The Number Nobody Likes to Say Out Loud

There's a number in direct-to-consumer fashion that nobody likes to say out loud.

Depending on the category, somewhere between 20% and 40% of everything sold online comes back. Not because the product was bad. Not because shipping was slow. Because the customer couldn't try it on, guessed, and guessed wrong.

That single unsolved problem is the tax every online fashion brand pays for existing. It kills margin on the back end, and long before that, it kills conversion on the front end, because a meaningful percentage of people who load a product page never buy anything at all. They hesitate. They think I don't know if this would look right on me. They close the tab.

Every discount code, every free returns policy, every "shop risk-free" banner is a brand paying money to work around a problem instead of solving it.

So I spent a few months solving it.

The Brand

TopFoxx is a women-owned luxury eyewear label out of Brooklyn. Tangle-free aviators, plant-based frames, sunglasses that float. Real traction, real celebrity placement, a genuinely distinct point of view in a category that's mostly noise.

I'd been watching the brand for about a year, and what struck me wasn't the product. It was that the product deserved a better digital experience than the category standard, and the category standard is genuinely bad. Luxury eyewear online mostly looks like a catalog with a gold font.

So I built a complete AI-powered virtual try-on application for the brand and put it live at topfoxx.app. I've since written up the full technical and strategic breakdown of the build at proscris.com/apps/topfoxx, but the short version is what follows.

The Core Mechanic: See It on Your Actual Face

The center of the app is simple to describe and hard to execute. A customer uploads a selfie. AI renders any frame in the catalog onto their own face, realistically, in seconds.

Not a cartoon overlay. Not an AR filter that slides around when you tilt your head. A generated image of them, wearing that specific product.

The commercial logic is straightforward. Uncertainty is the enemy of conversion. If someone can see the product on themselves before they pay, hesitation drops at checkout and returns drop after it. Both of those are margin, and margin is the whole game in fashion.

But the try-on turned out to be the least interesting part of what I built.

The Stylist Who Lives in the App

The brand's founder, Helen, has a signature on-camera persona with a distinct Russian accent. It's memorable, it's specific, and it's genuinely funny. It's also the single strongest brand asset the company has, and it existed nowhere in the shopping experience.

So I built Natasha.

Natasha is an AI stylist chatbot embedded in the app. She reads your face shape, asks what you're going for, and tells you what actually works. The version I want to ship next has a toggle: standard mode, or full Russian fashionista mode, where she teases you, grills your choices, and tells you the truth about the frames you're considering.

That sounds like a novelty. It isn't. It's the entire difference between a tool and an experience.

Nobody screenshots a product page. People absolutely screenshot an AI stylist roasting their sunglasses selection in a Russian accent. That's not a feature, that's distribution, and it costs nothing per impression.

The Part Most Brands Get Wrong: It's a Journey, Not a Page

Here's the thesis underneath everything I built.

Ecommerce is no longer retail. It's media. And a media business is measured by engagement, retention, and return visits, not by transactions.

Which means your website can no longer behave like a website. It has to behave like an application.

So the app isn't a collection of pages. It's a sequence, and every step is engineered to pull the user one step deeper:

Land → try on → get styled → take the quiz → share → buy → come back.

The Style Quiz recommends products and auto-applies a discount to the bag, which lifts conversion. But the more valuable output is data. Every answer is a first-party signal about face shape, style preference, and buying intent. That's a segmentation and retargeting asset that most brands would pay a fortune to acquire and instead throw away.

The sharing mechanic rewards customers with free shipping when they share their try-on. That converts existing customers into an acquisition channel and drags blended CAC down.

The whole thing installs as a Progressive Web App, so it sits on the customer's home screen like a native app. That's the difference between paying to re-acquire someone and simply being present when they think of you.

Every one of those steps raises engagement. Engagement is what actually moves conversion rate. And conversion rate compounds into revenue per visitor, which is a far better business to be in than the endless treadmill of buying more visitors.

Why It Doesn't Look Like a Shopify Store

The app runs headless. It pulls product data from Shopify's API, presents it inside a completely custom interface, and hands off to Shopify checkout only at the moment of purchase.

Shopify keeps doing what Shopify is genuinely excellent at: inventory, orders, payments, fulfillment. Everything the customer actually sees and feels is custom.

This matters more than it sounds like it should. Almost every store in the fashion and accessories space carries the same visual signatures. The same grid. The same theme patterns. The same tells. Customers register these subconsciously, and what they register is generic.

Strip those signals out and a brand immediately reads as a different class of business than everything it competes against. That perception is what allows a brand to hold its pricing instead of training its customers to wait for the next 20% off email.

You cannot build a luxury brand on a template that ten thousand dropshippers are also using.

Eyewear Is the Doorway, Not the Destination

This is the part I find genuinely exciting, and it's bigger than sunglasses.

Eyewear occupies a strange and privileged position in fashion. It's the most visible accessory a person wears, it sits directly on the face, and it's the single fastest signal of taste that exists. It's also the lowest-friction luxury purchase in the entire category. Nobody needs to know their size. Nobody needs a fitting room.

Which makes it the perfect entry point into a customer's entire wardrobe.

Once you have a customer's face, their style preferences, their quiz responses, and their purchase history, you're no longer running an eyewear store. You're running a personal styling platform that happens to have started with eyewear.

The natural evolution of what I built is a full AI fashion board. Customers build looks, save outfits, mix categories, and get styled across their whole wardrobe by an AI that already knows their face and their taste. Eyewear becomes the first product a brand sells someone, not the last.

Every fashion house that survives the next decade will make this transition. The ones that own the styling layer will sell the entire outfit. The ones that don't will sell one item and lose the customer to whoever does.

And there's a hardware chapter coming right behind it. Smart eyewear with embedded AI and audio is going to be a genuinely enormous category within a few years. The brands positioned early, with an existing AI experience and an established eyewear identity, will be in a dramatically better position than the tech companies trying to convince people that their gadget is also a fashion object.

Eyewear brands have the harder half of that problem already solved. They just need to move before the window closes.

What I Actually Learned

Three things, and they apply to any DTC brand reading this.

Uncertainty, Not Price, Is the Real Conversion Problem

Most brands discount their way around a problem that AI can now genuinely solve.

Brand Personality Belongs Inside the Product, Not Just on Social

Natasha exists because the founder's on-camera character was doing enormous work on social and zero work in the shopping funnel. That's a wasted asset in most companies.

Build It Before You're Asked

I could have sent a pitch deck. Instead I sent a working application. The gap between those two things is the entire difference between being considered and being taken seriously.