A good rule of thumb is to always start with the simplest possible transformation and then move on to more complex ones if needed. For example, if you have a dataset with a lot of outliers, you might start by trying to remove them. If that doesn’t work, you could try transforming the data to make it more normally distributed. And so on.
