Ummm...

This is somewhat of a complicated question, but I'll go through the steps for you:

1. First I have to get the color information. Currently I am doing this by scanning the color chips in using what is more or less a normal flat bed scanner.
2. Once I have that data it is converted from the RGB gamut into the CIE Lab gamut via the XYZ gamut.
3. After the coor data has been translated into the Lab gamut, it is then stored in the database until needed.

That all is pretty normal stuff, but you then have to do the comparisons. Depending on how the source color is chosen one of two things happen:

1. If a color was provided, the program takes that RGB data and converts it to the CIE Lab colorspace just as above.
2. If a paint was selected the program will get it's information from the database.
3. The color information is compared to that within the database by effectively drawing an ovoid in the 3 dimensional colorspace and returning any colors which fall within that area. The specific size is dictated by the color range block of the web application.
4. Once that is done, I use another formula which double checks the results to remove any junk results.
5. Finally I return the closest six colors (assuming six colors exist) to be displayed.