SQL is a central component of any database course. Despite the small number of SQL commands, students struggle to practice the concepts. To overcome this challenge, we developed an intelligent tutoring system (ITS) to guide the learning process with a small effort by the lecturer. Other systems often give only basic feedback (correct or incorrect) or require hundreds of instance specific rules defined by a lecturer. In contrast, our system can provide individual feedback based on a semi-automatically/intelligent growing pool of reference solutions, i.e., sensible approaches. Moreover, we introduced the concept of good and bad reference solutions. The system was developed and evaluated in three steps based on Design Science research guidelines. The results of the study demonstrate that providing multiple reference solutions are useful with the support of harmonization to provide individual and real-time feedback and thus improve the learning process for students.