Tracking Brands: Emotions will tell us more than sentiments ever could about how consumers perceive brands (case in point Volkswagen)

If you were responsible for spearheading the communications mandate for Volkswagen after the dieselgate scandal then you’d be wondering some of these questions,

• How does social media sentiment change as a consequence of a public relations crisis?
• How does the public react to recovery efforts initiated by the company?
• How do topics of conversation shift as a consequence of a brand scandal and subsequent recovery efforts?

And some of the input that your marketing teams will get from social media could be comments (sample only) of this nature.

“how could VW lie for years and not care about the environment, this is horrendous!” — Owner of a Passat since 4 years.

“It’s just so sad to know that unknowingly there was so much harm done to the environment.” — Owner of a Polo since 8 years.

“It is nice to see that they’re apologising now and owning up to their wrongdoing” — a potential customer.

Imagine the task is to create the first ad campaign for Volkswagen after the dieselgate scandal, what kind of analysis of the above listed comments be of more use to you.

Sentiment Analysis: which will tag two comments as negative and one as positive.

Or

Emotion Perception Analysis: which will tell you that customers are angry; sad and guilty; and accepting of the apologies.

No points for guessing it’s the second, emotion perception analysis. So why does the industry just stick to sentiment analysis, you may ask, well the answer is, it’s only a matter of time. All marketing communications teams would prefer emotions over sentiments but as of now it is more time consuming and has more manual work involved. But with natural language processing systems gaining pace it is going to be a reality soon.

Most people express their feelings by talking about their favourite aspects of a product/service. These emotions do not get captured by the binary approach of sentiments (positive or negative). In many cases a hatred towards an aspect of the product could express extensive use of the product. Which would typically be tagged as negative in the sentiment based approach.

There has to be another way to glean the finer nuances of these emotions and their intensity by analysing both the conversation and the reason for which it would have been mentioned. In fact, its not very often that people mention an outright rejection or acceptance. Human emotion is usually complex and needs to be treated as such.

The task ahead is to just train machines to learn our analysis and repeat it over bigger data sets. Just like we trained machines to analyse sentiment, it is time for us to train the same machines to analyse emotions.

Picture this, “Arrghhh! How can <insert brand name> do this? Why can’t I get a refund?”. As a human we can understand the context and conclude that the customer is angry because of some issue with the product/service. However, according to current systems this is probably a neutral sentiment tagged comment. Which is not enough since it does not grab the actual emotion behind the comment.

We have to do better. The only way this is possible is to start feeding the system with human analysis of the comments and overtime the system would know how to tag the comments for emotions too.