You've successfully subscribed to Thematic
Great! Next, complete checkout for full access to Thematic
Welcome back! You've successfully signed in.
Success! Your account is fully activated, you now have access to all content.
Success! Your billing info is updated.
Billing info update failed.

Top 3 reasons why most NLP fails

Understanding customer comments, on a large scale, needs to be automated. So, it requires Natural Language Processing (NLP) or Text Analytics. Unfortunately, most open-source NLP tools were developed on text researchers have easy access to. These are typically news articles, research papers and movie reviews. I learned that the analysis of customer comments is quite different, and here is why.

1. Customer comments are messy

What people share with companies in free-text is extremely messy. There are no regular sentences, no attention to white space, misspellings, abbreviations, autocomplete errors etc. In open-ended questions in a survey, many write meaningless idk (I don’t know), nothing specific, all good, yes, p, etc.

2. Customer comments are diverse

There are many people who just comment on one thing, e.g. too expensive. But just as many like to list various aspects. For example, let’s take this sample sentence.

He made a painfully long process bearable with regular touchpoint.

It relates to several issues at once. It describes dealing with the company process overall as negative. And, it describes the respective customer support person and their communication as positive. Complaints are particularly lengthy with stories explaining in detail what went wrong.

3. Customer comments are hard to interpret

Often, capturing themes is hard. Let’s say, a customer says:

 She made me feel as if she really had time for me.

How would one categorise this? Even a person reading this would spend a lot of time linking this aspect of customer service with similar ones that may appear in the customer responses.

So, understanding customer comments is fascinatingly difficult. One size does not fit all when it comes to NLP. Therefore, each type of text requires a different approach. And, because most open-source NLP packages (or NLP APIs for that matter!) were developed on grammatically correct and uniform text, they often fail on customer comments.

If you work in a Voice of Customer role and need to make sense of your customer’s comments, don’t trust those who claim it’s a piece of cake. It’s not as easy as plugging in a library.

Ready to scale customer insights from feedback?

Our experts will show you how Thematic works, how to discover pain points and track the ROI of decisions. To access your free trial, book a personal demo today.

Recent posts

How to theme qualitative data using thematic analysis software
How to theme qualitative data using thematic analysis software
Members Public

Become a qualitative theming pro! Creating a perfect code frame is hard, but thematic analysis software makes the process much easier.

How to measure customer satisfaction: the complete guide
How to measure customer satisfaction: the complete guide
Members Public

Everyone says they want customers to be satisfied, but what are you actually doing to make customers happy? How do you know if you’re on the right track? How do you know if your customer satisfaction efforts make a difference? Why even aim for customer satisfaction at all? We

Churn & Loyalty
How to super-charge your Qualtrics setup with Thematic & Power BI
How to super-charge your Qualtrics setup with Thematic & Power BI
Members Public

Qualtrics is one of the most well-known and powerful Customer Feedback Management platforms. But even so, it has limitations. We recently hosted a live panel where data analysts from two well-known brands shared their experiences with Qualtrics, and how they extended this platform’s capabilities. Below, we’ll share the

Customer Experience