An unsupervised approach to demystifying customer-to-customer interactions in the banking sector: the case of the Moroccan banking sector

Purpose Most online customer-to-customer (C2C) interactions occur in uncontrolled environments. This paper explores such interactions in the Moroccan banking sector by analyzing customer discussions on social media. Design/methodology/approach Data were collected from three Facebook groups representing Morocco's leading banking institutions. After preprocessing, two analyses were conducted. The first used word clouds to identify frequently used terms in customer discussions. The second employed clustering to group interactions and uncover dominant topics. Findings The word cloud analysis revealed “deductions” as the most frequently used term, while the presence of words like “thieves” reflected strong negative sentiments. The clustering analysis identified four main themes. The first cluster focused on issues such as unexplained deductions, loan eligibility and mobile banking concerns. The second reflected customer distrust, with complaints about unauthorized insurance and perceived dishonesty. The third addressed account management challenges, particularly the difficulty of closing accounts. The fourth discussed general financial topics, including taxation and stock trading. These findings highlight the need for banks to monitor and engage with C2C interactions proactively. Originality/value This study provides a novel perspective on how customers communicate with each other online and outside of formal banking channels. Unlike traditional surveys, our approach captures real-time, unsolicited feedback that can guide banks in improving communication, service delivery and reputation management.