Traditional product selection methods – especially for high involvement products like refrigerators, cars and diamonds – use customer specified multi-attributes of the product to select products of interest to the customer. However, such methods tend to generate lot of false positives and false negatives due to conflicting, imprecise and non-commensurable nature of product attributes. In this paper, we present a novel methodology for product selection in Internet business to effectively handle the nebulous nature of product attributes. The system enhances the customer desired product attributes by utilizing his/her past profile, which is built by using his/her past purchases in the related product category. The suggested system offers the product variants as recommendation in a ranked order with customization to individual user’s needs. We experimentally evaluate the system on a real-life dataset in order to assess its potential usefulness. The methodology discussed here can be useful for consumers in making a better choice of the final product. In addition, it can also be useful for e-commerce managers in providing personalized services to their customers.