귤곁 · Gyulgyeot

The idea

Households buying everyday fruit often have to compare many listings with inconsistent descriptions. Meanwhile, a marketplace can create too much administrative work for small suppliers if stock, packing, support, and payments are not reduced to clear tasks.

귤곁 is a consumer-marketplace pilot for affordable, mixed-size Jeju mandarins. A customer chooses a box size and receives one recommendation with at most two alternatives. The recurring farmer experience centers on three tasks: update remaining boxes, prepare orders, and check payments.

What I built

The pilot has separate consumer, farmer, and operator experiences. It covers offer selection, reservations, checkout, order history, packing and shipment, receipt confirmation, customer cases with private images, refunds, stock management, and settlement records.

Core business changes use explicit PostgreSQL transactions and row locks. This includes races such as the final available box being purchased while stock changes, or a customer cancellation competing with the start of packing. A transactional outbox and background worker make provider-facing work durable and repeatable.

Authentication supports private customer accounts, staff roles, recovery codes, and an optional phone-verification path with SMS rate and budget limits. Uploaded support images are decoded and re-encoded, stripped of metadata, and protected by purchaser/operator authorization.

Product decisions

• Reduce choice instead of expanding a catalog. The interface recommends a small number of eligible offers using delivered price, availability, and dispatch constraints.
• Separate promises from assumptions. Weight, cosmetic variation, dispatch timing, and supply evidence are presented as distinct facts.
• Design around operational reality. Farmer, support, refund, and settlement workflows are part of the pilot rather than deferred back-office concerns.
• Make uncertainty visible. Unknown payment outcomes hold stock for reconciliation instead of pretending an operation succeeded or failed.

Current state

The public site is a private-account journey pilot. It uses fictional supply and simulated payments and shipping; it is not a commercial marketplace launch and has not validated demand or unit economics.

The implemented workflows have automated rule, provider-contract, PostgreSQL integration, and Chromium acceptance coverage. Before real orders, the project still requires real supplier onboarding, merchant and carrier arrangements, verified customer policies, provider acceptance, production recovery work, and evidence from actual fulfillment.

Still to add

This page will later explain why I explored this market, how the Korean customer experience evolved, what I learn from farmer and household interviews, and whether real-world economics support continuing the project.