<p>This paper analyses the performance of master’s graduates in Science, Technology, Engineering, Mathematics (STEM) disciplines in Italy. Data are from the Almalaurea Survey on Graduates’ Employment status in 2022. The analysis considers nine indicators that account for graduates’ academic performance, employment characteristics, and satisfaction with the degree programme and their current work. STEM as a broad fields of study includes 41 master’s degree classifications that correspond to over 600 degree programme. Given the nature of the problem, this paper proposes a new variant of the well-known sorting method AHPSort II, a multi-criteria method that, among those proposed in the literature, allows handling a large number of alternatives and criteria. In particular, a new methodological approach to determine the central (or limiting) profiles (traditionally defined ex ante by decision makers) using thresholds defined as a function of quantiles is introduced; this prevents arbitrariness which represents the main weakness of the AHPSortII method. To validate the new approach, we also test automatic classification algorithms, and which fall within machine learning techniques, which allow the definition of classes downstream of the method’s structuring (ex post). The paper includes a first phase in which the information on individual master’s programme is analysed and synthesised to identify any similarities among STEM disciplines and/or associations among the indicators examined. The results obtained by applying Principal Component Analysis (PCA) show that degree programme with similar characteristics define groups mostly coinciding with the four STEM fields of study. In the second phase, the processed information is then automated through the use of the new quantiles-AHPSort II (q-AHPSort II). The results obtained are compared with those of the classic AHPSort II and validated with machine learning techniques.</p>