Every morning, five outfits on your own body, one swipe from the cart
scroll to read · skip to the designs
scroll to read · skip to the designs
A concept for Zalando: each morning the app builds five outfits on your own body, from one photo you give it once, every piece in stock in your size, then deals them out as a deck. Left to pass, right to like, up to send the whole look to your cart.
Browsing fashion online means scrolling photos of other people, and picturing the clothes on yourself is left entirely to you. When the guess goes wrong, the jacket goes back: "didn't look like I expected" sits reliably near the top of return reasons.
Five looks a day, rendered on your body, dealt as a deck and gone at midnight. A ritual, not a feed: finite, considered, cheap to render, and every swipe is a clean preference signal, the kind most retailers never get.
Handed to twelve people cold: all twelve worked out the three swipes on their own, eleven finished their first deck unprompted, nine superliked a look. Small numbers held loosely, but the gesture needed no manual, and that was the whole bet.
Zalando has one of the deepest catalogues in Europe, and browsing it still means scrolling photos of other people. How it looks on you is left to your imagination.
Returns tell the story: rates for clothes bought online are high, and "it didn't look like I expected" sits reliably near the top of the reasons. Virtual try-on exists, and almost every big retailer has piloted one, but always as a feature you go and find, buried on a product page. Nobody has made seeing yourself dressed the normal way to browse:
"A jacket. Looked amazing on the model. On me it just… wasn't that jacket. What would have changed my mind? Seeing it on my body. Not a size chart, not a 360 view of someone else. Just me."
Seeing yourself in the clothes is the marketing. The swipes underneath are the business.
The honest move was admitting which value is the real one. The loop underneath the renders: the app generates looks, you swipe, the model learns your shape, the colours you reach for, the brands you trust, what you'll pay and what you're dressing for, and tomorrow's looks get a little better. Five looks a day adds up to a couple of hundred small preference signals a week, first-party, gathered without anyone filling in a survey.
That decided the shape of the thing: a ritual, not a feed. A finite morning deck keeps generation costs sane, makes each swipe scarce enough to mean something, and gives the model a tidy daily set to learn from instead of endless scroll noise.
Noisy signals, and the render bill never stops.
Too rare to become a habit.
Scarce enough to mean something, frequent enough to become a ritual.
Why it worked: scarcity earns its keep here. A daily deck gives people a reason to come back, makes the swipes more considered, and keeps the cost predictable: five renders per person, generated and cached overnight.
the screen that can sink the whole thing
Nothing else matters if this screen fails. It has to sell what you get, say plainly what you give up, and make handing over your face feel safe, so it leads with the trade, puts the promises in the same visual weight as the pitch, and keeps the exit one tap away.
One photo. Five outfits a day, rendered on you, every item in stock in your size.
ADD YOUR PHOTO
✓Delete your photo anytime, in one tap
✓Never used to train on anyone else
✓Rendered privately, processed in the EU
CREATE MY MODEL
OPT-IN · ONE TAP TO LEAVE · GDPR-FIRST
The verb gives the user authorship over what happens to their face. The wording was written before the screen was drawn, because the wording is part of the consent.
Delete in one tap, never used to train on anyone else, rendered privately in the EU, all carried at the same visual weight as the pitch itself.
"Skip for now" keeps normal browsing one tap away. The model is an invitation; a wall here would sink the whole concept.
don't pitch a gesture, hand it over
A gesture you have to explain is already dead, so this page doesn't explain it. The phone below runs the same prototype the twelve testers held, coded with an AI assistant over a few evenings, which is the only reason a solo concept got this far. Drag the cards, or use the buttons and arrow keys; the superlike sits biggest because it's the one that makes money, and price, size and stock are answered on the card so you never go digging.
It runs here on a stand-in model; in the sessions, every tester swiped a deck rendered on their own body. When the fifth card is gone, the only ways forward are the cart or tomorrow.
the render gets the whole card, the interface never competes
AI imagery is unpredictable, so everything around it had to be rigid. One system, locked before a single screen was drawn: orange only for wanting something, green only for what you own, and the render gets the whole card.
1 / 5
‹
♥
RENDERED ON YOU · TODAY
THURSDAY'S LOOK · ALL 3 IN STOCK
HUGO slim blazerHUGO · EU 36€229
ARKET satin shirtARKET · SIZE S€89
Vagabond chelsea bootVAGABOND · EU 39€139ADD LOOK TO CART · €457
OR ADD ITEMS SEPARATELY
THU · LATE RESERVATION
SUPERLIKED ✓
HUGO slim blazerYOUR SIZE · IN STOCK€229
ARKET satin shirtYOUR SIZE · IN STOCK€89
Vagabond chelsea bootYOUR SIZE · IN STOCK€139KEEP SWIPING · 4 CARDS LEFT TODAY
CHECKOUT
PAYMENT AND RETURNS HANDLED BY ZALANDO
The render sold a look, not a list, so the button buys the look. Pieces separately stays one tap below.
Named, priced, in stock, in your size. No size chart, no guessing between W27 and W28.
Intent is captured the second someone loves a look, but no money moves until Zalando's checkout takes over.
Five cards, one body, gone at midnight.
A regret machine, and returns would spike.
Things go into wishlists and never come out.
Intent captured at the moment someone loves a look, and no money moves until checkout.
Why it worked: the cart keeps both the excitement and the brake. You capture the intent the second someone loves a look, but no money moves until checkout, and because the sizes come from the same profile that made the render, what's in the cart is ready to buy as-is.
Twelve people, cold, zero instructions. All twelve worked out the swipes. Everything past that is a projection, and it says so.
Eleven finished their first deck unprompted, in a median of about 74 seconds. Nine superliked a look. Twelve people is twelve people, so I read that as encouraging, not proof.
The projections stay labelled as projections: buying looks rather than single items should lift order value, and seeing the clothes on yourself should cut returns. The part I trust most isn't a number, it's the preference data the feature throws off just by being used.
"That's me in the clothes, not a model. I stopped reading the descriptions and just went on how it looked."
01
Hand over the gesture, don't pitch itThe prototype was the argument: twelve cold starts, zero instructions, twelve people swiping inside a minute. A deck you can hold beats a deck you describe, which is why it's embedded on this page instead of screenshotted.
02
Scarcity keeps the signal cleanFive considered swipes teach the model more than an hour of scroll noise. The midnight expiry is what made people read the deck as a ritual instead of a feed. The limit is the feature.
03
The render has to tell the truthA flattering render that lies about fit just moves the disappointment from the model photo to the mirror. Returns only drop if the picture tells the truth, and honest pixels are the whole retention story.
The deck is day one. The wardrobe is the roadmap.
Your closet joins the deck: green is already reserved in the system for what you own, so V2 styles new pieces with clothes already hanging in your wardrobe, and deals a "wear it tonight" deck that costs nothing to act on.
Decks that know your week: an interview on Thursday, a wedding in June, a run every Wednesday, so the morning five start dressing the life you're actually living.
Guest decks: a brand takes over Friday's five, labelled, skippable, still rendered on you. The ad format fashion keeps trying to invent, minus the guessing.
And underneath it all: the taste model keeps compounding, and the longer you swipe, the harder it is for any other store to know you this well.
Tomorrow at seven there'll be five new looks waiting, already in your size, already on you. All that's left is the swipe.