Whitevector: Making the most of sentiment analysis
Whitevector’s latest research focus report looks at positive and negative online discussions and what they mean for brands … surely all publicity is good publicity?
“What do people say about us online? Is it positive or negative?” These are often the first two questions asked by companies about to embark upon social media analyses. Even though it’s natural to be interested in the tone of the discussion, this sort of analysis, known as sentiment analysis, shouldn’t be the first priority, or the work risks getting off on the wrong foot.
In Whitevector’s experience it is far better to start any project by studying the general nature of the discussions to see what their overall tone is and in what way people praise or criticize the product in question. This overview is an important phase in analysing discussion sentiments since consumers tend to talk very differently about different products: brands in some categories get very detailed and passionate praise or criticism, whereas others are mentioned in varying contexts with less detailed feedback.
The general way in which people debate issues online should also be born in mind as discussions typically include roughly as many positive as negative comments about the discussed brand. For example, if the discussion starts with a praising comment about a brand, someone with a contradictory opinion notices the thread and replies. It is rare for discussions to include only positive or only negative comments as people tend to try to add new aspects to any discussed topic. Because of this, discussions inevitably produce negative comments even if people aren’t against the discussed product per se.
After assessing discussions on a general level, it is often useful to include sentiment analysis in any online discussion study. A good way of going about this is to compare the sentiment levels of the studied brands on a monthly basis. One can also compare shares of negative and positive discussions to see if there are any changes in the usual ratio between them. A more detailed drill-down analysis then enables the marketer to understand what has caused the change in the ratio between positive and negative.
Many times companies concentrate their analyses on the negative comments; Why are people complaining about the product or service? However, are all negative comments bad? Is it not the case that all publicity is good publicity? Of course it is impossible to generalise as sometimes a large number of negative comments might affect sales, whereas any occasional negative remark might serve to raise the public’s top-of-mind awareness of the product, and lead to online correspondents posting multiple positive comments.
Thus it follows that the objective in social media marketing shouldn’t necessarily be about decreasing the absolute number of negative comments, but rather in increasing the number of fruitful and meaningful discussions.
A common conclusion from sentiment analysis is the recognition that the product or service is facing some form of crisis. Indeed, if you think of almost any well-known negative piece of news related to a company, Whitevector’s data does show up a related negative discussion peak. However, just analysing the peak doesn’t usually result in any better understanding of the situation as it is often relatively easy to know what the main reaction to any well-known ‘bad’ news will be.
Thus, we believe that the focus shouldn’t be solely on noticing that something is wrong but rather on more thoroughly understanding what aspects of the brand consumers perceive as positive or negative.
Example sentiment analysis – British Airways
In December 2009, British Airways’ cabin crew staff threatened a strike, which (not surprisingly) produced a substantial number of discussions online. Yet, by browsing through the discussions, one can see that the majority of them were either written by people concerned about the possibility of their own booked flights being cancelled, or were speculations as to the impact on BA’s business within business related forums.
Chart 1. Sentiment Trend of British Airways in December ’09. Source: Whitevector
The increase in the total amount of the discussions due to the strike can be seen in the below Monthly Sentiment Comparison. However, by looking at the other months or by excluding the word “strike” from the query, one can see, what other topics create opinionated discussion about the company.
Chart 2. Monthly Sentiment Comparison of British Airways from July to December ’09. Source: Whitevector
Surprisingly, when comparing different airlines and their sentiment shares in December, it turned out that Air France received many more negative comments than British Airways.
If BA or their agency were to repeat this analysis on a monthly basis they would be able to identify and assess any change in attitudes towards their own, and their competitors’ brands.
Chart 3. Sentiment Comparison of different airlines in December ’09. Source: Whitevector