Social Media Sentiment: Bells, Cups, or Shark Fins?
April 30, 2010 | 2 Comments
When do consumers choose to talk about brands in social media?
1) When they’re really, really annoyed and need to vent.
2) When they’re surprisingly pleased and want to share the love.
In fact, those are the messages people love to read. The hate, the disgust, the love, and the adoration make for interesting and provoking reading. These types of messages are also really fun to share with other people because the extreme opinions strike a very deep cord.
As with all things though, there are more than two options. There must always be the neutral or disinterested or uninvolved vote. Some people have wondered whether this third option has any meaningful presence in SM data. Is the neutral vote simply overwhelmed by all of the extreme opinions, or does it actually show up in significant counts? Do we have a bell curve of emotions ranging from negative to positive or a cup curve of emotions that only includes very negative and very positive?
We evaluated the sentiment of three different brands to determine the distribution of 5000 randomly selected scores for each brand. We chose one brand that should skew negative, one that should skew positive, and one that should sit nicely in the middle.
#1 Toyota - We selected Toyota for the obvious reasons. All of the recalls they’ve experienced recently haven’t caused a lot of positive sentiment towards the brand. However, you can clearly see that though the negative side of the curve is well filled in, so is the positive side and the middle section, the neutral section. There is no denying a well rounded bell curve for this set of data.
#2 Taylor Lautner - Again, Taylor was selected for obvious reasons. He has received nothing but positive reviews as of late. It is clear that his sentiment lies more closely on the positive side of the curve, but once again, it does not pile up on the most extreme position. Even among the positives, there is a wide distribution of sentiment.
#3 Walmart – Finally, we selected Walmart as a solid overall brand that hasn’t received any particularly positive or particularly negative review as of late. The distribution of scores is fairly even, and focuses neither solely on the extreme positive nor the extreme negative.
It’s clear that SM does more than elicit extreme positive and extreme negative opinions. In fact, in all three cases, the distribution of opinions ranged nicely from extreme positive all the way to extreme negative. Though the mid-point of the distribution shuffled a bit from left to right, the overall distributions were similar. They may not be perfect bell curves, but there is no other curve that better describes the resulting data (except perhaps a shark fin curve for the Walmart frequency!).
In the offline world, people are more than extremes. They express a wide range of emotions from positive to negative with all points in between. We have just seen that this range of emotions transfers to the online world as well. We’ve definitely got a bell curve of emotions.
Category conversition | Tags: bell curve,conversition,cup curve,distribution,evolisten,frequency,market research,mr,sentiment analysis,shark fin curve,smr
Social Networks : Share
2 Comments so far
by Bob Fichtner
On April 30, 2010 at 4:32 pm
Hmm… Is this is based on automated sentiment scoring? If so, then I would question the conclusions drawn.
http://www.socialmediaexplorer.com/2010/04/26/trusting-automated-sentiment-scoring/
by Conversition Team
On April 30, 2010 at 4:42 pm
You’re right that there is a lot of discussion around sentiment analysis and what processes work better. You have to make sure that the processes you use are valid and reliable. It’s no easy task to do this but a solid research foundation turns automated processes into quality processes.