Home > blog > Sentiment Analysis in banking

The banking sector has undergone a major revolution with the advent of digital transformation. The entry of Fintech and tech giants such as Google, Amazon, and Facebook have introduced convenient banking that is easy to understand and use. In this competitive environment, banks are realizing the importance of customer service and satisfaction and want to pay close attention to the Voice of Customer to improve the customer experience. By analyzing and getting insights from customer feedback, banks will have better information to make strategic decisions. In their quest to better understand their customers, banks are seeking artificial intelligence (AI) solutions in the form the of sentiment analysis.

What is sentiment analysis? In simple words, sentiment analysis is the process of detecting a customer’s reaction to a product, brand, situation or event through texts, posts, reviews, and other digital content. Using sentiment analysis, business leaders can gain deep insight into how their customers think and feel. The analysis can help in tracking customer opinions over a period of time, determine customer segmentation, plan product improvements, prioritize customer service issues, and many more business use cases.

Sentiment analysis, also called opinion minion or emotional AI, is a series of algorithms based on natural language processing (NLP), text analysis, and computational linguistics. The algorithm is designed to identify the types of comments or reviews (positive, neutral, or negative) based on the words used by the customers. NLP, often confused with text mining, is an advanced analysis technique used to filter large amounts of research and extract relevant information. NLP forms an integral part of text mining but uses a variety of techniques to understand the complexities and sentiments of human speech and natural text. Today, the use of NLP is widespread, hidden behind chatbots, virtual assistants, online translation services, and much more.

The NLP algorithm determines a customer’s sentiment using either a dictionary (Lexicon-based wherein words are annotated by polarity score), machine learning (constructing a classifier to identify specific text) or hybrid (combination of machine learning and lexicon) method. Usually, sentiment analysis tasks are modelled as a classification problem, i.e., classifying a text to a class. Here, two problems must be resolved:

  • Subjectivity classification – subjective or objective classification of a sentence
  • Polarity classification – a positive, negative, mixed or neutral opinion of a subject or object

For example, a sentence like “The customer service of XYZ bank is frustrating” – the system identifies “customer service” as a feature, “XYZ bank” as the object, and “frustrating” as a negative opinion. The algorithm arrives at a relationship between the opinion and object to extract relevant information.

Today, several banks study and track customer behavior through websites, transactions, voice notes, social media, and other digital channels. The aim being, to map and monitor a customer’s journey with a bank and how those paths affect the quality of service or the sale of financial products and services. Financial institutions are collecting data through polls or interviews to capture customers opinions towards specific product or service. Analyzing the unstructured data through semantic processing offers a comprehensive view of customer satisfaction; classifying it under negative, neutral and positive feedback. Using the insights, banks can deliver better customer service by:

Personalizing customer engagement

Keeping a record of customer sentiments would help guide customer service teams to engage with their customers better and deliver personalized experiences. Social media listening tools can be deployed to understand customer behaviour and interaction and arrive at a data-driven marketing strategy. For example, Amex’s Go Social program delivers insights to merchants to create social and mobile offers for their customers.

Prioritize customer issues

Customer support systems can use sentiment analysis to categorize customer support tickets or comments based on the criticality of the issue. The automatic analysis of sentiment of the text can then prioritize the issues, helping customer support teams focus their effort and time on highly critical issues first.

Improving banking products and service offerings

Social media monitoring is helping financial institutions gain a comprehensive understanding of how customers react to their offerings. For example, BBVA Compass analyzed social media comments to improve its rewards system. Through the insights, BBVA was able to identify trends, capture competitor product benefits, and understand how social media users comment on the bank. The result – BBVA raised the cash back rewards on its credit cards.

While sentiment analysis is used across industries, it comes with its own set of data challenges that can be classified under – volume, language ambiguity, and text size. The new-age, tech-savvy customer is generating a huge amount of unstructured data. Combing through this mountain of emails, texts, support tickets, chats, etc. is a difficult task that is time-consuming and expensive. Using machine learning, banks are able to reduce the computational burden of analyzing text, but with text comes the challenges of general sentiment analysis issues (irony and sarcasm, word ambiguity, negation detection, neologisms, idioms, and multi-polarity). In addition, NLP algorithms are generally written to analyze large body of texts that offer context and more information. In  the case of short texts (reviews/comments), the syntax  and context of the written language is lost and cannot always apply to traditional NLP techniques. Incorrect segmentation of text leads to incorrect semantic similarity and increasing ambiguity of data.

NLP has been successful in handling the syntax of a natural language, but the technology is far away from meeting the challenges of semantics of natural languages. Delivering exceptional customer service does not end with sentiment analysis alone. As the volume, diversity, and complexity of customer feedback data grow in the banking sector, there are further challenges to be addressed with sentiment analysis. With the evolution of technology, semantic analysis will grow to become an integral part of customer service.

Customer Experience is the new battlefield.  The harsh reality remains that a bad experiences can be shared with millions of people around the world within a matter of seconds. A good reputation takes years to build and seconds to destroy, hence managing risks with a technology that flags high risk social posts (reviews, mentions, tweets and blog) in real time is key. Listening to customers, and moving with them, can be a game-changer for the Banks.

Article by

Maveric Systems