Understanding the contemporary stress state in rock volumes is crucial for applications such as reservoir management, geothermal energy, and underground storage. Geomechanical-numerical modelling, which predicts the 3D stress state based on geological structures, density distributions, and elastic properties, requires calibration using stress magnitude data records acquired in-situ. However, these data records can include outliers—stress measurements significantly deviating from expected values due to errors or localized geological anomalies. These outliers can skew model calibrations, leading to inaccurate predictions of boundary conditions and stress magnitudes, particularly in sets with limited numbers of data records. A systematic approach to identifying and handling outliers is essential to mitigate inaccuracies. The Python-based script DOuGLAS (Detection of Outliers in Geomechanics using Linear-elastic Assumption and Statistics) was developed to address this challenge. The software is part of the FAST (Fast Automatic Stress Tensor) suite of programs. Its function is to identify outliers in sets of stress magnitude data records by assessing the respective impact of individual data records on boundary condition predictions, using iterative combinations of data records. Results are analysed through dimensionality reduction and statistical scoring, providing visual and quantitative tools for outlier detection. The script aids users in improving model reliability by identifying and addressing anomalous data. It supports sets of different numbers of stress magnitude data records and integrates seamlessly with tools such as Tecplot 360 EX and GeoStress. This manual provides a comprehensive guide for using DOuGLAS, interpreting its outputs, and understanding its application in geomechanical modeling.