Is Your Social Listening Program Hampered By Ineffective Sentiment Analysis?
August 25, 2015
Although social media is an important and growing source of customer data, some marketers find themselves struggling to effectively listen to, measure, and use the information available through social channels. Worse, they are unable to capture perhaps the most valuable aspect of social feedback—the sentiment expressed by individuals. Understanding the emotions being shared lets you understand the motivations, frustrations, and desires of your customers.
Capturing accurate sentiment data and effectively including sentiment in a social listening program poses a challenge to many businesses, but it’s definitely possible. In fact, accurate social sentiment analysis is not only possible, it is critical to successful social listening. In order to be truly effective, though, sentiment analysis shouldn’t happen in a vacuum. Businesses need to take the following actions to make sure that sentiment analysis is accurate and effective.
Measure the right things
If you don’t know what to measure, you will certainly miss out on the insights that are available from social listening and your sentiment analysis will undoubtedly be inaccurate. You can start by measuring:
- WHO is talking about/to you (age, gender, nationality, language)
- WHAT are their issues (specific pain points, recurring questions)
- WHEN are they talking to you (business hours, weekends)
- HOW are they talking to you (sentiment, type of language, tone of voice)
- WHERE are they talking to you (source, region)
As you begin to notice trends, you’ll be able to use that information to refine your strategy. If a new topic bubbles up, you can swoop in and address it quickly. Customer engagement will increase, which impacts sales and customer retention—so measuring the right things is a critical first step in proving your social listening program’s ROI. As in all kinds of customer analytics, specific numbers and percentages are less important than benchmarking your progress and seeing improvements as you go.
Integrate social with other sources
Social feedback has certain defining characteristics. It is:
- Highly emotional
- Full of hashtags, abbreviations, and slang
Therefore, social feedback must be analyzed in a way that takes these features into account. At the same time, social is merely one source among the dozens that are available to today’s businesses—one Clarabridge customer uses more than 50 different sources of feedback regularly. Taking an omni-source approach to customer analytics (including all kinds of data across all channels) while respecting the differences of each type can be tricky, but it is necessary to view your social feedback in the context of all of your customer data if you want it to be relevant and actionable. The Clarabridge CX Intelligence platform acts as a hub for doing this sort of omni-source analysis, with connectors to internal and external data sources.
Choose the right solution
There are many ways to approach sentiment analysis (for social or any other kind of feedback). Recognizing the limitations of other methods, Clarabridge developed a proprietary technique that combines lexical and grammatical approaches to maximize accuracy. The system understands how sentiment is altered or intensified by taking into account grammatical constructs and function words. This gives Clarabridge the unique ability to correctly understand the trickiest parts of your feedback:
- Negation: Recognizing that adding “not” changes the positive sentiment of a word like “happy” to a negative one (and vice versa).
- Conditional sentiment: Adjusting the sentiment of the text when phrases like “could have been” or “used to be” appear.
- Amplification: Adding or deducting sentiment scores in the presence of modifiers. For example, “The food was great,” (+2 positive sentiment) jumps to a +3 positive when written as, “The food was so amazingly great.”
- Context: Recognizing that in some cases a word (like “thin” or “sick”) is positive, while in others it is negative.
- Exceptions: Automatically considering the oddities of the English language that have a bearing on sentiment. For example, adding the word “too” to a neutral word (“too orange,” or “too tall”) makes it a negative. There are over 500 pre-existing exception rules within Clarabridge.
In addition, Clarabridge indexes the sentiment score on a normalized minus five (-5) to plus five (+5) scale. This provides significantly greater accuracy over the common “positive,” “negative,” or “neutral” rating system—since we all know that there is a big difference between a good meal (+1) and the best food you’ve ever eaten (+5).
For further analysis, users can “tune” the sentiment of particular words and phrases to match their own experiences, perform sentiment filtering to isolate the degree of positivity/negativity, and do root cause analysis to determine what is driving the sentiment.
With the right measurement, the right context, and the right solutions, your social listening program will be a valuable source of business intelligence and customer engagement that won’t be hindered by sentiment analysis—or anything else.
If you want to learn more about Clarabridge’s proprietary language analysis methods, check out our white paper: The Truth About Text Analytics and Sentiment Analysis.
Lisa Sigler is Sr. Manager of Content Marketing at Clarabridge. For over 16 years, Lisa has used her writing and editorial skills to bring the value and benefits of technology to life. In her current role, she works to demonstrate Clarabridge’s position as thought leader and trailblazer in the Customer Experience Management market. Lisa holds a B.A. of English from Kent State University. Read more from Lisa on Twitter @siglerLis.