We all know by now that online reviews matter to businesses, whether it’s in regards to SEO, word-of-mouth marketing or just creating a positive online presence. But, it can be easy to get caught up in the amount of stars you have and not pay attention to what users are actually saying. Here is why review sentiment is something your business should take a look at.
What is Review Sentiment?
Sentiment Analysis is a fancy term for looking at a large set of data and extracting the opinions reflected in the text. The algorithms used to interpret this data usually focus on one of three things:
- Polarity- Are the opinions reflected positive or negative?
- Subject- What is being talked about?
- Opinion holder- Who is doing the talking?
Sentiment Analysis has many practical applications in our digital world. This analysis technique allows data scientists to create structured data out of paragraphs and words—something that has not been a possibility until recently.
Why is Review Sentiment Important?
Sentiment analysis can be helpful in determining several things about your data.
Sometimes, you may want to dive into your data granularly. Rather than simply classifying data as “positive” or “negative”, you can categorize reviews as “very positive”, “positive”, “neutral”, “negative” and “very negative”.
Certain keywords can be used to detect emotions such as happiness, anger, sadness and more. However, sometimes words such as “kill” can be used in reviews that are both positive and negative. Example: Your staff is garbage and the customer support is killing me! Vs Your staff is killing it on this busy night!
When looking at product reviews, you may be interested in the features that people are talking about. For example, if your company manufactures and sells cameras, your product reviews could tell you that people think the battery life on your new camera is too short, or that they love the new color options.
The AirBnb Experiment
When data scientists at Airbnb wanted to know what customers were saying, they turned to a little bit of data mining. Using a spreadsheet of Google reviews, they were able to determine that, while customers left positive ratings, the text that accompanied those ratings didn’t always line up with the 5 star rating they left. By writing their own sentiment analysis algorithm, they found something that alarmed them. Users did not want to offend their hosts, so they left four or five star ratings, but hid their true feelings in the comments of the review. They discovered the importance of actually reading the reviews, rather than focusing on the cumulative rating.
In Airbnb’s case, they have used this data to fix the pain points brought up in these reviews. Ultimately, this will lead to happier customers who feel valued.
We wanted to put the review sentiment concept to the test, so we took a look at one of our clients’ review sentiment ratings. Awesome Company (name changed for privacy reasons) has a 4.3 star rating on Google, but has several five star ratings with long paragraphs of text accompanying them. We wanted to take a deeper dive into what these reviews were actually saying. So, we built a spreadsheet (our favorite thing) full of their reviews and put it into a review sentiment analyzer. We were surprised by the results.
A sentiment analyzer like we used takes a deep look into the text you feed it. It will look for positive and negative words, but will also consider the words around it. For a good control, you can view what one of these analyzers says about the Declaration of Independence here.) Sentiment analyzers are “trained” by “feeding” them thousands of text documents across genres and domains and “teaching” them what is positive and what is negative. (A lot of coding goes into this.)
Sentiment scores range from -100 to +100, where -100 indicates a very negative or serious tone and +100 indicates a very positive or enthusiastic tone. The higher the number is on the positive side, the more satisfied a customer is with your business.
Despite Awesome Company having a “positive” star rating, the comments with the reviews were only slightly positive. A lot of the comments mentioned wait times or other perceived issues, which brought the overall positivity down.
We took the time to go over these findings with our client, and talk through how some of the expressed concerns could be improved upon. Awesome Company had no idea patrons were experiencing these issues and were happy to have it brought to their attention.
Taking the time to uncover your brand sentiment is a valuable asset to your business. It provides an excellent source of data to provide consumer insights on brand reputation, improve customer experience, stop issues from becoming a crisis, determine future marketing strategies, improve marketing campaigns and product messaging, identify influencers, test OKRs, generate leads, etc.
Curious what your customers are saying? We’d love to chat about it.