Guest Post: Exploring Natural Language Processing to Categorize Customer Feedback

This week, we feature an article by Sheila Bugal, Head of Marketing at Caplena, a market-leading text analysis tool. She shares how companies can use Natural Language Processing in conjunction with human capabilities to enhance customer service.

Natural Language Processing (NLP) helps computers understand and process human language, which has powerful implications for businesses that want to offer increased communication with customers and clients without the high cost of hiring additional staff. Ever chatted to a chatbot online? Chatbots can answer your questions and offer help because they rely on NLP to evaluate natural human language. However, A.I. – and NLP – aren’t perfect solutions for processing valuable information. 

NLP: A Powerful But Imperfect Solution

Some tasks can’t be replicated by a machine. For example, advanced customization, abstract thinking and some types of complex problem solving cannot always be effectively performed by a machine. Businesses who want to use NLP to process customer feedback will find that this type of A.I. has limitations. 

When it comes to processing feedback, categorization is king. Categorization helps to efficiently organize feedback into categories like “Customer Service,” “Price,” “Ease of Use,” and “Features.” Then, categories may be divided into subcategories. Categories are key to producing actionable insights on top of rating questions, such as CSAT score (customer satisfaction), net promoter score (NPS), how customers are responding to specific features and why some customers may be unhappy with the product or service. 

However, NLP does have its limitations, for example: 

Cannot Design MECE Categories

Mutually exclusive and collectively exhaustive (MECE) categories are often used by management consulting firms to help problem solve. The basic premise is that to effectively fix a problem, all potential solutions must be able to fit into only one category (mutually exclusive), and all solutions must fit into a category (mutually exhaustive). MECE categories aim to eliminate confusion and help pinpoint actionable solutions. The ultimate result is that problem-solvers are better able to hone in on a solution that will fix the problem. 

While this is a specific application of MECE, this problem-solving framework is also an efficient and effective approach to any type of organization, including that of customer feedback. It helps reduce duplication that could potentially warp metrics, and it allocates every piece of feedback into a category, making it actionable. 

Unfortunately, NLP cannot perform MECE organization on its own. MECE requires human intelligence to design a framework of categories that will factor in all possible results and ensure that each result has a mutually exclusive “home.” 

Has A Reduced Ability to Customize

Another key to efficient categorization is customization. Each business may have a unique set of feedback categories that are best suited to its product, service or type of insight that it’s looking to gain. 

For example, most businesses will have feedback categories that apply to satisfaction with customer service or pricing – but some products or services may require additional categories. For example, a time-tracking app will need a unique set of categories that help process customer feedback on its ability to deliver accurate reports. A budgeting app will require categories that help process feedback on how accurately it categorizes purchases and so forth. 

NLP can customize categories to a certain extent – but still cannot match human intelligence in terms of creative, insightful customization that will allow owners and analysts to get the most out of customer feedback. 

Has Limited Adaptation to Customer-Specific Topics

Customers may surprise us with “creative” answers they give to otherwise straightforward questions. For example, they may bring up an entirely new issue that hasn’t been addressed by any customer in the past. 

In the case of open-ended questions, customers may provide feedback that doesn’t fit into a designated category – and could potentially be handled incorrectly by NLP. 

Ultimately, businesses that rely exclusively on NLP will miss out on some of the most valuable outcomes of gathering feedback – such as deep customer insights and accurate metrics (like CSAT and NPS). 

And yet, NLP is still highly useful and efficient and well worth investing in. Relying on the power of A.I., it helps automate many aspects of processing customer feedback, which helps businesses save valuable time, money, and energy that can be used on other tasks. 

Augmented Intelligence: Combing the Best of Both Worlds

Good news: There is a solution. 

Augmented intelligence combines artificial intelligence with human intelligence to get the maximum benefits of both: The speedy, automated capabilities of A.I. with the creative, conscious, and even emotional abilities of human intelligence. 

In terms of processing customer feedback, businesses that want to use NLP to gather feedback on their websites can still do so. But by relying on augmented intelligence to organize and process this feedback, they’ll get a deeper level of actionable insight into what their customers think, need, and want from their product or service. 

Augmented intelligence also has exciting potential to “fill in the gaps” left by artificial intelligence in other industries, such as education and healthcare. For example, augmented intelligence might be used in education to give teachers insights into their student’s learning behaviors and capabilities, but it won’t necessarily replace the teacher. It simply makes the teaching and learning process more efficient. 

Ultimately, augmented intelligence is a more effective approach across multiple industries. While artificial intelligence is an exciting solution offering increased efficiency and speed, it’s even better when used with human capabilities. 

Sheila Bugal is the Head of Marketing at Caplena, a market-leading text analysis tool, that empowers researchers, consultants, and insight teams to spend less time on the analysis and more time on the results.