Sentiment is a funny topic. Why is it so important and why do we rely on tools to tell us what it is and what it means? Why are we content to assign a positive or negative score to something, but resist the idea of understanding the ideas presented within the comments themselves?
This has been an interest of mine long before social media monitoring came on the scene. The Knight Digital Media Center offers great perspective on the idea of distilling sentiment from user comments left on blogs or mainstream news sites. We’ve all seen them: those nasty, sometimes hateful reactions and commentaries. All contain some truth or insight. It’s up to the reader to determine what that is. Social media shouldn’t be any different.
I hope to start a dialog with this blog post, because after all, that’s where sentiment begins to break down, right before the conversation begins.
Tip #1 – Forget about automated scoring
It’s bunk. Extremely few of the social media monitoring (SMM) tools on the market do this idea any justice. Most, on average, seem stuck in the 60-70% accuracy range through Boolean keyword matching. Others promise closer to 80-90%, but these often require natural language processing (NLP) technology that doesn’t yet exist in most SMM products.
Pick any tool you might use to score sentiment. Look at a recent range of verbatims in that tool. How many scores are just plain wrong? Consider these tools as directional aids for evaluating your social content, no matter its source. How many offer the ability to override the automated score?
What about Net Promoter Score (NPS)?
In 2011, a hot topic started to coalesce around creating a “social NPS.” It doesn’t exist. At least not yet. Social sentiment and NPS are two very different things with very different parameters. Social sentiment scores are ostensibly based on very organic responses to some stimulus (earned and unearned), but in reality are tainted by certain influences. I see a correlation to what a brand says and how consumers respond. NPS is a very controlled means to measure customer satisfaction.
In any case, automated scoring runs the risk of precluding a marketer from truly understanding what’s behind a sentiment, what the consumer might be thinking, and what the consumer isn’t saying.
Tip #2 – Leverage sentiment as an “early indicator”
Social content is great for surfacing ideas that you might not be thinking of, nor being aware are out there. Use existing sentiment engines to identify and develop trending models for any topic of interest. These could be product reviews (good and bad), support issues, feature requests, competitor insights, and so. Use sentiment as a directional sign post to take the temperature of any given environment.
Tip #3 – Engage with all sentiment types
We all feel the need to react to bad situations, whether an emergency situation or a PR crisis. If someone out there is critical of us, we want to defend ourselves. That’s human nature. What’s uncomfortable is engaging with unfavorable sentiment. That should be job #1 for social media managers, PR departments and thought-leaders. Dig into why folks are responding as they are. Embrace your detractors.
“Pig Pile Syndrome”
Crisis moments might seem obvious (scandals or natural disasters), but sometimes people will simply complain to complain without any particular context. If enough comments of a certain type are appearing, it might be impractical to respond individually, but assuming one or two primary points can be addressed (good or bad), a blog post or YouTube video might help make a connection. Then you might engage with people still interested in what you have to offer.
Tip #4 – Don’t look at sentiment in a vacuum
Consider what stimuli are contributing to sentiment. Can you correlate sentiment trends with other activity? Is that activity causing sentiment? If you see a spike in sentiment-based activity, what’s driving that commentary? Did you announce a product? Issue a press release? Did a disgruntled employee post a YouTube video? Did an unhappy customer? These are but a few questions to ask yourself.
I often see three waves of activity in response to major topics as they propagate across the social web. The first is reaction. Whatever the topic, folks will react and often their followers will react to those reactions. The second is blame. Once reaction trends are established, folks will join the conversation (often ill informed) issuing blame. The third is analysis. Everyone loves to be a Monday morning quarterback. Sentiment swirls around these waves. Pay attention to the drivers and understand the meaning behind them.
Tip #5 – Use caution in defining sentiment as a key metric
Not only is sentiment a very subjective matter, the review of sentiment is also so. If you choose to consider sentiment as a business driver, set clear expectations around what you’ll learn from it. It’s not enough to report on positive versus negative mentions. At the minimum, juxtapose sentiment over time with social activity over time. Conceptually, these two measures should increase or decrease somewhat proportionally. Annotate spikes or outliers or anything else that can provide context to the meaning of sentiment. Haters and brand advocates alike can disproportionately skew results. Above all else, know what’s being said about you and who the primary voices are who might be influencing conversations.
Will we see improve sentiment technology in 2012? We’ll see. This remains a hurdle for most SMM vendors. Vendors traditionally in the customer voice space will fare well here. Big data companies are exploring the idea as they develop solutions to analyze social content. I predict some strategic partnerships and perhaps and acquisition or two amongst these categories in the coming year.