Sentiment Analysis … So What?

In early January, Next Generation Market Research lead Tom H.C. Anderson shared over 70 predictions for 2012 from clients, Research company CEOs, text analytics providers, and recruiters. I’m always interested in looking for themes, and I’m certainly curious for perspective on where our industry is headed, so I did a little text analysis of my own with the verbatims that Tom posted on his blog.

Many of the predictions were fairly predictable … lots of focus on Social Media and Mobile Research, along with some comments surrounding emerging methodologies like Neuroscience and Gamification. The idea of consolidating data was also frequently mentioned. As a qualitative researcher, the text analytics conversations were particularly interesting, so I took a deeper look into those comments. Roughly 22% of the predictions referenced Text Analytics in some way, (skewed a bit by the fact that over a third of the sample represented Text Analytics companies.)

Many of the comments tied to the larger concept of Social Media analysis, and using text analytic tools to extract insights from this vast amount of unstructured data. Several talked about the need for “education” and focusing on user training to remove the “black box” mystic surrounding many text analytic tools. Sentiment analysis also factored into these predictions for about 11% of the sample.

While many were just general comments on the importance of measuring consumer sentiment, a couple took the idea a bit further, expecting to see some pushback over the next 12 months:

• “It’s not that marketers don’t care how consumers feel about them but that there’s not much they can do with a pie chart of percent positive neutral or negative sentiment. The sharpest marketers will be looking for ways to make sentiment analysis pay.”

• “Businesses who pay for sentiment analysis are going to start asking ‘So what?’”

These comments regarding sentiment analysis were particularly interesting to me given the work we do at Quester. I’ve been asked at least three times just this week about sentiment analysis, and how Quester applies it in our practice. Although we have a suite of text analysis tools called Aristotle to help us organize and process the data, which significantly reduces our turnaround time, the majority of the Quester analytic process is done by hand by our team of linguistic analysts.

Because our analysts are digging into the data, really understanding the key themes and linkages between those themes, they can tell the story of what is going on in the data … not just in terms of frequency of mentions, or by using an algorithm to assign a “score” on the sentiment, but by communicating the nuances in the way the respondents are using language to communicate their thoughts.

It’s this in-depth review of the data, really “listening” to what our respondents are saying, that enables Quester to uncover the key insights within that unstructured data … ultimately making the “so what” crystal clear.

That said, we are always open to the possibility of incorporating new techniques and methodologies, so Quester is undertaking an investigation on the currently available sentiment tools. We’ll be sharing our findings in an upcoming post.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.