Three weeks is enough time to draw a marketplace. It isn't enough time to draw every state a marketplace needs: every modal, every empty result, every "insufficient funds," on desktop and on mobile. The design had to be built in a way that made those states cheap.
Art first, data second
Marketplaces drift toward the spreadsheet: floor price, volume, 24h change, owners, supply. All of it matters to a trader, and all of it competes with the thing people came to look at.
So every page leads with the art and puts the numbers underneath. The home page opens on a carousel of featured collections, with the full collections table below it, sortable by trending. The NFT detail page gives the piece the whole left half, uncropped. Price, the buy and offer actions, traits, and contract details sit beside it rather than above it, with the item's open orders and trading history one tab away.
The marketplace home, and an NFT detail page with the art given the left half. (2 screens)DecisionWhere the filters went
The home page has no filter panel. It has the featured carousel, the collections table, and one sort control. Filtering by status, rarity, price, and traits lives inside a collection instead, where the collector already knows what they're looking at and a trait filter actually means something.
Search sits in the nav on every page, so someone who arrives knowing the name never has to browse at all.
Every collection tells its own story
Collections arrive with their own art direction, and a marketplace that flattens them into identical cards throws away the thing that makes each one worth buying. So each collection page opens on that collection's own banner art. Below it, the chrome stays constant: a stats strip (floor, top bid, volume, owners, listed supply) and the same four tabs, Items, Offers, Holders, Activity. That way you always know where you are.
The same idea carries into a collector's profile. Owned items show as art cards, and each card carries its own state (listed, delisted, or ready to list) instead of pushing that into a separate table. Offers split into the ones a collector has received and the ones they've made, each editable or cancellable from its row.
A collection page carried by its banner art, and a collector's inventory as art cards. (2 screens)StructureFour tabs, every collection
Every collection gets the same four tabs. Items is the grid or list of pieces, with filters. Offers shows open bids. Holders shows who owns it. Activity has a volume chart over the past year above the buy, sell, and offer history.
A collection can look like anything above the tabs. Below them, a collector who has learned one collection has learned them all.
One modal, every trade
Trading on Loomia goes well past "buy." Collectors can sweep a collection (pick a count and buy the cheapest listings in one go), place an offer on a whole collection, or offer on items with a specific trait. Owners can list, edit, and delist, and accept the offers that come in. A balance sheet in the nav swaps between STT and wrapped STT. Each of those needed its own confirmation, loading, error, and success states.
Rather than design each flow from scratch, every trade uses one modal anatomy: what you're trading, the floor and top offer for context, the one input that matters, a fee breakdown, and what you'll pay or receive. Buy, offer, trait offer, list, and edit all differ in the middle row and nowhere else. The failure states follow the same rule: "insufficient funds" takes over the action button, so the fix is always in the same place.
The shared trade modal, here as a collection offer in its insufficient-funds state.AnatomyWhat changes between trades, and what doesn't
Every trade modal is the same five rows: the item or collection; its floor, top offer, and last sale; the input; a duration; and a breakdown ending in the number you care about. Only the input row and the breakdown's wording change:
- Buy: the items, each removable, and a total in STT and dollars.
- Offer: a price and duration, with the offer's distance from the floor as a percentage.
- Trait offer: the same, plus the selected traits as removable chips.
- List: a price and duration, with royalty and taker fees and "You will receive."
Three weeks, one component set
Three weeks doesn't buy exploration. It buys one good decision made early and held: define the components before the screens, then compose rather than draw. On mobile, the same pieces follow a few rules instead of a redesign: trade modals become bottom sheets, and the sweep and buy actions stick to the bottom of the screen, within thumb reach.
NFT detail, a collection page, and the menu on mobile, from the same components.InsightWhat a three-week build actually cost
Every trade flow and every responsive breakpoint came out of one component set built in the first few days: the trade modal, cards, tables, the stats strip, and the collection header. Once those existed, most screens were an assembly job rather than a design job, which is the only reason the surface area fit the calendar.
What that bought: consistency for free, and a developer who could build ahead of me because the pieces stopped changing.
What it cost: no exploration of a second visual direction, and no usability testing before launch. On a longer engagement I'd have spent week two proving the art-first hierarchy with real collectors instead of arguing it from first principles.
Where it landed
The marketplace shipped on Somnia with the whole surface designed: browsing, collection pages, NFT detail, collector profiles, and the buy, sweep, offer, and listing flows, each with its error and success states, from desktop down to mobile.
