Sentiment Analysis to understand your customers’ opinions at a large scale

Consumers  avidly express their opinions about the products and services they use on social media. Using sentiment analysis, this wealth of information can be analyzed to monitor your brand’s reputation, gauge customer satisfaction and target customers based on their satisfaction levels (polarity scores). Similar to gauging company performance using quantifiable performance metrics, sentiment analysis quantifies subjective information such as customer reviews. This changes everything!

But first, what is Sentiment Analysis?

Sentiment analysis is a process to identify, extract or characterize subjective information, such as opinions, expressed in a piece of text. Specifically, sentiment analysis is an algorithm that can predict polarity scores from given text. This kind of analysis is helpful in understanding the emotional opinion expressed in a Google review, tweet, Facebook post and so on.

There are two methods to build a sentiment analysis model: rule based model and machine learning based model. Whereas the rule based model is usually faster in inference time, the machine learning model is more accurate and reliable in many cases. A large amount of text data is required to build a machine learning model while a rule based model does not need any kind of data, just rules for common words such as:

  • ‘awesome’ equals a score of 1.0
  • ‘good’ equals a score of 0.9
  • ‘terrible’ equals a score of -1.0 

How does it work?

Let’s take a look at examples to see how it works. A marketing representative is monitoring Twitter to understand how people felt about their customer service. He reads the first two tweets: 

  1. “My favourite branch. Awesome customer service!”
  2. “Totally dissatisfied with the service. Worst customer care ever!”

For these two tweets, the representative can easily understand whether their experience was positive or negative. But what if there are thousands of tweets you have to read, plus, thousands of Google reviews and other posts? With such large data, sentiment analysis can be used to gauge polarity scores in customer reviews. Here, we are using Google Cloud Platform’s (GCP) Natural Language API to analyze the examples:

Example 1

Example 2

GCP’s Natural Language API calculates sentiment score which indicates positive sentiment with a value greater than zero, and negative sentiment with a value less than zero. The score for the first example is 0.8 which means that the customer’s experience was positive, whereas in the second example, the score is -0.9 which means that the customer’s experience was extremely negative.

Use Cases

Sentiment analysis can be used for all kinds of discoveries. Here are two examples of where sentiment analysis can be used in companies:

Brand Monitoring

You might want to keep track of your brand’s reputation. There are many different sources to retrieve data from such as social media postings, emails from your customers, product reviews and more. By using sentiment analysis, you can easily monitor the sentiment score for words used by your online customers.

Customer Targeting

Now that you understand people’s opinions about your brand and products in a quantifiable manner, you can use that information to make new marketing strategies. You may segment your customers on the basis of sentiment score, then offer different promotions to the different groups.

Do you have to build your own model?

Building machine learning model requires a lot of data, computational power and time. You don’t have to know fancy machine learning algorithms to build your own model for sentiment analysis. All the major cloud service providers offer their own models based on state of the art machine learning algorithms.

All three platforms give you free credit when you sign up. Try their sentiment analysis features to see how it can help your business!

GCP as a CDP

Create your own Customer Data Platform leveraging the power of Google Cloud A Customer Data Platform (CDP) is an important utility for analysts and especially