Sentiment analysis – what contact centres should consider

A potentially powerful tool for contact centres looking to make informed changes to their services in order to enhance the customer experience, is sentiment analysis. 

Sentiment analysis provides a means by which customer and agent emotions within interactions can be quantified. This can relate to interactions that take place on the phone or through alternative channels, with the purpose being to uncover processes, behaviours and situations that cause strong levels of sentiment, whether positive or negative. 

Learning more about such processes, behaviours, and situations is crucial for organisations operating contact centres. After all, if companies do not have insights into these areas, they could overlook key aspects of their service that have a real impact on the customer experience and business outcomes. 

Through the use of analytics and large data sources it is possible to search datasets as a way of identifying and inspecting the types of interaction that have the greatest effects on customer sentiment. 

Is the use of technology for detecting sentiment actually necessary? 

This is a very fair question for organisations to consider. It would seem obvious to many observers that agents – particularly those who have higher levels of empathy and experience – should be capable of identifying callers’ emotions. So, would using technology for the purposes of sentiment detection represent a redundant investment of time and money? 

There might have once been an argument for that. However, in today’s era of artificial intelligence (AI)-enabled analytics, it is now possible for the sentiment and emotions of millions of calls to be assessed in light of their ultimate outcome. 

This, in turn, allows for the identification of real-life situations where there is an increased likelihood of a negative outcome, so that steps can be made to help ensure a more positive outcome before it is too late. 

There are, however, certain downsides with language models 

As useful as 2020s language models can be for the identification of ostensibly positive and negative words and phrases, there are still fairly obvious limits to their sophistication. 

Current language models are, for example, less able to identify sarcasm or other non-straightforward forms of communication. Nor are they as capable as human beings when it comes to identifying the actual meaning in a series of comments seemingly alternating between positivity and negativity. If, for instance, a customer said to an agent that “I’m happy that the product has finally arrived, which is good, but not exactly great”, the human agent would be able to pick up the nuances of such a sentence.

There is, however, scope for the further training of sentiment models to make them more effective at noticing changes in tone, volume and speaking rate, as well as instances of the customer and agent talking over each other, and even silences or subaudible noises that can indicate emotion, such as a snort of disgust. 

What about the process of analysing detected sentiment? 

Presuming that efforts have been made to score each of a contact centre’s interactions on a sentiment scale from highly positive to highly negative, it is also possible for nuances within those interactions – for example, the conversation starting positively and ending negatively – to be selected for analysis of the root cause. 

Although sentiment analysis captures and analyses every interaction, it is generally believed to be of greatest use at an aggregated level, rather than as a means of judging specific individuals. Many organisations appreciate the ability that sentiment analysis gives them to identify the processes, interactions, and subject areas that are bringing customers the greatest stress and negativity. Sentiment analysis also allows for trends to be observed over time, so that the organisation can gauge the effectiveness of any business or process improvements it has made. 

With the potential applications of sentiment analysis for contact centres including – but by no means limited to – the discovery and categorisation of the products, processes and topics that most frequently provoke the strongest positive or negative reactions, as well as quality assurance and even the consideration of agent morale and motivation, it can undoubtedly be a very potent tool. 

Would you like to read more about the latest research and statistics that will be of relevance to your efforts to optimise the customer experience your organisation’s contact centres provide? If so, you are welcome to download and peruse our reports for various key national and regional markets here at ContactBabel.