Today’s post- How and Why Companies Should Use Sentiment Analysis – is written by featured author Federico Pascual, co-founder of MonkeyLearn, a powerful machine learning tool allowing you to extract valuable “opinion-based” data from text.
The first time someone tried to talk to me about sentiment analysis, I thought it was a joke. “Machines can do what?” “In how much time?” “With that much accuracy?” Baffled, to say the least. For a split second, I considered nodding along in feigned interest while slowly but steadily tuning out. Surely this had to be either over my head or useless to my daily life.
Happily, though, the generous person with whom I was speaking trudged on, and soon enough my curiosity got the better of me. I was hooked mostly by the scale of it all — the fact that machines could process and analyze millions of tweets in a matter of minutes (a feat that would take me years at least, never mind the speed reading courses), and what that meant for how we understand and act upon the vast information available to us.
After that initial conversation, I read more literature, I learned the processes, and I began to see use cases of sentiment analysis in real life. I co-founded MonkeyLearn, a company built around enabling all kinds of organizations to access and benefit from sentiment analysis. And I’ve spent the last four years watching how it changes our clients’ workflows, daily interactions, productivity, and long-term success. So now I stand here with a little more familiarity (and zero feigned interest) and assert that sentiment analysis is neither over our heads nor useless to our daily lives. Rather, it can be accessible for everyone — and in the future, it will even be essential.
In this piece, we’ll cover the basics of what sentiment analysis is and examples of how companies use sentiment analysis to improve their work.
What is sentiment analysis?
Sentiment analysis is the automated process of discerning opinions about a given subject from written or spoken language. It’s also known as Opinion Mining, and is a field within Natural Language Processing (NLP). It offers us information about both the polarity (positive, negative, or neutral) and subject of an opinion.
How Sentiment Analysis Works
We won’t get too deep into technicalities here, but the gist is that using machine learning algorithms, we can train a model to associate particular inputs (texts) with corresponding outputs (tags). The more we train our model, the more accurate it becomes. The algorithm is designed and guided by human intelligence, but maximized by the automating power of computers.
Why Sentiment Analysis Matters
An estimated 80% of the world’s data is unorganized, much of that in textual form such as emails, support tickets, chats, social media, surveys, articles, and documents. Manually sorting through it all would be difficult, expensive, and impossibly time-consuming. Using sentiment analysis allows us to make sense of this chaos through automation, yielding actionable insights otherwise unattainable.
- Scalability: Process vast amounts of data efficiently and at low-cost.
- Real-time analysis: Immediately identify and assess potential crises, from PR to customer support.
- Consistency: Apply the same criteria to all data, reducing variance. (Versus subjective human judgment, which is heavily clouded by personal experience and beliefs. When evaluating sentiment of a piece of text, humans agree only 60-65% of the time.)
How Companies Use Sentiment Analysis
Sentiment analysis can revolutionize the way we do business. The use cases are many, but today we’ll zero in on three: product feedback, communication, and customer support.
Think for a second about all the data available from customers about a given product: online reviews, social media, Net Promoter Scores (NPS), support tickets, chat conversations — to name a few. Now think of a product development team you know that has the time or bandwidth to read even a slice of that data. We know our team, for one, would be quite swamped.
Let’s say you work for Slack. In Capterra alone, there are more than 6300 reviews, rich in valuable and authentic opinions about Slack that could make or break business decisions. With aspect-based sentiment analysis, we can analyze these reviews and easily dissect what people praise or protest about their experience with the product. Using MonkeyLearn, we analyzed and visualized the reviews:
In one afternoons’ play, we suddenly have real, digestible information we can work with — and this is just Capterra. Imagine if we added analyses not only from other review sites (G2crowd, Siftery, Crozdesk, etc) but also from other sources of feedback like social media, customer support conversations, NPS, and surveys. Imagine if we ran all the same analyses but for our competitors’ products. This is the future baseline of product development.
In today’s digital, multi-channel world, it often feels like there’s no off switch for communication. More than ever before, companies are accessible to (and sometimes at the mercy of) current customers, potential customers, anti-customers, and really any one of the billions of people on the internet. Faced with such ambiguity, how can we distill effectiveness of our communications?
Just for kicks, we decided to do some analysis on how the four biggest US phone carriers (AT&T, Verizon, Sprint, and T-Mobile) handled customer interactions on Twitter. We downloaded tens of thousands of tweets mentioning the companies (by name or by handle), and ran them through a MonkeyLearn sentiment model to categorize each as positive, neutral, or negative. We used our Insight Extractor to find the most relevant keywords and sentences containing them.
We found that:
- T-Mobile had far and away the highest percentage of positive tweets.
- Verizon was the only company with more negative than positive tweets.
- Top keywords for positive tweets at Verizon included typical terms such as “new phone,” “thanks,” and “quality customer service.” Key sentences were typical, formal, somewhat dry interactions between the team and followers.
- Top keywords for positive tweets at T-Mobile included names of people on their customer support team, likely because they run higher engagement, back-and-forth about anything type conversations with followers.
This could imply that a more personal, engaging take on social media elicits more positive responses and higher customer satisfaction.
Even on the best days, customer support teams are faced with inefficient ticket prioritization, repetitive manual tasks, and lack of reporting and insights. Sentiment analysis is here to oust all of those.
Oftentimes, support tickets are listed–and therefore answered–in order of submission. This is perhaps the worst plan of attack, and could lead to a simple suggestion being addressed prior to a distressed key client in their time of need. While every ticket is important, not every ticket is urgent. We can use sentiment analysis to put top priority tickets at the top of our team’s to-do list — improving first reply time and alleviating stress for our agents. Instead of clicking aimlessly in and out of conversations, losing time and draining mental energy, sentiment analysis automatically tells us where to start.
What’s more, we can use sentiment analysis to detect what a given ticket is about and anticipate replies. Rather than having to manually type repeat answers, agents can select which suggested response is best and adjust from there.
Finally, sentiment analysis can improve our long-term game, something that often gets shafted in the immediacy of customer support. We can track trends in complaints and questions over time, and use that data to improve self-service channels and user documentation.
We’ve seen a snippet of the potential impact of sentiment analysis to analyze product feedback, prioritize and streamline customer support, and improve customer communications across channels — truly, this is just the tip of the iceberg. If there’s one thing you hear from us today, it’s that sentiment analysis is not over your head and that it can transform how you work.
Whenever you’re ready to start playing around with its potential, we’ll be thrilled to walk you through it!
If you are ready to learn data science full throttle check out some of the courses you can apply to here.