• by Blake Cahill , July 6, 2010

From focus groups to phone polls to man-on-the-street surveys, companies and organizations have been trying to understand consumer opinions for decades. And with social media producing a steady stream of updates and opinions from the masses every second, businesses have discovered yet another way to attain this valuable data. But distilling actionable information from social media often requires a deeper analysis, and this is where sentiment analysis comes in.

However, sentiment analysis is almost as misunderstood as it is talked about. To help marketers understand how to successfully understand consumer opinion in the social space I have outlined some common myths and best practices to consider.

Myth: Sentiment is Sentiment is Sentiment

"Sentiment" is actually a general term that can mean a lot, depending on how it's measured and translated into a piece of the social media puzzle.

At a very basic level, sentiment is an opinion that is deemed positive, negative or neutral. But just being able to put a tweet into a "positive" or "negative" category may not be enough for your company to craft a meaningful response -- typically, to really understand the implications of a customer voice sentiment analysis you need to go deeper. To do this, you can frame sentiment as a reaction to a defined subject.

In other words, gauge whether the negative comment on someone's blog was about your company in general, a particular product, or even a facet of that product. Another way to make sentiment analysis more actionable is to define the "voice" of the opinion. Are all negative opinions coming from the same user or type of user? What are they saying that other customers aren't, and how can you personally reach out to this audience?

Myth: Technologies are Technologies are Technologies

To understand the general sentiment of the masses, some brands are turning to technologies that automate sentiment analysis -- but not all sentiment technologies are created equal. Automated sentiment analysis is really only as good as the human thought that goes into it.

For example, if your technology measures sentiment based on a list of words that have been programmed in as "positive" or "negative," those words need to be appropriate and relevant to your industry. On the other hand, you could "train" your technology to identify sentiments based on a sample data set that has been scored by humans. This approach may be more accurate overall, but still depends on that initial human input of what's positive and negative.

And unless your technology can aggregate and present the results in a way that makes sense to the humans in your company (who are, after all, the ones using the results to reach out to consumers), it's not really helping you.

Myth: Sentiment is Black and White

Sentiment is a uniquely human phenomenon. If the sentiment analysis technology scores a blog post a .995 on its negative scale, it's likely that a human would recognize that post as negative -- but it's not guaranteed. Technology that orders its assigned scores from high to low gives your service team a better chance of finding most negative input, but it still won't be absolutely perfect. Which brings us to our next myth ...

Myth: "Our Sentiment is 85% Accurate"

When faced with such claims, it is fair to ask, "accurate compared to what?" A technology might correctly decide whether a certain word indicates a positive or negative opinion 85 percent of the time, but that doesn't mean the automated process will perfectly align with what you're trying to measure 85 percent of the time. In addition, beware of technologies that claim accuracy when defined against non-social data sets or "expert" answers -- these data sets are not measuring what you want: Social data defined by your company's needs.

Myth: Sentiment is Valuable

With all the buzz, it's easy to believe that sentiment is the key to instant social media enlightenment. Really, sentiment is just one facet of a huge (and consistently growing) channel where a lot of other factors come into play. Social media strategy requires a broad, deep, enterprise-wide framework that will help guide your company into the conversation.