The PCA graph breaks out the variance in the data and we can see that maybe colors that are close on the RGB graph have more differences in the PCA graph. The PCA map would probably hold more value if the color data was collected by a spectrophotometer, but that would require the purchase of lipsticks and the preparation of the samples for the machine that could take weeks. The RGB data was collected by each companies' website and gives a good representation of the color space
PCA is about differences and HCA is about similarities. Clustering the data into similar groups gives a better representation of the market space by reducing the SKU's into groups that are too similar for a consumer to notice. Where the delineation occurs is defined by good consumer testing to determine the limits.
There are several ways to use this data to position new colors. The first is to use the HCA with total sales and find the money spots. This would require doing the same thing as other companies and would rely more on brand loyalty than on innovative new products. The second way is to use the PCA and place new colors in areas that are not covered by any competitor. This would require consumer tests to verify that large gaps in the color space are desirable for the consumer.
A better way would be to use the PCA and HCA in conjuction and find gaps in the color space that are between clusters of high volume sales. This would differentiate a new color from the rest and position it in a space of higher market probability.
Data was collected from the various websites of each company.