To create realistic benchmark tasks it’s important to understand main characteristics of real data batch effects. This is why we characterised batch effects from 9 different real data batch scenarios with focus on batch strength, celltype-specificity and batch type. We used these insights to (simulate realistic batch effects)[https://almutlue.github.io/batch_snakemake/simulation_index.html], with celltype specific logFC distributions, while retaining covariance structure between them.