I think having a version with a sample synthetic dataset makes sense.