This weeks #SocialMedia Tweetchat Topic: Sentiment Analysis-Opinions Matter, If Only You Knew Which Ones


Listening is the first step in social media (everyone says so).  Not only do you have to listen, you have to listen for 6 months or more before you are allowed to do anything.  Just ask the experts!

Frankly, I think everyone says that just to buy a little time before they have to really figure out what to do with social.  At any rate, most of the people who are told to listen have no idea what to listen for or who to listen to.  I’m not going to get into the depths of all things social media monitoring because that would take all day.  So let’s focus a minute.  

  1. You want to listen for mentions of your company, brand and top executives
  2. You quickly determine there is no way to manually search every blog post, tweet or comment on the web so you turn to automation
  3. Yeah, now you’re tracking buzzzz, but what does it all mean?
  4. So you start running reports and determine they are inadequate at best.
  5. Now you’re back to listening again but still not sure what you’re listening for.

 There is a word in the industry called “Sentiment” that is used when trying to determine a person’s attitude.  Online it’s a digital attitude and you only have text to go by.  No voice inflection.  No hand gestures or facial signals.  Just a bunch of words (or “noise” as they call it in the bubble) with little signal.  The challenge, after aggregating all of the buzz or mentions of everything you are tracking, is to make sense of it all and to make it actionable back inside your company.  So the sticking point here is whether or not you can use automated analysis to provide sentiment or if it has to be all human interface.  For any local or small business, human processing of sentiment might be reasonable.  However with any size at all, you would need a small army to determine if people liked your new product or enjoy working with your company…or would you?

If you ask 10 people how to measure sentiment, you will most certainly get 12 answers (yes 12).  The popular themes of managing sentiment revolve around polarity and intensity.  Polarity meaning either good/bad, positive/negative, like/dislike, etc and intensity meaning the volume or amount of mentions.  These are not wrong by any means, but I use a little different formula and you might say it’s probably for different purpose.  I like to consider the following:

  • Mentions – which is broken into volume, intensity and opinion (polarity)
  • Influence – of the person it comes from. How many followers, how often they interact (like a TwitterGrader)
  • Severity – of the content itself. “X product just saved my life or killed my brother” would be Sev1, where “Boss caught me goofing off and fired me, X company sucks” would be low severity.  Further defined by a direct vs indirect mention and context of the content.

OK, try managing that formula through reports.  No way, Jose!  And, by the way, I usually change what I am monitoring (at least the focus) to match what I am working on.  There are companies who are working on ways to automate forms of sentiment through natural language processing and machine based or community based learning.  They have their claims on successes and what they have may work for a lot of people in a lot of situations.  It has to be an individual call.  So how do you know what’s right for you?  That’s where this week’s moderator Katie Paine comes in.  Katie, affectionately known as the “queen of measurement”, spends most of her day answering these questions for her customers.  She will host our next chat with the following topic and questions:

TOPIC: Sentiment Analysis: Opinions Matter, If Only You Knew Which Ones

Q1:  How do you define positive sentiment?
Q2:  How does that impact your organizational goals?
Q3:  How do you know that what you are measuring matters?

Please join us Tuesday 02/16 at noon est and become part of the conversation.  Learn insights and have an opportunity to capture Katie’s attention for a solid hour.  Follow along using #sm47 or simply go to our LIV

Posted via web from marcmeyer’s posterous