How Sentiment Analysis Can Improve Your Sales

What is sentiment analysis?

Sentiment analysis is the scanning of words written or spoken by a person to determine the emotions they’re most likely feeling at the time. If the person spoke verbally, sentiment analysis technology can analyze a transcription of the conversation for that purpose. The results of the analysis give businesses a better read on their customers.

Companies can use sentiment analysis to analyze direct communications, such as conversations and interactions between you and your clients via email, phone, WhatsApp, chatbots and other channels. They can also analyze online communications, such as comments made by consumers on social media, in blog posts, in news articles and on online review sites.

How does sentiment analysis work?

Sentiment analysis examines text mined from a wide variety of sources, including online forums, social media platforms ─ such as Twitter, Facebook and LinkedIn ─ chatbot conversations, support tickets, blog posts, emails and third-party websites. 

Artificial intelligence (AI) and machine learning (ML) run natural language processing algorithms to analyze the text. Sentiment analysis software attempts to understand the emotional content of the text from a human point of view. [Related article: What’s the Difference Between Machine Learning and Automation?]

What are the different types of sentiment analysis?

There are five main types of sentiment analysis:

• Graded analysis: This is one of the simplest forms of sentiment analysis. An example would be people scoring a business out of 5, such as rating a business on Yelp. Sometimes, numbers are replaced by choices like “excellent,” “satisfactory” or “below average.”
• Emotion-detection analysis: Analytics tools assign feelings like sadness, anger, frustration and happiness by matching text to a list of words tagged with one of these emotions. While this works well a lot of the time, some technology can be confused by colloquialisms like “bad” or “wicked,” which can also be complimentary in the right context.
• Fine-grained analysis: Here, sentences are broken into their constituent parts and analyzed in more detail. For example, in the sentence, “The wipers on my car snapped off after three years,” fine-grain analysis determines the object (“the car”), the feature of the object (“the wipers”), what went wrong (they “snapped off”) and when (“after three years”). It can determine comparatives like “x is better than y” and it can assess sentiment on a given subject ranging from “very positive” to “very negative.” Fine-grained sentiment analysis is used most frequently to gauge opinions on social media, particularly in cases of crisis management. 
• Aspect-based analysis: Like fine-grained analysis, this method looks for positive or negative sentiment based on input. An example of this would be a person writing to a chatbot, “The wipers on my car snapped off after three years.” The chatbot would recognize that the customer was in need of help and then transfer the conversation to a human operator for assistance. [Learn more about responding to live chats.]
• Intent analysis: This type determines whether a statement is a question, show of appreciation, complaint, suggestion, opinion, marketing collateral or news. A good example of intent analysis is how Gmail sorts incoming messages as “Social,” “Promotions,” “Updates,” and “Forums,” although Google uses other techniques in addition to intent analysis to achieve this.

How to conduct a sentiment analysis

Sentiment analysis can be a vital tool for businesses. However, sentiment analysis is a multi-step process. To conduct the analysis correctly, businesses need to take the correct steps in the right order. 

Define your objectives

Prior to starting a sentiment analysis, businesses need to have a clear idea of what they are measuring. Is it a particular campaign? A product launch? Is the business more interested in how potential clients view the business overall? 

Next, businesses need to determine where they want to conduct the analysis. Are businesses sorting through social media posts? Evaluating customer reviews? These questions will guide what type of sentiment analysis the business will need to conduct. 

Collect the data

Once the objectives are clear, businesses need to collect the data they intend to analyze. This approach changes based on objectives, but may include actions like:

• Sorting and collating documents or response forms
• Conducting interviews
• Scraping web data
• Building and leveraging application programming interfaces (APIs)

Prepare the data

Raw data can be difficult or impossible for sentiment analysis tools to parse. All collected raw data needs to be prepared and cleaned prior to analysis. Depending on how the data is collected, this step could include: 

• Removing noise and irrelevant information 
• Transcribing audio, such as from in-person interview questions, into text 
• Removing URL tags, emojis or Unicode characters 
• Normalize text by converting it to lowercase 

Choose the analysis method

Once the data is ready, choose how you would like to analyze the data. For simple survey results, a graded analysis could be sufficient. But if a business is looking for more in-depth analysis from sources like social media posts or comments, a more thorough method should be chosen. 

Businesses should also decide here if they want to do a rules-based or an ML approach to conducting the analysis. Rules-based approaches use predefined rules and procedures for detecting sentiment while an ML approach uses a learning model trained on a particular labeled dataset. 

Whichever analysis method and approach is chosen should also relate back to the overall objectives determined in the first step of conducting this analysis. 

Once a method is chosen, proceed to analyze the data. 

Evaluate the results

Once the analysis is complete, it should be evaluated to ensure that it matches the business needs and assesses sentiment accurately. Some of the tests may be overly technical, such as measuring an AI-powered tool’s F1 score ─ a measure of the model’s precision and recall. However, if a business is using a third-party sentiment analysis tool, concerns over model accuracy should be flagged to the model developer. 

Results should also be judged based on human evaluation, to see if the overall sentiment analysis makes sense according to basic human judgment. If not, there may be an issue with how the data was collected or processed. 

Interpret and share the results

Once the results have been evaluated and approved, share them with decision-makers within your organization. The results may contain valuable insights regarding how customers view the brand, recent campaigns, marketing decisions and more. All of this would prove extremely valuable for different departments within the business. Depending on the results, it may lead to shifts in business practice, marketing materials, product changes and more. 

How can you use sentiment analysis to improve sales?

Businesses can use the results of sentiment analysis to shape their sales and marketing plans, evaluate social media posts, improve crisis management and brand strength and translate digital public relations (PR) into tangible actions. Understanding your clients’ emotions and expectations can be the key to keeping customers. 

Sales and marketing

Businesses can use sentiment analysis to see how well their marketing campaigns are going on social media and third-party websites. With brand-new product launches, they can scan online comments to see if any customers are having issues. Companies can also get a sense of how well their target audience has received their new product. Based on the results of the analysis, they can adjust their sales and marketing plans to feed into or address consumer sentiment.

Social media research

Traditional social media monitoring often focuses on measuring the number of likes, comments and shares a post gets. While these numbers might indicate buzz around a company, they don’t give emotional insights into consumers’ likes, dislikes and expectations.

In contrast, you can use sentiment analysis to “understand whether consumers feel ‘positive,’ ‘negative’ or ‘neutral’ about a certain brand, product or topic,” said Maxime-Samuel Nie-Rouquette, former head of partnerships at EverIT and former account manager at Tempo Software. 

Sentiment analysis offers companies the opportunity to find more meaning in social media data, said Sean MacPhedran, senior director of innovation at marketing agency SCS. “The most straightforward use for sentiment analysis tools for marketers is the measurement of trends in general sentiment on social media ─ for example, tracking Macy’s mentions and looking at the words around it for emotion and modifiers. Emotional words are fairly intuitive for us to grasp. ‘Crappy’ or ‘hate’ are bad. ‘Awesome’ and ‘great’ are good.”

MacPhedran recommends diving deeper to determine any nuances in the sentiments expressed. “For example, is there a specific location associated with clusters of negative sentiment? Is there a specific issue that is associated? ‘Returns,’ for example, might indicate people are generally unhappy with a returns policy.” 

Crisis management and brand health

Crisis management is how companies attempt to seize the narrative and minimize damage following an emergency. In a crisis, it’s crucial businesses use sentiment analysis to find out how their brand’s supporters and detractors are reacting to the situation. They can also conduct analyses at regular intervals after the crisis passes to determine whether consumers have moved on from the incident.

For example, in 2019, Gillette experienced a PR disaster with its “The Best Men Can Be” video campaign, which addressed toxic masculinity, sexual harassment and bullying. The video got 1.5 million dislikes on YouTube and the company saw its YouGov BrandIndex buzz score drop by more than five points, plunging it into a negative rating. But the bad buzz eventually died down and a few months later, sentiment analysis of the follow-up campaign “#MyBestSelf,” featuring a transgender man being taught to shave for the first time by his father, indicated very positive consumer reactions.

In this case, Gillette recognized consumer sentiment to its maligned “The Best Men Can Be” campaign and was able to restrengthen the company’s brand health by adjusting its marketing content.

Digital PR

The goal of digital PR is to create a constant buzz about a particular brand and its products or services. You can measure the volume of content and consumer sentiment toward your brand and the stories people are talking about with sentiment analysis.

“By listening to conversations being held online, a company can understand consumer emotions and give them a connection that goes well beyond whether a product simply sells well or not,” said Nie-Rouquette, who offered examples for the retail sector. 

“Retailers can monitor their customers’ reactions and feedback to push content for ‘virality’ or exercise a damage control strategy during crisis management. Retailers such as Walmart, Target and Costco use sentiment analysis to understand what their customers care about and leverage that information to reposition their products, create new content or provide new products and/or services.”

Benefits of using sentiment analysis

A well-conducted sentiment analysis can pay dividends for a business. Some of the many benefits include several factors. 

Improved customer insights

One of the main draws of sentiment analysis is to better connect with customers to see how they view the business. Sentiment analysis can help with: 

• Understanding customer preferences through evaluating reviews, social media posts and online comments. This can help businesses understand what customers like or dislike about their offerings, marketing strategies, communication styles and more. 
• Identifying pain points customers may experience interacting with the brand, such as customers having poor experiences with a business’ web layout.
• Boosting customer satisfaction by using insights to improve areas of complaint and to focus on areas of the business that customers like. 
• Improving customer loyalty by tracking and responding to broad consumer sentiment. 

Better market research

Sentiment analysis can also help a business measure how it stands in the eyes of current and prospective customers compared to other businesses: 

• Competitive analysis helps businesses see how they are being discussed relative to other businesses in the field. 
• Trend analysis helps businesses see any changes in public perception or notable shifts in consumer sentiment. 
• Social media engagement tracking can help businesses identify what type of posts resonate with potential customers. It can also help identify reactions, positive or negative, to new marketing or brand strategies. 

Enhanced brand management

Related directly to market research is sentiment analysis’ help in brand management: 

• Reputation management can help businesses keep an eye on various platforms for any negative feedback, including changes in the frequency of such postings. This can allow businesses to manage their brand proactively before sentiment changes too drastically. 
• Crisis management helps businesses catch early negative reactions toward potential bad publicity, allowing for timely reactions to bad PR.  

Improved product development

Sentiment analysis can help businesses prioritize product development and features based on customer desires: 

• Prioritizing features becomes simpler with thorough sentiment analysis, as businesses can be sure they are developing or iterating upon actual consumer wants. 

Content personalization

Sentiment analysis can also help drive content personalization, leading to better results for a business: 

• Better marketing localization can help businesses expand to new locales or regions. By tracking sentiment, businesses can localize their marketing campaigns to attract the local audience. 
• Improved targeted marketing is also possible with sentiment analysis helping businesses understand consumer sentiment, allowing for iteration of materials to better suit the targeted market. 

What’s the future of sentiment analysis?

Sentiment analysis is still a developing field. There are likely to be several new developments as better tools and techniques are developed and refined. 

Improved algorithms and techniques

As sentiment analysis develops, new tools and better algorithms are likely to be developed. These improved tools and techniques are likely to lead to significantly more capable tools. Future sentiment analysis is likely to benefit greatly from the further refinement and development of AI-associated technologies. In particular, the natural language processing (NLP) computer science subfield is likely to yield dividends, as it is focused on teaching machines to understand and interpret human language. 

Advancements in NLP can allow for sentiment analysis to become better at understanding context and emotion. Currently, sentiment analysis can be constrained to understanding basic sentiments, such as positive or negative, while future developments may allow the analysis to track more detailed emotions like joy, excitement, anger or frustration. 

Similarly, better NLP techniques could allow for sentiment analysis to parse the context better of language. This could allow for a more accurate understanding when a person writes sarcastically, for example, which is currently difficult for algorithms to parse. This will also help algorithms detect implicit sentiment, which may be difficult for current analysis techniques to pick up on. 

Lastly, improved algorithms could also allow for enhanced analysis or nontext fields. This could allow for analysis of images, videos and audio recordings. 

Individual sentiment analysis

MacPhedran said the next generation of sentiment analysis is very exciting.

“Microservice APIs are able to measure emotion in written content, but also voice and facial expressions. For the sake of the example, assume that we have a CRM [customer relationship management] system that knows users’ social handles and has an image of the customer usable, with customer permission, for personalization based on facial recognition.”

With that knowledge, your business could better gauge that individual customer’s sentiment and target conversion strategies accordingly. [Learn more about facial recognition advertising.]

Prioritizing big data

You may need to invest in this analysis technology now or risk being outcompeted in the future simply because one company didn’t have key consumer data and another did. A business’s insights and, therefore, its success, will be limited by how much data it has.

“Because the backbone of sentiment analysis utilizes big data, using datasets that are comprised of thousands upon thousands of data points, retailers need to have enough data available (including customer conversations and reviews) to gain actionable insights,” Nie-Rouquette said. “So, in some cases where data is scarce, sentiment analysis might not provide good insights because of the lack of statistical validity.” [Related article: Big Data vs. CRM: How Can They Help Small Business?]

That’s a fixable issue and one that companies should address if they want to receive the maximum benefits of sentiment analysis.

“With the availability of data on various online sources, companies (and especially retailers) can leverage sentiment analysis to gather insights that would not be possible using traditional marketing methodologies,” Nie-Rouquette said.

Best tools for sentiment analysis

There are a wide range of sentiment analysis tools available for small businesses. Virtually all sentiment analysis tools can scan social media networks looking for mentions of your brand and your competitors. You get information back on the volume of content and whether that content was positive, negative or neutral.

You can also plug sentiment analysis apps into your email server and live chat systems, giving the apps instructions on what to do depending on how they interpret the message and the sentiment behind it.

Some notable sentiment analysis tools include: 

• Microsoft Azure AI Language: Microsoft’s AI language offering comes packaged with a range of features, including the ability to recognize core concepts, provide summaries, track sentiment, process unstructured data and categorize text automatically. This is an extremely powerful and flexible tool. However, for some businesses, it may be too complex with too many capabilities beyond just sentiment analysis. 
• IBM Watson Natural Language Understanding: This AI-powered tool uses deep learning techniques to extract meaning and metadata from text sources. IBM Watson can be integrated into an existing data pipeline to pull out text analytics, classifications, emotions, sentiment (in beta) and more. 
• Qualtrics XM Platform: Qualtrics is an experience management platform powered by generative AI tools. The customer-facing platform is focused on improving customer care by supporting frontline team members. This allows for reviewing customer sentiment during initial chat on the website, monitoring website interaction and reviewing customer sentiment across a range of websites and channels. 
• Sprout Social: Sprout Social is primarily a social media management tool. As it integrates and manages well with a brand’s social media accounts, it can compile data relevant easily to sentiment analysis across social media. This helps with tracking real-time consumer sentiment as well as competitive analysis. 
• Medallia: Medallia is a customer and employee experience management platform. The tool features its own native AI tool, Athena. Athena uses NLP to extract sentiment from across a range of channels, including customer surveys, short message service messages, emails, reviews and social media interactions. 
• Awario: Awario is a brand monitoring tool focused on a brand’s reputation on social media. It integrates with all of a company’s social media channels and then provides insights based on sentiment. Beyond tracking sentiment, Awario can help flag mentions or trends that may need immediate responses as well as tracking the reaction to marketing campaigns and new rollouts.  
• Meltwater: Meltwater offers a full suite of tools to help with brand management, business intelligence and social media listening and analytics. Among Meltwater’s offerings is the Meltwater API, which is focused on enterprise-level analytics. The API allows Meltwater customers to pull in data from over 10 million content sources, such as social channels and open web data to track overall sentiment.

The value of sentiment analysis

Sentiment analysis can be invaluable to a small business. For a company to succeed, it must be aware of how the marketplace is receiving its products and services. Sentiment analysis can tell a business how customers are feeling about the brand and its offerings. With that knowledge, companies can develop sales strategies that take into account consumer sentiment.

Going forward with sentiment analysis

Sentiment analysis is a dynamic subject area of technology undergoing rapid evolution in part due to the AI boom. Current techniques are already highly able to benefit businesses of all sizes by measuring consumer sentiment. This can help businesses identify pain products in their offerings or brands while also highlighting what consumers love about them. 

While current sentiment analysis techniques are somewhat limited to determining if sentiment is merely positive, negative or neutral, future advances in NLP and sentiment analysis algorithms should allow for a much deeper understanding of business. Such data is likely to be priceless for companies as they can get real-time information on how their consumers feel from social media channels, open web reviews, customer surveys and interviews. 

Although sentiment analysis may be difficult to implement fully in-house, numerous third-party applications and software make the process of analysis simple and an integral part of any future business strategy. 

Jeremy Bender and Brian O’Connell contributed to this article. Source interviews were conducted for a previous version of this article.