The first measurement made (between the red and blue groups) shows a difference between the average wool quality of the two groups. Since the members were selected at random, you might conclude that the new shampoo is working because, what are the odds that you just picked all the good alpacas at random? That's what we're checking.

For several permutations of different groups (each group is a new random slection of alpacas so a mix of blue and red alpacas) take the same measurement (average of group a - average of group b)

Count the number of times that the difference was as good or better as your original measurement and you will find out the odds that being in your treatment group made a difference. (If random groupings show similar measurements then it means it's more likely that your specific treatment/control group did nothing).