As an abstract, functional description; Principal component analysis (PCA) is a statistical technique which aims to minimise the number of variables to describe the data set by taking advantage of correlations in the data. Eg if you have intensities for A, B and C, but A+B are correlated, you will get (A+B) as one of your principal components (PCs). This means you can now describe the data with two variables (PCs); (A+B) and C.

The idea is then to use these to separate different classes of sample (e.g. the different types of coffee). Say, group 1 is high in (A+B) and low in C, group 2 has high (A+B) and high C, and group 3 has low (A+B) and high C

By finding a minimum number of variables to describe as much of the data as possible, it becomes easier to spot the trends.

I’m not sure how easily you will understand the Wikipedia page, it is a little heavy on the maths, but I’m not convinced it (or I) explain the conceptul background well enough.

https://en.m.wikipedia.org/wiki/Principal_component_analysis