In the past few posts we have highlighted the power of influencer marketing in this new social media age and how you can measure its effectiveness. When negotiating with agencies specialised in social, you’ll undoubtedly hear terms such as reach, engagement or sentiment analysis – all three holy grails to qualm nerves and measure branding success (if you’re unfamiliar with these, check out our post below!).
The latter, sentiment analysis, is a buzzword that has taken the online marketing scene by storm. Here we break it down for you by explaining what is meant by sentiment analysis, why it’s important, how it can be measured, and how we at Socially Powerful use it to inform our campaigns.
What is sentiment analysis?
Social media platforms have drastically changed the relationship between producer and consumer. The top-down one-way channel of communication common to traditional media has been torn down by UGC, opening up dynamic spaces for consumers to collectively and individually voice their opinions on brands. This, in turn, has made it far easier to understand how your target audience feel about your product and/or your marketing strategy, whether that be through the like/dislike ratio, influencer story polls or (and this is the most fruitful) the comment section. While engagement is a useful metric to gauge a post’s relative popularity or the amount of interest it peaked, sentiment analysis allows you to further refine and optimise your content strategies to effectively maximise ROI.
So why doesn’t everyone carry out sentiment analysis?
They do. Sentiment analysis is featured as a metric on many social media insight tools used by agencies – hence its buzz on the marketing scene. However, often these metrics are used without fully understanding how they work. Some rely on the like/dislike ratio mentioned above. Other more specialised analyses will examine the language used in the comment section, and here is where the problems arise.
Sentiment and opinions are highly subjective and open to interpretation. As such, the grammatical and syntactical conventions used to express positive or negative emotions are hard to generalise with precision. To circumvent this, some tools such as LIWC use sentiment lexica, i.e. list of words organised by their bipolar semantic orientation (positive/negative). However, this offers only a crude interpretation of language, which ignores the intensity of a certain sentiment or the contextuality in which words are used – a feature particularly crucial as words often have multiple meanings. Even tools that incorporate valence scores for intensity (e.g. VADER) ignore the lexical features native and ubiquitous in UGC like acronyms, emojis and slang.
Other more recent attempts at sentiment analysis (e.g. Naïve Bayers classifier, Support Vector Machines, etc.) have made use of growing expertise in machine learning and natural language processing to learn and identify sentiment-relevant features of text. However, the issue with such tools and UGC is that they require large sets of validated training data which represents as many of the lexical features as possible. Such data sets of UGC are hard to acquire due to the spare and short nature of text on social media.
How then does Socially Powerful analyse sentiment?
Here at Socially Powerful we understand why sentiment analysis is hard and we, therefore, like to do everything in-house to ensure the highest degree of quality and certainty for our clients. We carry out our own comprehensive sentiment analysis, integrating easily identifiable metrics such as like/dislikes and influencer polls, validated sentiment analysis tools and analyses of multiple samples in comment sections, carried out by different expert analysts. This way ensures we cover the drawbacks of each method. It also means we can be more creative and offer a more fine-grained bespoke analysis for each piece of content.
Why is this important?
Thinking back to last year’s Pepsi-Kendall Jenner advert provides a perfect example of the importance of sentiment analysis. Viewing it on the basis of engagement, the advert was a huge success. However, as everyone knows by now, it drew widespread criticism from around the globe for its insensitive and farcical content.
In influencer marketing, for example, an influencer may post to Instagram holding the product in hand so that it gains exposure to their followers. Any of the sentiment analysis tools mentioned above will then analyse the language in the comment section to get a rough picture of how it has been received. However, going the extra step allows us to fully understand whether the positive or negative sentiment recorded is actually directed towards the product or just other features of the post (i.e. outifts, quality of photo, background, etc.) – if the latter is the case then it is classified as neutral. In other words, it allows us to fully understand the context in which views are expressed, because ultimately it is that context that shapes our opinions.