Contact centers and customer experience (CX) leaders have no shortage of data and insights into the consumers’ level of satisfaction with the organization. Consumers are interacting with brands through multiple channels at a growing rate. In fact, results from the Customer Experience Management Benchmark (CXMB) Series 2020 Consumer Edition report showed that over 70% of consumers used multiple channels to resolve a single customer-care issue. This level of interaction across multiple channels provides a big window into consumers’ needs, expectations, and feelings. The challenge is how to collate and harness that data to improve the entire customer journey, from awareness to purchase to support. With the continued evolution and capabilities of natural language Processing (NLP) and machine learning, customer sentiment analysis is increasingly becoming an important tool for organizations to be able to leverage this data in a meaningful way. There are many facets to sentiment analysis, so in this article we will explore what it is, why it is growing in importance, practical applications, specific uses in customer support, and common challenges.
What is sentiment analysis?
Sentiment analysis is a set of tools that uses machine learning and NLP to extract opinions to provide a measure of sentiment (or opinion) of voice, text, and images. Algorithms work behind the scenes to find specific nuances of text to identify consumers’ positive, negative, or neutral attitudes toward a product, brand, or service.
Why is sentiment analysis growing in use?
It is no secret that CX is a competitive differentiator for organizations, and consumers are not shy about voicing their negative opinions about their experiences in very public ways. Eighty-three percent of respondents from the CXMB Series 2020 Consumer Edition indicated as much. When this happens, the domino effect that can occur can be devastating to a brand’s image. Not to mention, consumer needs, expectations, and feelings change rapidly. So, with multiple channels and the sheer volume of data that is available, it is near impossible to effectively mine the data or rely solely on customer surveys or monitoring customer interactions. Customer feedback measurements such as customer satisfaction, NPS, and customer effort are all important measures of satisfaction and loyalty, but the ability to proactively collect and act on nuanced, real-time data will not only improve customer loyalty and retention, but lower costs as well. In short, it can be a game changer.
What are the practical applications of sentiment analysis?
There are several use cases for sentiment analysis. These are some of the most common:
- Brand monitoring / reputation management – This is one of the most common uses of sentiment analysis, allowing organizations to quickly identify negative or harmful comments about their brand online and react quickly before they go viral. Conversely, it can also be used in the event of ‘positive’ mentions – brands can quickly jump in on these conversations and maximize the positive exposure in online forums.
- Marketing and market research – Sentiment analysis helps brands measure performance of campaigns and marketing activities while also providing insights on how marketing messages are perceived. This allows brands to pivot strategies if needed and better personalize messaging for more relevant content. From a market research perspective, sentiment analysis is not the most common tool utilized, but it can be used to gain perspectives from customers AND competitors.
- Product analytics – Using sentiment analysis in the context of product analytics is like brand monitoring. Instead of focusing on brand mentions, it mines comments about specific products. Customer feedback can be categorized for further improvements, which is particularly important in the early stages of product development. As products mature, this data is often combined with brand monitoring in a more multidimensional view. All of this allows brands to understand how products are perceived by various target audiences, their performance in the market, strengths that should be maximized, and areas that need improvement to increase user reception.
- Customer support – Below we will review, in more detail, the specific use cases of sentiment analysis in customer-support operations.
How is sentiment analysis leveraged in customer support?
Sentiment analysis, or text analytics, is not necessarily a new concept for customer-support organizations. However, it can seem a bit of a mystery as a concept, but it has some very real and practical applications.
- Route consumers to the ‘right’ resolution channel – This is perhaps the most common use of sentiment analysis in customer support. Identifying the emotion or tone of the consumer upon contact with customer support allows quick routing of that consumer to a higher place in the queue, or to specialized agents trained and equipped to handle difficult interactions. Similarly, if consumers are interacting with chatbots, upon recognizing a consumer is dissatisfied, the consumer can be routed to a live agent better trained to resolve the issue.
- Identify friction points – By monitoring and collecting data in real-time, in all channels, and on all interactions, the data can be used to identify pain points in the consumer’s journey. Organizations can then utilize this information to remove those barriers and improve the overall experience.
- Uncover agent challenges – Using sentiment analysis to understand how consumers feel when interacting with customer-support agents allows leaders to identify which agents are most successful and those who struggle. It can also uncover gaps in knowledge so training can be modified or created to better equip agents with the tools they need.
Having a pulse on customers’ feelings has never been more important than it is today. The events of 2020 have understandably introduced high levels of stress and strong emotions. So, the ability to recognize negative emotions early in the resolution process and proactively address those concerns before they spiral beyond control can provide a competitive advantage if implemented effectively.
What are the challenges with sentiment analysis tools?
As illustrated above, sentiment analysis is a valuable tool for many applications, including customer support. That said, there are some inherent challenges that should be understood when exploring the use of sentiment analysis. These include:
- Detecting tone and subjectivity – Detecting tone is generally one of the primary features of these tools, and typically fairly easy for them to determine based on the verbiage, keying off certain words such as “nice” or “horrible.” It becomes more difficult to determine whether the message is objective or subjective.
- Identifying irony and sarcasm – This is reportedly the most difficult of issues for these tools to navigate. In reading a text string at face value, it could easily be mischaracterized as either positive or negative. For example, a customer sarcastically types in a chat, “That’s just GREAT customer service, isn’t it?” A tool might misinterpret that as a positive sentiment, when in fact the customer was being sarcastic. This requires constant training of tools to deliver more accurate results.
- Understanding context – This too can create a real issue for these tools. A human can often quickly understand context. However, an algorithm cannot ‘guess’ what they need to do – they are configured to get to the right answer based on a set of rules. So, the model must include an additional component to get to the heart of the context.
- Changing vocabulary and word ambiguity – Vocabulary is constantly changing, slang is introduced, and even existing words take on new meanings virtually overnight. As a result, again, researchers and systems engineers must continuously work to train these tools for these changes.
Sentiment analysis can be incredibly valuable to organizations, though as with any technology, it is important to understand its capabilities. Its specific application can differ from organization to organization so those who implement and manage this technology must know how to use it, and where to apply it for maximum benefit. Because it can be leveraged by many parts of an organization, its usefulness is maximized if internal departments work together to agree on a common goal, the areas it will be applied, who will mine the data, and how it will be applied.
If you are a senior leader running CX operations on the corporate side and would like to engage your peers about Customer Sentiment Analysis and many other topics in a private, online community, we invite you to join our KIA online community. You can also find out about everything happening within the Execs In The Know community by joining our mailing list.
Blog post, written by: Execs In The Know