Sentiment Analysis: A Thorough Guide For The Data Geek

Feb 27, 2020

13 mins read

Tanuj Diwan


In 2020, business owners don’t get and don’t have to meet their customers in person. Every single business transaction happens online requiring little or no direct interaction.

However, the need to know and understand what customers feel and say about a brand has not changed. In fact, with virtual interactions becoming the norm, businesses are more curious to know customer sentiments towards their brand.

But, when there is customer feedback data pouring in from all sides — like email, social media, and other communication channels, it is difficult to filter out signals from all the noise.

Even if a business is able to set up an entire team of ‘customer listeners’, they may not be able to process all the information to figure out accurate customer sentiments.

With sentiment analysis, a business can get a clear pulse about customer sentiments, determine positive, negative or neutral sentiments and take corrective actions.

What is sentiment analysis?

Sentiment analysis is an analysis of textual data that gives a polarity of sentiments — positive, negative or neutral along with an associated score around it.

For example, a customer review of “I like the product’s minimalist interface.” would have a positive polarity of 70% or more.

 sentiment Analysis definition


Since what sentiment analysis does is more like a judgment of the opinion that textual data carries, it is also referred to as opinion mining.

Although Sentiment analysis has become popular in recent years, work on it has been progressing since the early 2000s.

During that period, there was a lot of buzz around how data can be collected and analyzed to form clusters of similar topics.

The data was either curated by a human or fetched from a dataset. Then, the data was analyzed manually to determine the polarity of sentiments as positive, negative or neutral.

How does sentiment analysis work?

Sentiment analysis works with the help of a tech architecture that looks at text not just as data but as data which when combined together makes up a holistic meaning.

The text data could come from diverse sources like social media, blogs, websites, email, short messages, metadata of images and so on.

For example, in a sentence, “The food that we had yesterday was great.”, the first three sentences may not provide much context in sentiment analysis. But when each word is built upon another and read together as a string,

it gives a holistic meaning. Sentiment analysis does this with the help of Artificial Intelligence and Machine Learning algorithms.

how Sentiment Analysis works 2020


There are also several approaches to sentiment analysis that helps determine the sentiment involved in data.

Rule-based algorithms

Sentiment analysis is carried out based on manual rules set by programmers.

  • For example, identify positive or negative words in a sentence.
  • Collate the number of positive.negative words.
  • Compare the overall score of positive and negative words.
  • If the overall score is above a specified threshold, polarize the sentiment as positive or negative.

The image below shows how rule-based sentiment analysis works.

Sentiment Analysis Rule-based algorithms


In rule-based algorithms, the task of curating data, feeding it to the system and also maintaining the system is done manually.

As a result, when the data needs to be changed, or when complex textual data has to be analyzed, rule-based algorithms might seem to be insufficient.

Automatic systems

Automatic systems used for sentiment analysis follow the same process, except that they are more independent and do not require extensive manual intervention.

A machine learning algorithm takes over the task of classifying the input text, extracting the cues out of it and giving the output as positive or negative.

The system is taught to identify sentiment as positive or negative from the data used for training.

Automatic systems Sentiment

There are specific Machine Learning algorithms that are used for this purpose. Some of the most common ones are:

  • Naive Bayes
  • Linear Regression
  • Support Vector Machines
  • Deep Learning

Hybrid systems

Hybrid systems, as the name suggests, is a combination of both rule-based and automatic systems. Since they have a balance of automation and also manual processes, they are considered to be more efficient in sentiment analysis.

How business can leverage Sentiment analysis to deliver the superior customer experience

Now let’s discuss the burning question. How can a powerful technology like sentiment analysis help a business deliver superior customer experience?

Although the possibilities are endless, there are specific ways of how sentiment analysis can help. Some of them are:

Social media monitoring

Social media is the new amphitheater for the world’s events. In this amphitheater, good and bad news spreads like wildfire within a matter of hours.

To make things great or worse, it also reaches a global mass of audience. That said, businesses can’t afford to ignore what people (or customers) are talking about their brand in social media channels.

Even a random tweet of poor customer service, which if not attended to promptly could snowball into a PR agency nightmare.

Like it happened with Times Warner Cable. A customer’s rant tweet went on to make 1,840 take notice and also express their empathy with the same.

social media sentiment analysis


From a customer point of view, social media empowers them to voice their opinions and experiences with the power of a public relations agency.

These opinions largely come in the form of Facebook statuses, tweets or retweets, Facebook likes, emojis, comments and so on. In other words, all kinds of textual data that a business can make use of for monitoring their social media image.

Where does sentiment analysis fit in amidst all of these? Sentiment analysis helps monitor the kind of conversations that customers are having about a brand on Facebook, Twitter, LinkedIn, Instagram and similar websites.

Timely responses to negative comments and topics can help the business steer far from PR catastrophes.

A typical sentiment analysis used for social monitoring will showcase the reports as under. It will depict the number of positives and negatives and how they rose and fell during a given period of time.

The report would look similar to most analytical dashboards available in the market.

sentiment analysis Dashboard report


There are several other benefits that social media monitoring using sentiment analysis can bring to a business:

  • Develop a data-backed social media strategy that appeals to target market sentiments
  • Understand the most talked about topics by customers
  • Measure the reach of an organic (non-paid) social campaign
  • Score the brand perception amidst customers against competitors

Analyzing customer sentiments towards competition

Every business worth its salt exploits competitor shortcomings to its advantage. But, spotting competitor weaknesses is not a mean feat. Sometimes one has to spy on customer feedback on competitors to spot such weaknesses.

Thanks to the widespread use of user reviews and ratings on Google, Twitter, Amazon, and similar review websites,

it is easy for customer reviews and analyzes them for customer sentiments. Sentiment analysis can make this task easy as well.

Take the example of an Amazon customer review about a competitor’s product. With sentiment analysis, it is possible to figure out the key points about the competitor that disappoints customers.

Sentiment analysis can sift through customer reviews and cherry-pick specific words that indicate positive, negative or neutral sentiments.

Some examples are:

  • Good quality
  • Average product
  • Poor customer service
  • Easy installation
  • No phone support
  • Waste of money
  • Will buy again

By understanding what customers dislike about competition, a business can work to build on those capabilities. The end result would be a better customer experience and an inflated bottom line.

Identifying promoters and detractors

For a business that uses NPS® surveys to measure customer loyalty, identifying promoters and detractors is a critical activity.

Promoters are those customers who would remain loyal to the brand and would also refer others to it. Detractors are unhappy customers who could portray a negative image of the business.

The good thing about NPS surveys is that customers have the option to leave an optional note on why they would or would not refer to a friend or dear one to the product.

Sentiment analysis can analyze all the customer responses, collate them and create polarity of the sentiments.

For example, to an NPS survey question that asks what is holding you back from referring us to others, the possible responses could be:

  • High price
  • You don’t care for customers
  • The Product didn’t work as expected
  • Great features
  • Like the interface, etc.

Sentiment analysis can review these responses and group them into specific groups for a better understanding of customer responses and their loyalty. Ultimately the findings can also be used to strategize ways to better customer experience.

Why should businesses take sentiment analysis seriously?

Sentiment analysis is an excellent way to discover how people feel about a product, topic, service or idea based on their textual data.

For businesses, this is as good as getting to know the true feedback that customers won’t otherwise provide through surveys, emails or questionnaires.

Today, there are millions of bytes of data that are being floated on the internet. From social media posts to blogs, tweets, and comments,

there is an abundant variety of data that can be subject to analytics. If analyzed properly, it can tell businesses the gaps that they have to fill to meet customer expectations.

Although the list of sentiment analysis applications is endless, there are some solid applications that are mainstream today.

Track customer sentiment over time

In September 2019, Nike released an ad featuring the controversial National Football League quarterback Colin Kaepernick.

Kaepernick became a hot topic of discussion in America when he, like many other footballers knelt during the national anthem to protest against police brutality.

Nike’s announcement that Colin would be one of the athletes helping commemorate the 30th anniversary of the brand’s iconic slogan took America by storm.

Angry customers who were upset with Colin’s ‘kneeling’ took to Twitter to show their dissent. The hashtag #justburnit soon became a viral trend.

Sentiment analysis shows that Nike’s net sentiment for the month nosedived.

sentiment analysis example,


However, on the flip side, despite the negativity, Nike’s sales resulted in a 31% jump from the controversial ad (MarketWatch). According to 4C Insights, Nike’s social media mentions jumped by 1678%, while that of Kaepernick spiked 362,280%.

Needless to say, these figures would have been impossible to track if not for sentiment analysis. Without sentiment analysis, it would have been easy for anyone to make a wrong judgment that the ad was a blunder while the reality was the contrary.

Brand mentions

Social media monitoring restricts itself to the ambit of social media websites. Whereas, brand mentions encompass the entire web, including news articles, blogs, press releases and so on.

Similar to social media monitoring, brand mentions look for specific instances on the web where the brand was mentioned.

The use of text analytics helps in finding the exact instances where the mentions are made, with contextual correctness.

Here is an example of how brand mentions would look like in Twitter:

Sentiment Analysis Nike's Example


Sentiment analysis, with its capability to identify the emotion behind the brand mention, can enable the business to determine whether it is good news or bad news that is being spread about them.

For example, in the TW Cable example above, the tweet got picked up by regional news portals and websites which further blew up its widespread distribution.

Benefits of using sentiment analysis for brand mentions:

  • Set up alerts when the brand’s name is used anywhere on the web
  • Identify the polarity of overall customer sentiments by analyzing blogs, news articles, website content, etc.
  • Have a bird’s eye view about public notion about a brand

Analyze customer feedback

Every business that deeply cares for customers spends time and resources to collect their feedback. This feedback is usually collected in the form of email questionnaires, NPS surveys, social media polls, user reviews and ratings, phone conversations and so on.

However, all the customer feedback that is received will be in the form of unstructured data, meaning the data will not have a common shape or form. This makes data analysis cumbersome.

Thanks to sentiment analysis, which uses Machine Learning algorithms, even the unstructured data can be sifted through to pick up the underlying sentiments in the data.

For example, all responses to an NPS survey can be collated to identify the overall polarity of customer sentiments.

It also segregates positive and negative sentiment-based responses separately thus enabling the business to reach out and serve specific customers in a better manner.

Conduct market research

For newbie market analysts as well as veterans working in the internet era, an online search is a preliminary form of conducting market research.

Of course, there are other offline forms of market research that deliver valuable insights. But, an online search is a simplest and easiest way that is available right now. And it has an inherent flaw.

Web search results are always biased towards web pages that are search engine optimized.

There is a fat chance that web pages that conduct solid material for market research may not get featured in the top SERPs at all. That could handicap the market research activities.

In such a scenario, sentiment analysis can provide an added advantage. It can scan over the internet and handpick instances, news and market reports that are relevant to the business.

In other words, it helps market research executives find new sources of information that are otherwise unavailable.

SurveySensum for sentiment analysis

SurveySensum is an AI-enabled customer experience management tool. It was created with the vision to help businesses serve customers better by understanding their feedback,

Net Promoter Score responses and also sentiment analysis.

As a sentiment analysis tool, SurveySensum offers several unique features that help businesses ramp up their customer service efficiently.

Multi-lingual support

Speak the language of your customers. Understand the language of your customers. SurveySensum’s sentiment analysis capabilities will help you decipher the hidden meaning in your customer interactions no matter which language they are in.

Targeted Sentiment Analysis

Sentiment analysis is a broad topic. Without a proper direction, you are bound to get mixed responses about customer sentiments. With SurveySensum, you can target a specific range of customer interactions or data — like email, chat, NPS survey responses, etc. to derive maximum insights.


Almost every AI and ML-based software have some kind of inherent approximation. If the approximation is skewed disproportionately,

it can make decision-making difficult. However, SurveySensum’s sentiment analysis is primed for accuracy. Rest assured, you will have a perfect sense of what your customer sentiments are all about.

Trusted by all

SurveySensum’s sentiment analysis is used by customers big and small. From global players with continent-spanning business presence to startups who are just getting started with their big journey.

we have helped businesses of all sizes and types save their precious time and resource with sentiment analysis.

Tanuj Diwan

How much did you enjoy this article?