Three things you need to know, if you are using Twitter to study consumer emotions

With 313m monthly active users around the world, of which 82% access Twitter via their mobile handsets, Twitter is likely to be a great source of insight into what customers are doing, paying attention to, or talking about. As Pratik Thakar, Coca-Cola’s head of creative content for Asia-Pacific, said, it is like a big focus group.

 

 

In particular, Twitter conversations can be a great source of insight into customers’ opinions about the brand. However, if you are going to use Twitter to study consumer emotions you need to take these three points into account.

 

1 Consumers praise brands in private, and criticise them in public

Social media users use different social media platforms for different purposes – for instance, in terms of photo sharing, Snapchat is for conversations, Instagram for capturing special moments, and Facebook for events.

 

Likewise, according to research done by Jan Kietzmann and myself, customers tend to talk about their positive experiences on Facebook, and about their negative ones on Twitter. As a result, while Twitter may give you a view into the sources of customer dissatisfaction, which is really important and should not be ignored, it will not give you the whole view about how the customer feels about your product or customer service experiences.

 

2 Consumers complain more about big brands

Research by Andrew Rogers and colleagues based at Cardiff University, reveals that Twitter users not only talk about big brands disproportionally* more than about smaller brands, but they are also more negative about those big brands. The authors call this the ‘negative double jeopardy’ effect:

“Larger share brands attract a larger percentage share of tweets which have a more negative sentiment amongst the larger number of tweets.” (page 21)

 

Therefore, if using Twitter to investigate whether consumers are happier with brand A or brand B, we need to take into account that the volume and sentiment of conversations about a brand is skewed by the relative size of that brand.

 

* I.e., the relative percentage of tweets about big brands is higher than their market share, whereas the percentage of tweets about smaller brands is smaller than their market share.

 

3 Automated analysis of tweets may underestimate the strength of negative sentiment

Conversations on Twitter are, often, laden with humour and irony, as illustrated by these exchanges about the Samsung’s product recall of the Galaxy Note 7 phone, due to safety issues.

The problem is that it is really difficult to detect (and correct) for irony, when using sentiment analysis software to code Twitter posts. This is not a criticism of the software but, rather, a fact; and it occurs because tweets are very short, because users may express themselves with emoticons, emojis and abbreviations that can not be processed by the software, and because there are significant contextual variants to how people communicate their emotions publicly.

 

This means that the software may classify as neutral or positive, messages that are actually negative and, thus, give you an incorrect view of what Twitter users are saying about the brand. This post discusses some ways to remedy this problem.