# Enterprise Marketplace Design: product design case study Taking Wholesale Seafood Buying Off the Phone. A marketplace for a seafood trade that still runs on phone calls. By Krishnachandran Ramachandran (Kaezee), end-to-end product designer. Page: https://kaezee.com/work/whalelord - **Product:** Whalelord - **Role:** Lead Product Designer - **Team:** 3 people - **Focus:** B2B Marketplace, Enterprise Workflows - **Timeline:** Ongoing - **Platform:** Desktop web app, B2B seafood marketplace - **Status:** Ongoing - **Domain:** B2B wholesale seafood trade **TL;DR.** US seafood is bought on trust: buyers stay with the suppliers they know, because quality is hard to judge and they sell it on to their own customers. Whalelord is a B2B marketplace where checked buyers ask checked suppliers for quotes, accept one to create an order, and pay through Whalelord. We started with a verified directory to test how many of these businesses would take their buying online at all, and moved payments in once most users preferred to pay on the spot. I designed the whole product end to end, and Compass, the design system behind it. The project is ongoing, and growing slowly, as trust markets do. --- ## A trade worth billions, run on phone calls and trust Seafood is one of the most traded foods on earth. The FAO expects global seafood exports of about $184B in 2025. The US alone imported over $25B of seafood a year from 2023 to 2025, and about four in five pounds eaten there are imported. Between the boat and the kitchen sit brokers, importers, distributors and wholesalers, and their roles overlap. For a buyer, that means a few suppliers with all the leverage, quality that is hard to check, and thousands of species to keep track of. For a supplier, finding new buyers is slow, and trust starts from zero every time. Tariff changes in 2025 (average duties on US seafood imports hit 8.43%) sent buyers looking for new sources. A tool for finding them was timely. **The brief.** Whalelord sits between seafood buyers and suppliers. Its marketplace had to feel trustworthy to a trade with good reason to be sceptical, and be simple enough for people who have never used a B2B platform. - Constraint 1 **The users** Experienced operators who are not comfortable with tech, next to a younger generation that expects consumer-grade speed. - Constraint 2 **The team** Three people in three countries. The developer needed a design system packaged to be used fast, so I built it and packaged it to be replicated one to one. - Constraint 3 **The starting point** A directory and a four-step quote flow already existed as a rough prototype. Good data, weak experience. **Who did what.** Our PM, in the US, gave us the brief and ran sales and the expo. Our developer, in Indonesia, built the product. I designed the whole product end to end, and Compass, its design system, from the UK. ### Why buyers stay with who they know US seafood buyers mostly buy from people they already know. They serve the fish or sell it on to their own customers, so a bad lot costs them their reputation, and the safest supplier is the one who delivered last time. Studies of wholesale fish markets find the same pattern. At New York's Fulton Fish Market most customers were repeat buyers, and every price was quoted privately. At Marseille's wholesale market, buyers stay loyal to their sellers. The caution is earned. Seafood is often not what the label says, and the checks only go so far. - **1 in 3** US restaurants and shops tested sold mislabelled fish - **½** Of US seafood imports covered by federal traceability - **11%** Of seafood firms get trade data straight into their systems Sources: Oceana, “Casting a Wider Net” (2019), DNA tests of 449 fish; NOAA Fisheries, Seafood Import Monitoring Program; Hardt, Flett & Howell, Journal of Food Science (2017), a survey of seafood companies in North America and the EU **Why they don't move online.** For a business that already has suppliers it trusts, going online solves a problem it doesn't feel. Research on seafood and food trading platforms keeps finding the same four reasons. - Friction **The handshake** Deals are made person to person, often by phone. A platform can look like it takes the relationship away. - Friction **Private prices** Each price is set for one customer. Listing prices invites comparison, and suppliers fear it squeezes their margins. - Friction **Quality you can't touch** Buyers judge fish by sight, smell and past lots. A photo and a spec sheet don't replace that. - Friction **New fees and paperwork** Commissions, onboarding and credit checks: new costs for a way of buying that already works. Sources: Graddy & Hall, NBER (2009), on the Fulton Fish Market; Kirman & Vriend (2001), on the Marseille fish market; Day, Fein & Ruppersberger, California Management Review (2003), on B2B exchanges; Eurofish Magazine (2024) The friction comes from how this trade works, since the same buyers already order much else online. Big foodservice distributors already sell mostly online: US Foods says 77% of its independent restaurant customers order that way. The gap is in specialty seafood, where each deal is priced privately, credit runs on relationships, and every lot is different. ### Why a directory came first The US has about 3,600 fish and seafood wholesalers with about $24B in revenue, around 4,800 seafood markets and about 17,300 sushi restaurants (IBISWorld). That is 25,000+ possible accounts before counting other seafood restaurants, hotels and caterers. Seafood Expo North America, held in Boston each March, is the biggest seafood trade show on the continent: about 22,000 seafood professionals on average, and more than 1,000 exhibitors from 48 countries in 2026. Our PM went to the 2026 show and made contact with 100 to 200 people there, most of them buyers rather than sellers. In a market this size, that doesn't scale by hand. And in a trade this loyal, the first thing to learn was whether these businesses would take their buying online at all. So the first product was a verified directory: the cheapest test of that question. From market to contacts the accounts out there, the people at the expo, and the ones we reached Possible US accounts 25,000+ Seafood Expo North America ~22,000 Contacts our PM made there 100–200 Sources: IBISWorld (wholesalers, seafood markets, sushi restaurants); Seafood Expo North America, 2026 (average attendance, exhibitors). The last bar is our own count. **Who else is out there.** Platforms sit at two ends. Listings directories connect buyers and suppliers and stay out of the deal. Sale-fee exchanges take a fee on each sale and bundle in shipping or trading. The biggest global B2B marketplace leads with verification: suppliers pay for a checked membership, buyers join free, and fees come on payment. Whalelord borrows that model and adds what seafood needs: catch data on every listing, and a quote before every order. Directory Transactional Light Integrated Listings directories Sale-fee exchanges Global B2B marketplace Whalelord #### Who the product touches, from the centre outward **Environment** US regulators and import rules, tariffs, competitors, seasonality **Enablers** Certification bodies (BAP, ASC, HACCP, BRC), freight and customs brokers, inspection providers, the payment provider **Direct users** Buyers (restaurants, distributors, wholesalers, importers) and suppliers (processors, exporters, wholesalers) **Core** The Whalelord product and team: verification, payments, the order record ## A marketplace you can trust with the whole deal Three of these decisions changed after the first version. The closed test runs through our PM, who holds its numbers, so each change is told as what we assumed, what we heard back, and what we changed. *Figure: Find: the marketplace, and a featured listing* #### Decision 1: The deal happens inside Whalelord: suppliers pay a plan, buyers pay a small fee on each payment. *The cost:* Money moving through the product raises the bar on trust. Every business is checked before it can quote, and the fee has to be shown at every step. What came back from the directory, through our PM, was a clear pattern: businesses would list and browse, and most of them preferred to pay on the spot, inside the deal, over settling another way. *Figure: The first model, what replaced it, and what that decided* *Figure: Paying a deposit: the fee on each payment, and a bank transfer matched by its reference* #### Who pays what: the plans Deposits and balances go through Whalelord, matched by a payment reference. The supplier receives the full agreed price, and the buyer's fee is added on top. *Figure: Pricing: buyers join free, suppliers pay a plan* #### Decision 2: Every order starts with a quote, and a repeat order comes with the quote already filled in. *The cost:* No instant checkout for a buyer in a hurry. Speed has to come from the request being filled in, and from how fast suppliers reply. In the old quote flow, buyers lost track of what they were quoting, and the round floating buttons were too small to hit. I rebuilt the request as four steps with the product in view, labelled Back and Continue buttons, shipping terms, and nothing sent before the last step. The supplier answers in the same thread, and accepting the quote creates the order. *Figure: The last step of the quote request: everything checked before anything is sent* *Figure: The supplier's side of a quote, then the quote as the buyer sees it* **Buying again.** Most of this trade is repeat business, so a repeat order had to be quick. *Figure: Order again: past products with their last amounts, sent as one quote request per supplier* #### The quick-order page it replaced, and the accept step *Figure: The quick-order page, what replaced it, and why* *Figure: Accepting a quote says what it does and shows the fee again* #### Decision 3: Trust at the top of every card: the verified supplier, then the catch data, then the action. *The cost:* Every card carries more frame than data, and a featured tier only means something because it is paid for. Buyers can't check quality from a screen, so the card has to say who stands behind it first. The old cards gave every field the same weight and buried trust. I split each card into three zones: trust at the top, data in the middle, the action at the bottom. *Figure: A featured and a standard listing card: trust first, then the catch data, then the action* *Figure: The marketplace and the quote request in three stages* #### The product page, ongoing checks, and the audit of the old prototype *Figure: The product page's buy panel: who sells it, what it roughly costs, what was checked* Checks keep going after sign-up. A new account starts in review, and buyers can browse but not ask for quotes until they are approved. *Figure: A document close to expiry raises a banner on the supplier's dashboard* The old directory had a solid structure and real trade data, but: - Cards showed exporter ID, origin, FAO zone, specs, certifications and product ID at the same weight - Descriptions cut off mid-sentence, which read as missing information - Buttons changed style from section to section - The filter pills looked dated - The four-step quote flow had abstract icons, tiny floating buttons and near-empty steps The sell form was the strongest part. Its toggle-able certification chips became the pattern for the buy side. #### Decision 4: Only claim what Whalelord knows: an order board the supplier moves by hand. *The cost:* The supplier does the moving, and a board left untouched goes stale. Automation is planned but not yet built. *Figure: The board's first idea, what replaced it, and what comes next* *Figure: The supplier's order board: cards move by hand, and a new order waits to be confirmed* #### One menu per role Buyers, suppliers and visitors have different jobs, so each gets its own menu. What a supplier does today and what happened last quarter are different needs, so Overview and Performance are separate tabs. *Figure: Each role gets only the menu its job needs* #### Decision 5: Compass, a design system packaged so the developer could replicate it one to one, fast. *The cost:* A system this size is a second product to maintain. Every rule in it is a call the team can no longer make on the fly. - **389** Components, in 38 sets - **556** Variables, in 13 collections - **66** Text styles - **2** Modes, light and dark Colours are named for the sea (Waves for action, Shores for surfaces, Depth for text, Reef for status and catch data), and each has one job. So does each typeface: DM Sans for the interface, Inter only for prices and numbers, Roboto Mono for IDs. The developer got a 997-line `design.md` and a prompt tuned for Stitch, with exact tokens, layouts and behaviour, so every screen could be rebuilt without guessing. *Figure: Compass colour jobs: each colour has one job* #### The first prototype that was rejected, and the rules it produced **The first prototype was rejected.** The sell and quote flows were missing, featured cards looked the same as standard ones, the page was white, and the cards were cramped. Each failure became a written rule. The page is never white: cards sit on a raised page. A missing flow is a blocking failure. Featured cards must look different. Tokens are used through variables, never as raw hex. Those rules now live in the handoff spec. - Compass tokens - design.md 997 lines - Stitch prompt - Built screens #### How a deal moves, front to back A deal has eight stages. Some move on an action in Whalelord, some when a payment arrives, and the rest when the supplier moves the card. *Figure: What moves each stage: an action, a payment received, or the supplier's board* *Figure: Service blueprint: buyer and supplier above the line, Whalelord below it* #### All the screens *Screen: Marketplace* *Screen: Product page* *Screen: A quote in Messages* *Screen: Order and payments* *Screen: Supplier overview* *Screen: Orders board* ## A slow-growth product, by design Trust markets move slowly. Buyers switch suppliers one lot at a time, so Whalelord was never going to grow like a consumer app. That is why the directory came first, and why progress is counted in checked accounts and first quotes. The project is ongoing. The marketplace is in a closed test, and the public site collects the businesses that will fill it. Its numbers sit with our PM, so none are quoted here: UNMEASURED at the time of writing. **The gaps, plainly.** There are no usage numbers yet. The research is secondhand, through our PM and published studies. The buyer fee and plan prices are samples. The board's automation is still to be built. **What we will track.** Five events define the funnel, and each one is a moment the product already records. - Quote request sent event 1 - Quote sent event 2 - Quote accepted event 3 · order created - Deposit received event 4 - Order again event 5 - Next · 1 **Grow the closed test** Turn sign-ups into checked accounts, live listings and quote requests. - Next · 2 **Set the real numbers** The buyer fee and the plan prices are samples until the pilot sets them. - Next · 3 **Automate the board** Move cards from payments, documents and buyer confirmations, keeping the manual override. ## Trust decides where the deal happens - Learned **Test the habit before the features** In a trade this loyal, the first question was whether anyone would buy online at all. A directory answered it cheaply. - Learned **Let users pull the money in** Payments moved inside once most users preferred to pay on the spot. The order became something both sides could see. - Learned **Repeat buying belongs where the orders are** A separate quick-order page caused visibility problems. Order again, on the Orders page, kept every deal on one path. - Next time **Start from the money** Where the money moves decided more screens than any other choice. I would settle it before the first prototype.