This work describes a data cleaning procedure for the detection of eddy covariance fluxes affected by systematic errors. We believe that the proposed procedure can serve as a basis toward a unified quality control strategy suitable for the centralized data processing pipelines, where the use of completely data-driven and scalable procedures that guarantee high-quality standards and reproducibility of the released products constitutes an essential prerequisite.