How to improve customer loyalty with sentiment analysis

Knowing what your customers think and feel about you can be a huge advantage in today’s competitive world. If you’ve been using a tool such as NPS you already know that getting unfiltered feedback from customers is incredibly valuable, and that the information collected can inform business and marketing choices.

Sentiment analysis takes that to the next level, helping you improve your brand experience and increase loyalty with the insights that it provides.


Sally up the Sentiment Analysis

Sentiment Analysis, also known as ‘opinion mining’, gauges the opinions of individuals or groups. Monitoring customer care conversations or social media posts – essentially any information about your brand or product – it can quantify attitudes, opinions and emotions in the language.

At its core it helps brands understand what people think and feel about a particular ‘thing’.

It helps identify

  • Brand reception and popularity
  • New product perception and anticipation
  • Company reputation
  • Flame/rant detection
  • Potential crisis
  • Overall brand health

You can also use it to do a general market analysis or compare yourself to your competitors, as well as identify and take action on a potential crisis that could see something negative go viral.


Not just for social butterflies

Many people make the mistake of thinking that sentiment analysis is only for social media. And while this provides a great source of raw human sentiment, there are many other channels where this data analysis can be applied. This can include:

  • live chat conversation
  • call centre data
  • product reviews
  • forums
  • even emails sent to you by your customers

When Qrious delivered a sentiment analysis project for Spark NZ, we gathered data from emails, surveys, call centre applications (this includes both live via phone and live chat via the website) as well as social media to identify the common themes and issues Spark needed to address to improve their customer experience and satisfaction levels.


Advancements in analysis

Early sentiment analysis tools based their assumptions on basic linguistic indicators – namely whether a word was considered ‘positive’ or ‘negative’ and then provided a sentiment score.

There were a few limitations with this model. Namely, if a comment included both positive and negative words – depending on which there were more of, would determine if the sentiment was positive or negative. Furthermore, this type of model doesn’t take into account irony, sarcasm or overall context. A sarcastic comment with ‘positive’ words would be given a positive score based only on the words – though if read by a real person would be considered negative as they have a better understanding of the nuances in language. 

Through advanced analytics techniques, such as Natural Language Processing (NLP) and Machine Learning (ML), these types of nuances can now be ‘understood’ by automated systems and provide a much more accurate sentiment analysis.


Three vs Six

Advancements haven’t just been made in the understanding of the sentiment behind the words – but also the range of sentiment categorisations. 

The simplest breakdown is Positive, Neutral or Negative sentiment. More advanced analysis breaks it down into universal human emotions: Anger, Fear, Disgust, Joy, Surprise and Sadness (wonder if this is where Facebook found inspiration for their ‘reactions’?)

Of course, the real value lies in taking this type of text analysis even further to identify topic clusters, as well as track sentiment over time.

One of the pieces of work Qrious undertook for Spark NZ was to analyse the live chat data to understand which topics were discussed. When matched with outcomes or next steps, this identified if there were topics better suited to chat than another type of contact, and the customer journeys most likely to result in success. Qrious later overlaid this with data from the call centre to compare the behaviour and sentiment between the two, and discovered that sentiment differed depending on topic, identifying clear opportunities for improvement.


So, you know the sentiment, now what?

Take action. If your customers (and the general population at large) love you, use this to your benefit. In fact, it’s probably already working to your benefit through referrals and (unpaid) influencers singing your praises.

If it’s negative – find out why. Often, this will be included in the feedback provided by customers. They generally want you to know why they’re unhappy so you have the opportunity to fix it and improve. Not all of them are out there to watch the world burn.

You may find that your sentiment score rises or plummets based on a new product or campaign out in market. These provide great learning opportunities for what does (or doesn’t) resonate with your audience and adapt.

Lastly, it’s important to acknowledge the general sentiment – positive or negative – and if there is a single reason why, to acknowledge that too. The worst a brand can do is bury their heads in the sand and keep doing what you’re doing.

Marketing has come leaps and bounds over the last few decades; no longer is it about trying to manipulate people into what they should think about your brand or product – rather it’s about understanding our customer’s needs and wants, and providing them with the best solution.

Learn more about Sentiment Analysis here, or if you'd like to talk to one of our team, get in touch.


Subscribe to the blog