This similar clothing recommender seems like something that Netflix or Amazon Prime will add to their pause screens on their original content - letting viewers purchase similar things to what the characters are wearing right on screen. I remember searching for years for some dumb shirt that Owen Wilson wore in some dumb movie, but I was so happy when I finally found something like it.
Not surprising at all! Instagram is probably next for this, then. Every single non-ad post could be a way to sell you the clothes or widgets that your friends are wearing/using in their fabulous lives.
A 4X dress on Amazon will often wind up being the equivalent of an American medium because it's from a Chinese vendor. It's baffling to me that Amazon doesn't make their clothing vendors enter a basic bust/waist/hip set of measurements and let people search by those.
I worked on this problem a few gigs back at Polyvore! tl;dr: It's nearly impossible to solve.
As you may already know, "Standard Sizing" is a misnomer. There is no single standard for a size. US vs EU vs JP vs other are all the same number (size 4) but different nominal sizes.
Add to that the chaos of standard sizes between brands (brand A size 4 != brand B size 4 != brand C size 4) and even within the same brand (brand A product 1 size 4 != brand A product 2 size 4) and you find that there's no single answer to matching a human to a size.
Even your suggestion of measurements won't help without more data around it. Clothing shape and fabric type matter nearly as much as the measurements, since one may be a form-fitting 90cm and another may be a breezy and stretchy 90cm. Personal preference matters as well, so you may prefer breezy while I prefer form-fitting.
So I get that it's likely to be impossible to fully solve anytime soon, but I'd think at least requiring a couple of basic measurements would make a dent.
I've seen exactly what you describe - sizes varying between countries, between brands, between individual items in a brand. A lot of the listings on Amazon include a custom size chart now, but it's not semantic data - it's just a screenshot in the product listing.
This one lists a XXXX-Large as a 34" waist, and a Large as a 28" waist. It shows up if you search Amazon for "plus size skirt"; if Amazon made vendors put in the waist measurements and let them be filterable based on that, someone in a US size 18 dress wouldn't be offered this as an option.
Again, I get the difference between a 90cm loose fit and a 90cm tight fit. I should be able to tell a major ecommerce site "look, I'm 150cm... fuck off with the 90cm stuff".
That would be great to get as a measurement and filter by. It still won't solve the problem you want it to.
This is a product with 1 measurement, maybe 1.5 if you count length. Due to flare, waist position, ratio to waist, etc. the length is subjective even with an absolute value assigned. Providing those measurements in this one case will help, but then you have to deal with "skirts" as a whole category with many sub-categories. And your measurements need to apply universally, so you need style, waist, waist position, length, flare, hobble and more. Broad categories for any of them won't help you, so "> 150cm" is not going to help beyond an existing "plus size" filter. Where the skirt sits matters too, so 150cm is fine for https://amazon.com/dp/B01G0YFAWM/ but not for https://amazon.com/dp/B07HB2NXPY/ .
https://amazon.com/dp/B01KXBI2VE/#g2s2SizeChartView_15496401... shows they have the data in some cases. There's a proper HTML chart with all the measurements as "Size Chart" in the pictures section. I would love infinite knobs to filter on every aspect, but most people (as proved per $ spent) prefer fewer simpler options for how to find products.
Outdoor clothier PrAna has already done this. They ask for your height, weight, age, and what size of what brand fits you the best to predict best match. ML at work.
None of it works. I'm in the fashion e-com space and returns/fit is a serious issue, but no one has really made any progress when ordering online. Even services like TrueFit/FitFinder barely make a dent.
Do you think if cameras can capture real depth values, or capture volumetric information, could you at least prevent shoppers buying clothes that will not likely fit. I don’t think personalizes fit would be solved, but if you could measure dimensions and include that in recommendation engines, it could provide more insight.
Still too many features to make a difference - all we end up doing is shifting one distribution for another, and the net result is the same number of unsatisfied customers but just for different reasons. Neck & shoulders, length of waist vs. legs, arm length or face shape. All of the tech I've seen in this space does great work for improving conversion rate - which I take advantage of myself - but overall there has been no noticeable impact on net sales.
I’m working with my team on a solution that might be slightly different from what you’ve seen thus far — check out Drapr.com or download our app on the App Store (iOS only for now).
Photorealistic, physically accurate virtual try-ons.
Wish there was some kind of international standard for sizing - where toddlers / kids / adults lived within the same namespace. As a recent parent I have been going crazy seeing the inconsitent sizing labels for toddlers / children.
An alternative interpretation of the data is that people are buying clothes on impulse, so it doesn't really matter if the clothes fit or not.
Surely the graveyard of doing-it-smarter retailers lend some credence to the theory. You're not the first person nor the last to want clothes to fit you well. But you just don't buy new clothes all the time like the impulsive crazies do, and they're the "whales" of the fashion industry.
Gulnaz here from Easysize. We use a range of factors (combined with ML) to recommend the right size and fit: from gathering product-specific post-purchase feedback, analyzing orders & returns and asking shoppers about their unique sizing challenges.
IMHO when it comes to recommending the right size at scale, the key challenges are:
1. taking into account fit and style: the same shoppers may like slim-fitted shirts and oversized sweaters.
2. keeping UX super simple: shoppers get bored and annoyed when they're asked too many questions or asked to perform some other actions.
3. taking into account unique product features: fabric, clothing cut etc.
Working in this area the biggest challenge is getting the details right. Small difference like cut-outs on a dress make a huge difference in terms of end-user relevance. Same for colors, millennial orange comes in many shades. Deep Learning is pretty bad with such details.
