Customer feedback is central to understanding and improving the customer experience. That feedback is only useful, however, if brands understand the meaning behind it.
Customer sentiment is the feeling and emotion that customers have about a brand. Understanding customer sentiment is crucial to delivering a positive customer experience.
Customer sentiment isn't just a list of keywords; it's a complex set of subjective targets businesses can find on social media, from user behavior, and -- in the case of a service desk -- from the tone of voice when a customer or user calls. When a customer buys something, posts a review online, chats about a brand with others on social media, clicks on ads, phones a call center or takes any other action that brings him in contact with a brand, there is an emotion involved.
Sentiment isn't always positive, of course; negative sentiment is out there, too, but it also can be useful information, offering improvement cues. For example, "I think the new website design really works" is clearly positive, while "The new website design doesn't work for me" is only slightly different but is clearly negative. It's useful feedback either way.
This article is part of
Knowing how customers feel about a brand, for better or worse, is invaluable information for businesses. Here are some ways to gather and act on that information.
How to gather sentiment
The first step for gathering sentiment is knowing where and how to find it. There are several ways to do so.
Word-of-mouth. A customer's direct emotional expression of satisfaction with the purchase of a product or service is the most valuable sentiment data available to businesses. Word-of-mouth is the most trusted expression a customer can make because it's self-motivated and authentic. Companies can detect this kind of sentiment with the simple addition of a referral question to a customer survey: "How did you hear about [the product]?"
Social media. By far the most complex -- and potentially rewarding -- data collection process for customer sentiment is gathering data from social media. Actual conversations between customers occurring on the major social media channels, such as Facebook, Twitter, Instagram, etc., offer a fly-on-the-wall perspective, not only on which products and services are of interest to customers, but also the specifics of what they do or don't like.
To gather this data, major CRM platforms bundle social media listening software into their systems. Standalone products are available as well. Software can monitor all the major social media channels for brand mentions, and most of the important analytics are also built in. Parsing this data can be as simple or as complex as the tools allow, answering questions from "Is the product launch going well?" to "What don't they like about the interface?"
Likes. Less revealing than social media mentions and discussions are prompts within apps that present products to customers and ask for binary feedback, such as do you like/love this? Despite being a thin measure, it's an easily managed one, and it results in a simple ratio.
Reviews. Reviews are likes on steroids. They don't just reflect a thumbs-up for the brand; they provide the why of the thumbs-up -- or, conversely, the thumbs-down. This is invaluable detail, and increasingly easy to harvest; consumers now routinely turn to review sites such as Yelp, Angie's List and Google+ to see what others think. And even if you aren't able to work with unstructured data, star ratings are more useful than a Like count: "The new product is getting an average 4.2 rating" is worth more than "Sixty-two percent of the people who visited the page Liked the product."
Support/service desk. Most negative brand feedback comes from service calls. Customer difficulties and dissatisfaction flood this point of contact day in and day out. The problem is that data collection here is generally very poor, with the only actionable analytics traditionally being the, "How satisfied were you?" question at the end of an interaction.
But there's more data of value in this exchange. How intense was the customer's dissatisfaction? What was the customer's tone of voice? What is the customer's history of calls and complaints? All of this has great value, especially when it is aggregated with data from other customers.
Chatbots. Chatbots are an enterprise Swiss Army knife embedded in apps and webpages for a range of purposes. It's a simple matter of adding one of the above measures to a chatbot's script as appropriate. This can be simple -- "Were you satisfied with the answer you received?" -- or more complex -- "Tell us what you didn't like about the service."
Indirect sentiment. It's important to make a distinction between direct versus indirect sentiment. The examples above generally characterize the former -- the customer saying something explicitly about the product/brand. But there's also indirect sentiment -- drop-off in site visits, reduced browse time, abandoned carts and chatbot sessions -- all of which are also measurable and valuable.
Tools to measure sentiment
Would you recommend ...? Surveys mailed to homes and restaurant comment cards often include this question -- and it's trustworthy because it's a direct report communication.
Customer Satisfaction Score (CSAT). One might expect some guesswork when measuring customer satisfaction, especially given the multichannel nature of that data. There are standards, however, and the CSAT is one prominent choice.
The CSAT is a Likert scale measure -- a 0-10 rating, or something similar -- of, "How satisfied were you with your experience?" This score gauges customer receptivity to changes to a product, the effectiveness of a service, and the relative success of a campaign or any other event where customer response matters. The score is simple to calculate and businesses can add it to any CRM analytical process.
The value of the CSAT is not so much in any individual response, but instead, the application of this measure throughout the customer journey, including during product presentation, purchase, upgrade and service.
Net Promoter Score (NPS). The likelihood of recommendation is a strong measure of sentiment. The trick is to reduce that emotion into data that is consistent for analytics.
NPS is the standard for taking this measure. It's frequently a back-end measure of customer loyalty taken as part of survey processes, recommendation requests, customer interviews and customer service desk follow-ups. Like the CSAT, it's a simple calculation that's easy to incorporate into any customer data collection process.
The NPS is generally based on a Likert scale broken into three categories:
- Promoters -- very satisfied customers who are brand-loyal.
- Passives -- customers who may be loyal but open to the competition.
- Detractors -- unhappy customers who prefer other brands.
The NPS is calculated by taking the percentage of promoters and subtracting the percentage of detractors.
Mystery shopping surveys. Businesses may choose to create an in-house program or hire an outside agency to conduct mystery shops -- under-the-radar audits -- on products and services. Surveys include a number of multiple-choice questions using the Likert scale and written comments.
Many mystery shopping surveys also include an NPS rating. However unconventional this technique might seem, it leads the pack in terms of the resulting data quality. Information gathered by mystery shoppers has high authenticity, broad utility and is very useful for generating unanticipated insights.
Customer sentiment analysis tools. Companies that are serious about sentiment analysis should obtain a customer sentiment analysis tool suite. Tools such as Lexalytics, MeaningCloud, MonkeyLearn and others bring natural language processing and sophisticated data science together under one roof, going far beyond keywords and brand mentions to extract mood, intentions and other subtleties.
Sentiment analysis tools work by scanning a body of text for emotional expression, and then scoring the polarity of that expression: Did it lean positive? Negative? Neutral? These tools can go beyond scoring negative versus positive sentiment by detecting patterns, and sometimes pinpointing specific emotions.
At its most advanced, customer sentiment analysis can be aspect-based; it can break text down to the level of identifying aspects of the product or service being referenced, scoring each one. For example, "I love the product," followed by "The instructions that came with it are really complicated" is a positive and a negative in a single review, but both pieces of information are valuable.
Act on sentiment
All of this data plays a role in informing CRM decisions as well as product development and service policy. But it's not enough to have the numbers and see how they trend; the data itself will raise important questions, and, if possible, businesses should act on this data.
When a customer's rating or answer changes, it's important for businesses to understand why. What is the timing? Is there an upgrade or new release of a popular product or service? Or perhaps a competitor launched some new features that are drawing customers away?
A customer's positive rating going negative affects the overall numbers of a product's popularity, but it's also a cue to reach out to the customer for more data to address any problems and determine what the business can do differently.
It's also important, in using sentiment as the basis for action, to view it as a whole. While businesses should constantly monitor for important cues, they should also keep an analytical eye on the general trend of sentiment where various aspects of the brand are concerned.
Monitoring whether the same customer praises a brand repeatedly in recommendations or on social media is also important, as this person may be a potential brand ambassador -- a customer who can be incentivized to promote the brand to others.
Customer sentiment, when properly studied, can show the why of underlying customer behavior. This behavioral cue can signal to an organization what to do next and how to keep the relationship fresh.