Some of the above comments are so true, as part of my research I have run an online body shape /clothing preference survey and these are so similar to the responses. My interest ultimately is custom made knitwear, either on a small scale, or in mass made to measure in some way.

At least there are quick fixes for those problems–for fitted tops, I am just SOL, because they’re always cut for DDs and I am a B cup. Sure, I may be an outlier, but I have met a number of women with similar issues. In fact, among people I know (not statistically significant, I realize, but still), I would say there’s approximately zero correlation between height and size.

The publications also include common size ranges (based on statistical studies), so for example a man with chest 104 (cm) might usually have a waist of 92 (a drop of -12 cm), but a range of other drop sizes (between -20 to +8) are also seen.

Not to quibble with your example but there is much less variation btwn men’s sizes than women’s. Men’s sizing has always more readily lent itself to standardization.

This European standard is based on labelling garments not with arbitrary sizes (size “2″ … what does a “2″ mean? different things depending on the target demographic), but with actual measurements on a language-neutral pictogram.

e.g. A garment would be labelled with:
Bust 94-98 (37-39 inches)
Hips 102-106 (40-42 inches)
Height 158-174 (5′2″-5′9″)

Note the measurements are in metric (which, incidentally, were initially adopted by the USA in the 1800’s).

There is also a short form of codes using the middle of the range for the primary measurement and letter codes for the others, e.g. 084DH would be a slender but tall size.

The use of actual measurements makes mail order and Internet ordering much easier, and using multiple measurements allows a much wider range of body types to be accommodated.

The publications also include common size ranges (based on statistical studies), so for example a man with chest 104 (cm) might usually have a waist of 92 (a drop of -12 cm), but a range of other drop sizes (between -20 to +8) are also seen.

Suppose you measure 10,000 or so women and put them into a program that lets you slice and dice data, say my favorite, Interactive Data Language (IDL). Then you can do simple regression analysis between measurements (say bust and back width). The slope would give a linear constant to improve grading.

I like to do a scatter plot of things that should be related. Often, clusters would appear. If you plot hips vs. waist, clusters might appear representing apples and pears. Do the regression analysis for each set of figure types separately. That gives you grading guidelines for different figure types.

Surely, someone has thought of this before. Has anyone done it?