Demographics: Size & Shape

Everyone knows I love demographics. Understanding sizing and shape trends and anomalies is the key to creating a succesful ready-to-wear range that actually sells.  All too often design houses use hand me down standard size tables with no real appreciation of their target audience. I often use examples like a size 10 teenager being a completely different shape to a size 10 senior citizen, or the variation in shape increasing disproportionately with increasing size interval to demonstrate how carefully you need to examine for whom you’re designing.

Over the last few years on my site I’ve been gradually accumulating the data collected in the Lazy Person’s Pattern section. Now this collection hasn’t been about identifying the size characteristics of a particular group by age, interest or whatever … it’s plain and simply a record of anyone visting the site who want’s a pattern made by my server. Of the well over 35,000 measurement sets entered so far, many have been rejected by the statistical filter system we use to remove perculiarities (people experimenting with the system, men trying to use the womens pattern, etc) leaving me with around 27,000 results that I can be fairly sure are representative of adult women typing in what they really believe they measure.  The filter (designed by a statitics major and myself) initially runs over 50 comparitive tests to analyse each measurement set for anomolies like arms being longer than legs, arms being proportionately too long compared to legs outside a number of deviations, underbust being greater than bust or very large underbusts with ever so slightly larger bust (yes guys we know who you are) when compared to small waists and even smaller hips, etc. It’s very sophisticated but still runs the risk of eliminating real data believing it to be erroneous. The system then runs more statistical tests on the kept data and works out means and deviations … the previously rejected data is then retested against the deviations and either accepted back or rejected forever. Sounds very complicated, but what we’re left with is a huge collection of measurement sets that we can be sure are safe to use in our demographics studies … a collection that I should point out is the single largest private (non-govt) study of female size demographics in world history!! The only proviso to this is that the collection does not include the much larger body sizes which make up 0.05% of the entire population (my historical understanding based on sales figures and can be confirmed by statistical prediction models).  I’m sure some people will argue that some of my filters are unreasonable as well, but they need to understand that only tiny percentages are disqualified while leaving the filter open could potentially cause significant contamination by false data. None of the filters prevent people from getting patterns made by the server … they just decide (twice a week) what data gets kept by the server and what gets rejected.

The main problem I believe my system has is that it’s only as accurate as each person’s ability to take measurements. This is very significant in my opinion. I’ve seen a group of students each measure the same person and come up with wildly varying results. I’ve also had a big problem with metric versus imperial. Most of the world uses the metric system, but (mostly) Americans still haven’t found the 21st century yet as far as measuring goes. The problem comes when they try to convert their measurements by multiplying by either 2, 2.4, 2.5 or 2.54 (yes I’ve had people tell me all of these!!) or worst of all they just enter their measurements in inches and expect the system to work it out, which it doesn’t of course.  Unfortunately there’s not much I can do to solve this problem.  People also tend to round to the nearest half inch when using imperial which can cause a much bigger error than when rounding to the nearest half centimeter (think about it!).

A perfect study would use a fixed team of measurers trained in a particular technique and who can be tested to produce consistent results. I can’t do this via the website.

So why am I bringing all this up? Because I thought it would be interesting to actually do a study of measurements in a comparative manner in the hope of trying to  identify shape variations across the size categories … with the eventual aim of trying to find a better system for ready-to-wear sizing. An example might be comparing nape to waist against arm length … is this a constant value, a small envelope or a wide envelope? Does the golden 1.61803399 ratio really apply to leg length verses arm length? Are there constant relationships between other measurements? Most of all, can specific shapes be identified by specific comparitive values … eg; is there a direct relationship between back width and underbust when it comes to a degree of stooping and could the ratio/constant be applied to a new patternmaking rule?

You see, many people find one pattern making system (block/sloper drafting technique) works better than another. This may well be because the system was developed around a limited sample size, hence specific body shape (mine certainly was in the early days). By finding a relationship between comparitive measurements, hence shape and then ultimately system, we may well be able to create an all encompassing system … nice dream huh? Maybe it’s the fever I’m running again or maybe I’m having an epiphany (wiki: the sudden realization or comprehension of the essence or meaning of something). 

I’d love to hear peoples comments on this one. There’s so much to comment on really: ready to wear sizing problems, vanity sizing, pattern making systems, observations, measuring problems … the list could go on forever, so step up people and have your voice heard …

Ok maybe it’s the fever