television can be misleading and dangerous at different levels.

I’m not taking away the credit from Samsung’s marketing team on the feat, neither am I doubting the analytical capability of the editorial teams at our national news channels. All I think that it’s time to move ahead from the “social media and big data are like teen sex” phase. Its time to apply analytical frameworks and statistical models to build rigor into our work that may eventually culminate into a harmonious fusion with offline data models. That is when we would have ‘arrived’.