Customer segmentation has become crucial for the company’s survival and growth due to the rapid development of information technology (IT) and state-of-the-art databases that have facilitated the collection of customer data. Financial firms, particularly insurance companies, need to analyze these data using data mining techniques in order to identify the risk levels of their customer segments and revise the unproductive groups while retaining valuable ones. In this regard, firms have utilized clustering algorithms in conjunction with customer behavior-focused approaches, the most popular of which is RFM (recency, frequency, and monetary value). The shortcoming of the traditional RFM is that it provides a one-dimensional evaluation of customers that neglects the risk factor. Using data from 2586 insurance customers, we suggest a novel risk-adjusted RFM called RFL, where R stands for recency of policy renewal/purchase, F for frequency of policy renewal/purchase, and L for the loss ratio, which is the ratio of total incurred loss to the total earned premiums. Accordingly, customers are grouped based on the RFL variables employing the CRISP-DM and K-means clustering algorithm. In addition, further analyses, such as ANOVA as well as Duncan’s post hoc tests, are performed to ensure the quality of the results. According to the findings, the RFL performs better than the original RFM in customer differentiation, demonstrating the significant role of the risk factor in customer behavior evaluation and clustering in sectors that have to deal with customer risk.