I am frequently involved in discussions with brands about measuring the impact of conversations in social media and networks and frequently point out that one of the key metrics to understanding and measuring a baseline or ongoing change is tracking the sentiment of what is being said by consumers about your brand, products and services. Here at Visible Technologies we have always believed this to be one of the key ingredients of any successful social media measurement technology or program and why we have always provided it since even the early beta’s of our truCAST products. Recently, I had a great discussion with David Burcham, one of our EVP’s and thought the perspective was worth sharing.
Blake Cahill
Why is sentiment important and how do you define and measure it?
David Burcham
Sentiment is just one measure in the overall work of keeping track of your customer’s perspective online – but we have always believed that it’s an important one. When you’re spending millions of marketing and service dollars based on information, we believe that it should be as accurate and actionable as possible and not just a measure of the general tone of discussion.
Measuring the sentiment of crowds online is no simple task with over a billion people who might post information every day. False-positives, false negatives, sarcasm and irony are just a few of the problems that make it difficult. If you’re looking for the overall tone of the messages, then relatively simple measures can be used that count the positive and negative uses of words.
At Visible, we’ve chosen the hard path, but the one that we believe delivers the most value and the most actionable information. For definition, we consider tone to be the measurement of the overall tenor of a piece of content, and sentiment to be the specific measure of the Positives, Negatives, Mixed emotions related to a specific topic of interest.
Our systems first topically decompose a blog post or article so that we understand the subjects of conversation and particularly the ones that our clients are interested in studying. This is an area that is not talked about as often, but just as important to the overall measurement. The capacity to understand sentiment is most critical when you can be sure that the sentiment you are measuring is related to your product/company/issue.
BC
So how does Visible Technologies do this?
DB
Well, once we have broken down the post so that we can understand the multiple areas of discussion, we use a combination of automation and human-based analysis to understand the true sentiment related to that issue, NOT just the overall tone of the message. While a post might be really positive overall, the part where they mention your company or product might in fact be negative, mixed, or just neutral.
It’s also important to understand the sentiments that the post drives. When someone comments, is it positive or negative? When someone from your organization participated in the conversation, did they change the sentiment? Did they riot the emotions of the crowd or soothe them? Did they cause more people to talk or change the subject?
Because truCAST data is built up from a granular crawl of the web, with comments included in the thread of conversation, and with the multiple topics discussed in each post broken down, we can assess the level of sentiment – AND its IMPACT – in unique and powerful ways.
We can also help you understand the sentiment on the topic overall, the authors who are positive and negative on the subject, and the relative authority level of the sites and authors involved – again, on that topic, not just in a broad-reaching assessment of the post overall.
BC
Is their a right way or wrong way to be measuring sentiment?
DB
There are lots of ongoing debates about how much automation and how much human analysis gets applied and they will probably continue for a long time. Our answer is simple, though. We believe that there are advantages in both and we will continue to develop and apply unique and distinctive methodologies that deliver the most consistent and accurate possible data to our clients.
Thank you David for the insight on the topic of Sentiment measurement.
Blake Cahill
Visible Technologies
3 comments ↓
Blake,
I like the value from the human interaction and believe at some point in the process it has to play a role. But how do you get it to scale for large clients or many clients?
Automated sentiment analysis will always - yes, always - be a tricky process. But you’re right that a) it’s a complicated task best dealt with by a combination of constantly improving technology and a human layer of analysis and b) even if it’s not 100% perfect, it still simplifies and hastens the process for marketer’s needs.
Shawn - you bring up the million dollar question: scalability. Although there’s no simple answer, the best way to address the challenge is with better technology. The closer and more accurate the data processing, the less manual labor required. But for now I’m guessing the answer also means: more hands on deck.
Zach
Shawn and Zach - thank you both for the comments. It is very much a challenge for most organizations and why brands with volumes of mentions need to separate noise from insight. I think that some basic understanding of tonality assists some but dissecting and interpreting sentiment in original posts verse comments and ongoing threads requires not just some technology enablement but also human analysis in order to scale. That requires that an enterprise be “listening” and paying attention to it’s customers. This is not easy but most organization could and should tap one of their most valued (depending on company) or under leveraged organizations — the customer service groups to be involved real time. That is really the only way to “scale” or integrate social media and measuring social sentiment into an organization.
Blake
Leave a Comment