Appropriate spectral preprocessing can have a significant impact on the subsequently visualised data and output from more advanced techniques such as multivariate anlaysis and machine learning. The most commonly used preprocessing method, but also most contentious, is normalisation.