Customer satisfaction and natural language processing

•NLP is an effective method for capturing and analyzing the customer’s voice.•Satisfaction is not vertical or horizontal, it’s a combination.•There is a relationship between the NPS score and the number of topics mentioned.•Not all topics contribute in the same way to customer satisfaction.•Automatic NLP is the first step to establish models of response to customers. This study uses natural language processing in order to increase knowledge concerning customer satisfaction. A total of 12,000 customer returns were analyzed, 6,800 of which contained freely expressed qualitative feedback. Eight themes emerge from the analysis and bring to light the factors influencing satisfaction. It is also noted that satisfaction is not vertical or horizontal but can involve a more or less important combination of themes. This study also shows the link between the level of satisfaction and the number of themes addressed, thus challenging traditional approaches that do not seem to distinguish the discursive differences between satisfied and dissatisfied customers. Finally, this investigation lays the foundations for automatic and personalized processing of customer comments.