Personalization in e-commerce has potentials to increase sales, customers' purchase intention and acquisition, as well as improvement of customer interaction. It is understood that personalization is a controllable variable for successful e-commerce. However, previous research on personalization proposed diverse concepts from numerous fields. As a result, it leads to bias construct of e-commerce personalization development and evaluation by academia and industry. To address this gap, a study was conducted to unravel personalization features from various perspectives. A Kitchenham's systematic literature review was used to discover personalization research from Q1/Q2 journals and top conference papers between 2012-2017. A theory-driven approach was administered to extract 21 selected papers. This process classifies personalization features into four dimensions based on three characters i.e objective, method, and user model. They include architectural, relational, instrumental and commercial dimensions. The results show that instrumental and commercial personalizations have been proved as the most popular dimension in the academic literature. However, relational personalization has been consistently rising as a new interesting topic to study since the massive growth of social media data.