This week Amazon took the wraps off a new incarnation of its Alexa voice assistant, giving the AI an eye so it can see as well as speak and hear. The Echo Look also contains a depth sensor that’s being used, in the first instance, to create a bokeh effect for a hands-free style selfies feature that Amazon is hoping will sell the device to fashion lovers, by making their outfits pop out against the bedroom wallpaper, and making them more eager to socially share.
The Echo Look app is where users can view the style selfies (and videos) they’ve asked Alexa to record for them (she indefinitely stores a copy for Amazon too). But the flagship feature of the app is a fashion feedback service, called Style Check, which Amazon says will utilize machine learning to rate fashion choices and help users choose between outfit pairs. And ultimately, presumably, give their entire wardrobe a score. Albeit, the feature is using (human) stylists too, at least for now, to help train what Amazon surely hopes will be entirely robotic style recommendations down the line.
The app will also suggest clothes for users to buy based on their style selections — opening up another revenue stream for Amazon, and one that could prove pretty sticky if Echo Look delivers on its promise of furnishing users with a personal stylist whose killer feature is the ability to shop tirelessly on your behalf. This new voice-controlled, Internet connected Echo camera is designed to condition users to feed it with the training data Amazon needs to build a fashion savvy AI. As data grabs go, it’s exceedingly well dressed.
As I wrote in July 2015, adding a camera to Echo makes perfect sense for Bezos’ massive fashion ambitions. With an eye to see you, Echo Look promises to contain your self-image better than a mirror by claiming to know which of your outfits is the fairest of them all. Fashion is often sold as something feel good and confidence building — a way to belong and blend in within a peer-group. But equally style can be deliberately different; the essence of individual self expression. So whether there’s an AI that can usefully cater to all those different facets remains to be seen. But for many shoppers the primary desire they have for the clothes they wear can be boiled down to looking good. So Amazon is positioning Alexa to sell that hope as a service.
Buying clothes is a recurring need; both a practical necessity and a way to keep up with changes in style and taste. Like buying groceries, it’s a type of shopping without end. Which is why Amazon is fixated on both spaces. “In order to be a $200bn company we’ve got to learn how to sell clothes and food,” Jeff Bezos said as long ago as a decade — displaying the long term thinking that has enabled the ecommerce giant to slow-grow its business over more than 20 years from an upstart online bookseller into today’s sprawling digital marketplace whose upwardly thrusting arrow declaims its mission to deliver everything.
From household staples to fashion destination?
Amazon Prime is the membership club that sells a subscription to convince people to lock themselves in to buying more and more from Amazon. Notably, a recent addition to the Prime perk list is an Amazon own brand men’s dress shirt brand, called Buttoned Down. Here the company is selling wardrobe staples that, if they bore a different label, would cost a whole lot more.
And while a fairly uniform garment like a dress shirt can make an easy recurring purchase, i.e. once you’ve figured out which size fits you, a lot of fashion is intentionally far less predictable. Meaning there’s a much greater need for style-related try ons. Female fashion especially falls into this category — hence Amazon heavily focusing the marketing for Echo Look on women…
Amazon’s vast strength is how its various parts link and pull together (like Prime) to build lock-in by embedding more and more utility into the core platform. So Echo Look — which at first glance might seem a bit of a niche product (at least from a male perspective) — is designed to lay important groundwork for Bezos’ big bet on fashion. The company has already launched a swathe of its own clothes brands in key markets. Now, giving Alexa the gift of sight, opens the opportunity for Amazon to get a far more intimate perspective on how its customers think about fashion as they try things on in the privacy of their bedroom — asking Alexa to be their fashion judge.
At the same time the Echo Look aims to shift how Amazon users experience fashion buying via the marketplace. This latest Alexa-enabled incarnation means Amazon is no longer primarily a vast, impersonal warehouse that has to be manually data-mined to unearth threads you want to buy; rather it becomes a style destination in its own right; an app that’s savvy about fashion trends and understands personal taste so it can do the leg work and shop for you. At least, that’s the pitch and the promise.
The ecommerce juggernaut that Bezos controls — Amazon’s market cap is now a staggering $439 billion — has not merely been growing in momentum and marketshare because of the huge inventory it aggregates and makes discoverable in myriad different ways. But, at least in key markets, because it has fixated on delivering convenience at faster and faster speeds — building distribution and shipping infrastructure that can all but eliminate the practical difference between buying online vs shopping in store. It’s not quite instant gratification but in some cases Amazon is pushing very close indeed — with one-hour delivery in some urban areas for a sub-set of products, and even delivery within 15 minutes via its experimental Prime Air drone trials (and even eyeing some crazier delivery ideas than those).
Amazon’s ecommerce imperative is to work towards excising the middle ground and going direct to the customer where the most money is to be made. In fashion’s case that not only means luring consumers away from retailers’ bricks and mortar stores but, increasingly, moving in on the fashion labels themselves. After all, many clothes aren’t bought principally or specifically for the label they bear, but for the utility or passing color they offer. Amazon wants — and clearly feels — it can dominate those types of clothing purchases. It has the scale, ambition, speed and data to cut itself a very sizable chunk of the fashion market, reckons Tom Adeyoola, founder of UK fashion tech firm Metail — which has been working on rethinking the online fashion buying experience with virtual fitting room and garment digitizing technology for almost 10 years.
Not just basics and low cost garments that might otherwise have been purchased at a supermarket — but High Street fashion, glossy catalogue brands, and other online clothes retailers. Amazon’s own brand ambitions on the fashion front look almost absolute. At this point its own brand fashion labels include: women’s fashion (Lark & Ro, Society New York) and accessories (North Eleven); children’s clothes (Scout + Ro); mens’ suits (Franklin Tailored) and dress shoes (Franklin & Freeman). Workout clothes are also apparently incoming.
“I think the industry has been very complacent because they’ve said nobody’s going to buy fashion on Amazon,” says Adeyoola. “You’ll buy cameras and stuff but you’ll not buy clothes. But they’re forgetting that Amazon thinks in five to seven year timeframes. They’ve got all the customer relationships, and they’re good at data, and they’re good at logistics, and they’re good at inventory management… Amazon went from zero [in clothes sales] to number one in the US in four years.”
It’s not hyperbole to say many fashion brands and retailers are facing a doomsday scenario if Bezos is able to realize the scale of his sartorial ambitions. After all, the difficulty of being a brand trying to make yourself heard and monetize on someone else’s platform is already writ large in the smartphone space, where big name brands are boxed into same-size apps, competing with each other for discoverability and diminishing attention returns while the platform master sits above the fray, controlling access and — crucially — knowing what the user really wants.
Adeyoola says the retail opportunity Amazon is closing in on is the huge wastefulness of the traditional supply chain, where healthy profit margins are squandered with poor inventory management — in turn a consequence of a failure to understand the transformative power of big data.
“In clothing… RRP, in general, the starting point of margin is towards 80 per cent. So manufacturing is only about 20 per cent of that cost. But all of the retailers are making around, at best, three per cent profit. You look at somebody like Asos in the last four or five years, they’ve more than doubled sales but their absolute profit number is the same. They’ve added sales for no profit. So Amazon can look at this and say: hold on, you’ve got 80 per cent profit and you waste it all — this is our opportunity. Talk to a manufacturer and they’ll say that inventory management and everything to do with data is where the retailers are just lazy. They’ve been lazy for too long. And that’s where Amazon is really good. So, in my mind, I look at it and say Amazon could double manufacturing cost, take their standard five per cent or less margin and still be half the price of everybody on the retail market.”
Amazon can look at this and say: hold on, you’ve got 80 per cent profit and you waste it all — this is our opportunity.
Even if Amazon’s consumer frontdoor has remained fairly consistent (at least until the original Echo popped up), Bezos has spend years honing backend infrastructure and tooling up supply chain expertise in the areas he wants to dominate — positioning Amazon to be able to offer a more compellingly priced product than high street fashion retailers and still make its margin. So while retailers continue to waste money on inventory management, Amazon is aiming to use data to eliminate inventory entirely.
Earlier this month, for example, it was awarded a patent for an on-demand clothes manufacturing warehouse that suggests an intent to push the boundaries of fast fashion even further. Amazon’s end game looks very much like garments made on-demand, locally at the point of order — vastly shrinking its warehousing and shipping logistics costs in the process. Amazon could seed data-fed, just-in-time manufacturing hubs in urban centers to service demand locally, enabled by knowing exactly what its customers want, argues Adeyoola.
“There’s just much less friction,” he says of the Amazon approach. “So sure they don’t have a great consumer user journey now, but they will do — and in the interim they’ve been doing, effectively, what we’re trying to do, which is digitize the world. They’ve been accumulating all the brands and all the clothes, and getting them onto the system and then learning and understanding where the white space is.”
He points to b2b apparel maker the TAL Group, which claims to make one in every six men’s shirts sold in the US. But the question is, for how much longer? While its button down dress shirts are priced by retailer partners at $80+, last month Amazon launched their own brand (the aforementioned Buttoned Down) — selling shirts starting at $40. That’s called disruption, Jeff Bezos style.
And while Amazon has been using data to optimize its supply chain for years, traditional fashion retail and brands are still saddled managing networks of bricks and mortar stores where they’re seeing falling footfall. These physical locations have arguably convinced retailers to view the Internet as just another sales channel, rather than the vital data pipe needed to overhaul all their business and supply chain processes in order to survive as mobile platforms consume their world.
Again, you could draw a parallel with a former smartphone giant like BlackBerry fixating on its physical Qwerty keyboard as a new generation of app-focused touchscreen devices swept in to change everything. Even if some fashion retail giants technically have the scale and resources to adapt to the big data era, none apparently has the long term conviction to take the plunge — which is enabling Amazon to push in and sew up marketshare.
“I think the trend is very straightforward,” says Adeyoola. “In the old days of retail, the battle was for footfall in a shopping mall, footfall on a high street. Now that traffic is time on a phone. So if you think about it in those terms, the channels where retail is going to happen are going to be in those places which have the most time on your phone. So all of the guys who take up the most time and have those customer relationships are going to be those new retail channels — so it is going to be Amazon, it has to be Facebook, it has to be Google, it has to be Apple. Those guys are going to be those portals through which you’re going to do retail, because that’s where you spend all your time. And all the retailers that have been spending their time building their own websites and building their own apps — you can’t fight… You will be like a shop in a shopping mall and the shopping mall is just going to be the Amazon app, the Facebook app. And hence the middle retailers… just won’t exist. You’re going to have to have a real battle to position yourselves in terms of what are you about.”
Plus, remember the Echo Look’s on-board depth sensor? Such hardware could be used to size up people’s full length selfies, enabling the AI to automatically know its owner’s size and recommend correctly sized clothes to buy. At scale, taking measurements from multiple Echo Look owners, Amazon would start to build its own dataset for size and fit — to use to further feed its clothes manufacturing efforts by enhancing and better customizing garment fit for its own fashion labels. Even — ultimately — to offer garments that are custom tailored to individual Amazon users, on demand but at fast fashion prices.
Adeyoola’s prediction is an entire disintermediation of the retail manufacturing supply chain. “A natural end state for me is that somebody like an Amazon can put manufacturing right next to its distribution hubs — all of its distribution hubs are optimized for urban centers, you can put manufacturing in those urban hubs,” he tells TechCrunch. “So if you get enough consumer data then you just create the marketplace for design and you basically make everything just in time, made to order. So you get rid of the inventory problem, you get the stuff made right by the distribution centers to go straight out. And if you had body size and shape data — that’s what we’re aiming to produce and deliver at scale — you could then basically shrink that time, you can also move towards an end state of made to measure, and you can deliver within a Prime window.”
If you get enough consumer data then you just create the marketplace for design and you basically make everything just in time, made to order.
Tailored fashion at scale is not all that’s potentially unlocked by the Echo Look’s depth and trend sensing eye. Another technology that could be delivered via this connected camera plus app set-up is virtual try ons, with the product becoming a trusted conduit for Amazon to ask for and receive full length body size measurements, captured from users’ hands free Echo Look selfies. With that data it could build an accurate 3D body model for each shopper, and — combined with an inventory of digitized garments — the Echo Look app then becomes a virtual changing room where users could play around trying garments on digitally before they buy.
Virtual try on could be an important piece for Amazon’s fashion ambitions because buying clothes isn’t always just about about fit — especially if you’re contemplating buying higher priced, more experimental fashion vs wardrobe staples that people can be comfortable buying without trying. Technology that enables consumers to judge whether a just encountered fashion style suits without having to physically pull it over your head is what virtual fitting room startups such as Metail have been working on for years; aiming to remove a last key differentiator for Internet shopping vs bricks and mortar clothes stores. It may not yet be mainstream — but the promise is clear.
And with the Echo Look, Amazon may have created the perfect arena to slot in this last piece to crown its fashion ambitions. Then, with a savvy AI to recommend styles, a virtual body double to envisage how potential purchases look, and local manufacturing that supports very fast shipping the ecommerce giant is in a position to convince users there’s no practical need to ever visit an actual clothes store.
“People are time sensitive, and we’ve learnt and been programmed now, through the likes of Uber etc, that getting something quickly… is possible,” adds Adeyoola. “With Amazon Prime Now you can order something and it be there within an hour. So then why do I need to go on a busy Tube train, travel for an hour to a high street and get hot and bothered?
“Back in the day going to the shopping mall used to be a leisure activity and fun. It’s just not that anymore. There are other ways to have fun.”
So when does he think Amazon could be in a position to deploy virtual fitting room technology? Adeyoola reckons it could bring something to market in two years — should the company decide it needs to push into the 3D creation space. For now, he reckons it’s not clear they are convinced they need it yet. But Amazon is certainly paving the way to acquire data that could power it. (On that front it also bought another 3D startup focused on fit for shoes, Shoefitr, back in 2015.)
“The approach that they’re taking with that selfie Look is a low risk one — which is basically to see whether people will buy the product and use it. And the way that the product is coming out is one which is a computer vision way of learning style, so they want people to take pictures of themselves in garments, and then they’ll try and use computer vision — well, they’ll going to use real stylists to start with to rate stuff — and then off the back of that work out whether they can basically build an algorithm which can say this is good, this is bad, and then start to use that as a recommendation flow longer term. Which is big, smart value.
“They haven’t thought about going into the 3D creation space thus far. And I think that’s because they say, well we’ve got datasets and we’ll use our datasets to improve as a starting point. So I haven’t seen evidence of them wanting to make that jump. And I think they’re probably thinking they’re best place to buy it when they feel it’s tipped over into being a need, rather than still in R&D phase. So, for a startup like us, having been invited to speak at their European partner conference… I think it was a means for them to test where we’re at — and see at what point they might need to either consume or crush.”
At the end of the day, if Amazon can deliver on a big data vision of custom tailored garments in fashion styles it already knows customers want — and ship orders to buyers within a matter of minutes — why then, the bedroom in your own home effectively is the changing room.