Sentiment analysis is a powerful tool that enables us to reduce the effort and costs of collecting information from across our customer base. However, as with most things, there is great value in understanding how to use this technology properly so that we maximize its effectiveness and avoid pitfalls. In this blog post, I will share what I consider to be the seven deadly sins of sentiment analysis in marketing. They may not be actual “sins,” but if you do any of these things, they will likely lead to a less than optimal application of the technology and a potentially negative effect on its objectives.
Here are the three Deadly Sins of Sentiment Analysis in Marketing:
1. Failing to understand or recognize that sentiments exist in an organizational context
The sentiment is a powerful thing. When we ask for feedback from customers, we implicitly invoke feelings about our company, product, service, or brand. Asking for this data is actually more than just asking customers how they feel at that moment; it is also asking them to invest their time providing us with information from which we will form opinions that have the potential to affect how our behaves going forward. Because of this, it is common to see a powerful effect, known as the priming effect, in which the request for feedback affects the sentiment expressed by customers. In other words, when we ask for our customers’ opinions about us or our products, their judgments are altered due to this “priming” and they end up agreeing with what we’re saying more often than they would have if someone had just asked them directly.
2. Failing to recognize and account for differences in strength of opinion
One of my most important rules in life is: “Never doubt that a small group of thoughtful people can change the world.” This quote from Margaret Mead holds true today perhaps more than ever before. I believe we live in an age where companies must focus their energy on a series of narrowly defined constituencies and use technology to aggregate their opinions in order to gain a significant competitive advantage. In the case of sentiment analysis, these constituencies may include customers, prospects, employees, analysts, etc.
A similar kind of anomaly can happen between employee and client sentiment. Let’s face it: companies hire us to tell them what we think about their products and services so they can improve them. As such, we should expect that client-facing employees will be more critical of products and services than will the customers themselves. For instance, if a customer is asked to rate a product or service on a scale from 1–10, it is not uncommon to see scores near the top end of the range because customers want to give companies the benefit of the doubt. This same data from clients-facing employees might produce average scores in the 4–8 range depending on their relationship with management and other factors. The point is that different audiences have significantly different levels of positivity bias – even within an employee population – so you should use caution when comparing sentiment across groups. Taking these differences into account can yield significant insights about why your company’s perception rating differs from research conducted by third parties as well as from what you hear from your own employees.
3. Failing to account for context and personalized experiences
Another big obstacle when attempting to gauge the true sentiment of a customer is trying to determine whether their reaction is a genuine opinion or simply a knee-jerk reaction to a specific situation. A good example of this might be a comment like “I can’t believe how long it takes them to make changes.” The statement itself points toward dissatisfaction, but if we were able to ask more questions (e.g., Why do you say that? What specifically was frustrating?) We might find out exactly why the person is upset and uncover ways in which the company’s workflows could be improved, thus increasing satisfaction with the company.
What’s more, context directly affects both the sentiment expressed and how it will be interpreted by companies looking to understand customer views. People can feel very differently about a product or service depending on where they are when they are using it. At work, people might have very different expectations for an enterprise application than they would have at home or on their mobile devices. If you really want to know what your customers are thinking, ask them in all of these contexts because their opinions could vary significantly from one to the other.
Conclusion:
Most importantly: Ask for specific examples (and not just general overall sentiment)
The best way to ensure that we’re getting an accurate representation of how our customers feel is by asking for concrete examples that illustrate why our clients feel the way they do. There is no substitute for specific words that describe their experience with our company’s products or services, especially if we can verify that these examples are accurate by asking follow-up questions.
