Uncovering customer sentiment
Is social media analytics a priority that should not be overlooked?
Analytics can help companies understand public sentiment by extracting opinions or emotions from this enormous ever-growing source of texts. It often uses machine learning, a component of artificial intelligence, to improve the algorithms analyzing what the public thinks about a certain product or brand. However, sentiment analysis is a very complex task for a machine. Online comments ranging from everyday observations to involved discussions usually do not neatly fall into the categories 'positive' and 'negative' or 'like' and 'dislike'. It takes nuances to understand them.
One of the main pitfalls is context: Meaning is often rooted in the context rather than in the words used in a sentence. For example, a word like 'sick' could be used in a traditional or in the slang sense, all depending on context in which it is used. Measuring relative sentiment poses a similar problem; for example, how positive is 'I like it' vs. 'I love it'?
According to an article in the Guardian’s series on Big Data, a possible way to avoid the “context problem” would be to carry out a separate analysis of demographic data from social media content and to “marry” it with the real demographic data a business has about its customers. However, given the fact that a social network like Twitter already presents a skewed social group of young and urban adults, social media analytics might yield less accurate insights than for example more traditional data such as CRM data, web traffic or sales figures.
Despite the power of social media and the promises of sentiment analysis, combining multiple and diverse data streams including social media streams seems to be the best strategy for businesses to gain actionable insights. Or, as a consultancy representative is quoted: "Often the most valuable insight is based on transactional and behavioral data. But social media or sentiment analysis gives you more color to inform your business decisions and actions. If I only had one choice, I would take behavioral data every time, but neither is social media something to be ignored. It can add a richer, more human understanding to flesh out information by numbers."