Cytometry is an essential technique in the fields of immunology and oncology, allowing insights at the individual cell level at relatively low cost, complementary to genetics and other clinical parameters. Due to recent technological progress, researchers have started using more automated analyses. However, when applied to clinical studies, which typically include large amounts of samples, collected over larger time spans or in different centers, technical batch effects often impact the results of these models. While some normalization methods have been proposed, guidelines clarifying which one to use in which situation are still lacking. I will explore preprocessing techniques to remove batch effects before downstream analysis as well as modelling techniques which claim they can apply downstream analysis even when batch effects are present. Additionally, I will also focus on cases where it is not known upfront whether any batch effects are present at all and cases where a different set of antibodies was used when measuring the data. I will apply this on a range of datasets, including public data and data available through my collaborations with hospitals. For example, in a Primary Immune Deficiency (PID) study, more than 400 samples have been collected at UZ Gent over a period of 4 years and on two different machines. Correct integration of this data with minimal impact due to technical artefacts will lead to more trustworthy models which can be applied in broader settings.