SIFT
Streetwear shoppers could not find new brands, and new brands could not get found. I designed a discovery feed that solved both at once: no search, no filters, just an endless scroll of items from microbrands you had never heard of.
The Problem
Online streetwear had a discovery failure on both sides of the market. New brands had no organic path to an audience. The only reliable growth channel was social media, which meant running a clothing label was mostly running an Instagram account. Shoppers had the inverse problem: unlimited content, most of it irrelevant, and search tools that only worked if you already knew the brand name.
Both sides were failing at the same thing from opposite directions. That symmetry is what made a marketplace worth building.
Affinity mapping across 30 interviews. Brand owner concerns clustered around exposure and connection; shopper concerns clustered around irrelevance and search.
Research
We interviewed 30 people across both groups, testing each hypothesis as a working artifact rather than carrying assumptions forward.
Brand owners wanted exposure without becoming full-time content marketers, and more direct connection to the people buying from them. Shoppers described irrelevant content, indifference to mainstream trends, search that failed them, and no way to judge a garment through a screen.
Shoppers were also specific about price. They wanted it visible in the feed rather than one tap away, because an item outside their range made everything else about it irrelevant. Hiding price to drive engagement would have cost them time and cost us trust.
The two sides shared a constraint that shaped everything after: neither group would accept more work. Brand owners would not maintain another channel, and shoppers would not learn another search syntax. Any solution that added effort to either side would go unused.
That produced the problem statement we designed against. How might we help shoppers filter the clutter of online streetwear while helping brand owners market themselves organically?
Composite persona from the shopper interviews.
How shoppers browsed before SIFT. Every path started with a brand name they already knew.
Removing Search
My first instinct was better search. We prototyped several filter systems across category, price band, style tags, and brand size.
Testing killed them. Shoppers used the filters when handed them, but the behavior underneath contradicted the feature. Filters are a precision instrument and precision requires intent. These users had none. They were browsing, not searching.
We removed filters and tags entirely rather than refining them. SIFT stopped being a search tool and became a discovery feed.
The decision cost us the feature most marketplaces lead with, and it set the constraint the rest of the design had to satisfy. A feed with no filtering has to earn attention on every card, because there is no query to fall back on.
Revised flow after the pivot. Efficiency and simplicity replaced control as the guiding constraints.
The Product
A mobile-only web app. Brand owners submitted a product link and name through a form. Items entered a randomized catalog. Shoppers scrolled the feed, and tapping an item sent them to the brand’s own site.
Each card carried the product photo, item name, brand name, price, and controls to save or share. Nothing else. There was no cart, no checkout, and no account, because buying happened on the brand’s own site. The reference points were TikTok and Instagram rather than Grailed or SSENSE, because the interaction was browsing, not shopping.
Photo, item name, brand name, price, save and share. Tapping through exits to the brand's site.
Mobile-only
This followed from the research. Streetwear discovery already happened on phones, in the gaps of the day rather than at a desk. The feed was a touch interaction: thumb-driven, continuous, closer to an app than a website. Building for desktop would have meant translating a gesture into a mouse wheel and losing the quality that made it casual.
Randomized ordering
This is the decision I would defend hardest. Recency ordering means the newest submission always wins, which rewards whoever posts most often rather than whoever makes the best work, and buries every brand that submitted last week. Algorithmic ranking was not available to us: with 40 brands and no engagement history, there was nothing to rank on. Randomization solved both problems. Every brand got the same odds on every load, which was the only version of the feed that honored what we had told brand owners SIFT was for.
What we didn’t solve
We asked brand owners to submit at 9:16, which gave the feed a consistent frame to build against. We had no way to enforce it. Photography arrived as whatever the brand had shot, at whatever quality, and there was no review step and no quality floor. We worked on the assumption that brands would submit their best work, which was reasonable and which we never tested. There was also no cap on submissions, so a single brand uploading its full catalog could take a disproportionate share of a randomized feed. Neither problem surfaced at 40 brands. Both would have defined the product at 400.
Launch
We onboarded brands directly, marketed across social channels, and had a post trend on r/streetwearstartup.
The third figure was the one that mattered. It was the only direct evidence SIFT was doing what it existed to do.
Why It Closed
Our own traffic stayed flat, so we ran a growth push: paid social on TikTok and Instagram, a newsletter, and an interview series featuring brand owners on the platform. After a month it had not moved.
Running it made the problem legible. SIFT existed so brand owners would not have to build an audience through social media. Growing SIFT required building an audience through social media. We had designed a discovery product with no distribution of its own, and the only acquisition channel available was the exact work the product was meant to replace.
There was a second failure I weight more heavily now. Half the research concerned brand owners, and what brand owners received was a submission form. The shopper side was a designed product; the supply side was data entry. A marketplace where only one side has a designed experience is not two-sided, and that is part of why supply never converted into distribution.
We closed the marketplace after two months.
