Positive, negative, neutral. Thumbs up, thumbs down. Smiley face, sad face, neutral face. So straightforward – but so meaningless.
Most executives want their customer feedback to be analyzed based on sentiment. This makes sense on the surface. After all, it can only be a good thing to know what percentage of the feedback you get is positive as opposed to negative and be able to track those trends, right?
Unfortunately, although basic sentiment analysis can provide great data points to include in your presentation to the boss, that’s about all the useful information you’ll get from it. To gain truly actionable insights about what your customers are feeling, you have to go a step deeper than positive, negative, and neutral. Instead, you must look at sentiment in a more nuanced way – an approach that people are starting to distinguish as “emotion analysis.” Emotion analysis, which incorporates artificial intelligence and natural language understanding, is rapidly replacing sentiment analysis as the go-to approach for companies who need to understand how their customers are feeling.
Companies who continue to rely on sentiment analysis instead of moving to emotion analysis must grapple with some inherent flaws in this older approach:
1. It Oversimplifies Your Data
Traditional sentiment analysis takes a comment or sample of text, identifies how frequently positive and negative terms appear in that sample, and averages those out to assign the sample one of three labels: “positive,” “negative,” or “neutral.”
For all-positive or all-negative reviews, this can be a fine approach. But what happens when the system comes across a review like this one?
“We love this restaurant! We go there at least once a week. The service is terrific and the food is awesome! But we went for dinner on a Friday and it was really crowded. Plus, our steaks took a long time to cook… But they gave us a free glass of wine and we ended up being really happy.“
Most sentiment analysis engines would see that the occurrence of positive and negative terms is about equal, averaging out to this being a “neutral” review. The nuance and strength of emotion in this review is completely lost. Not only that, but….
2. It’s Misleading
This review is pretty typical as far as product and service reviews go; it’s full of nuance and specifics, and contains some criticism even though the customer was satisfied overall. However, standard sentiment analysis approaches would label this and similar reviews as “neutral.” Most of us wouldn’t say that the presence of both positive and negative comments means that they should cancel each other out – but this is exactly what happens with sentiment analysis. A skewed understanding of your data is the result.
3. It’s not Actionable
So, besides being oversimplified and misleading, what actionable insights can you take away from a standard sentiment analysis report that tells you that 57% of your feedback is positive, 23% is negative, and 20% is neutral? What can you do, specifically, to improve the customer experience with that information? Nothing.
Without knowing why your customers feel the way they do – or having an accurate understanding of how they’re actually feeling, beyond just “positive” or “negative” – it’s impossible to identify ways to improve. This is a key drawback to traditional sentiment analysis, and a big reason why companies are moving to a more comprehensive approach.
Time to Start Getting Emotional
For many companies, these are critical flaws that prevent them from truly understanding their customers and uncovering actionable insights. As a result, they’re turning to emotion analysis.
A key distinction between emotion analysis and sentiment analysis is that, rather than labelling data according to pre-existing categories, emotion analysis starts with the data and creates categories of emotions from there. An analysis starts by identifying what emotions are present in the data. Excitement? Surprise? Annoyance? Fury? In emotion analysis, these are all treated as the distinct feelings they are instead of being clumped into broader but less meaningful groups.
The next step in emotion analysis, which is often entirely missing from sentiment analysis, is a deep dive into the topics associated with each emotion. What, specifically, are people excited about? What annoys them? What infuriates them?
Having this information gives you actionable insights you can act upon. For example, if you identify that people are excited about the release of a specific feature, you can make sure that you’re promoting that more in your advertising. Conversely, knowing what issues are infuriating vs. merely annoying is hugely valuable when it comes to prioritizing issue resolution efforts and how you communicate with disgruntled customers.
Ultimately, capturing and analyzing the full spectrum of human emotion that appears in your data is critical to uncovering deeper insights and making smarter, more targeted business decisions. Sentiment analysis is no longer enough, especially now that artificial intelligence and natural language understanding have made it possible to analyze data based on specific emotions.
Companies that continue to rely on sentiment analysis will quickly fall behind. It’s time to make the switch from black-and-white to full color.