In such a diverse context as Luxembourg, educational inequalities can arise from diverse languages spoken at home, a migration background, or a family’s socioeconomic status. This diversity leads to different preconditions for learning math and languages (e.g. the language of instruction) and thus shapes the school careers of students (Hadjar & Backes, 2021). The aim of the project Systematic Identification of High Value-Added in Educational Contexts (SIVA) was to answer the questions (1) what highly effective schools are doing “right” or differently and (2) what other schools can learn from them in alleviating inequalities. In collaboration with the Observatoire National de la Qualité Scolaire, we investigated the differences of schools with stable high value-added (VA) scores to those with stable medium or low VA scores from multiple perspectives. VA is a statistical regression method usually used to fairly estimate schools’ effectiveness considering diverse student backgrounds. First, we identified 16 schools which had a stable high, medium, or low VA scores over two years. Second, we collected data on their pedagogical strategies, student background, and school climate through questionnaires and classroom observations. Third, we matched our data to results from the Luxembourg School Monitoring Programme ÉpStan (LUCET, 2021). We selected the variables based on learning models focusing on aspects such as school organization or classroom management (e.g., Hattie, 2008; Helmke et al., 2008; Klieme et al., 2001). We further investigated specificities about the Luxembourgish school system, which are not represented in international school learning models (such as the division into two-year learning cycles, the multilingual school setting, or the diverse student population). We will discuss the SIVA-project, its goals, and its data collection leading to data from observations in 49 classroom and questionnaires with over 500 second graders, their parents, their teachers, as well as school presidents and regional ...