Orders processed per day.
Across all sellers on the platform.
Platform-wide volumeWe built a shared back office for sellers on Amazon, eBay, Etsy, eMAG and Allegro. Stock, orders, buyer messages, returns and payments are connected in one multi-tenant platform.
E-commerce seller platform
Client project, in production
Multi-tenant web application
4 months
from development to production
One physical stock pool, five APIs with different limits and delays. We brought inventory and the seller’s daily work into one product.
The integrations underneath.
The workspace on top.
Five marketplace integrationsOrders, listings, messages and returns behind a shared model.
A central inventory ledgerReservations, channel queues and a path for stock conflicts.
The seller’s daily workspaceOrder processing, AI-assisted replies, return rules and payment tasks.
The ledger holds the stock the platform can reserve. Marketplace listings receive updates through separate queues, so a central reservation and an external stock update are different steps.
Bring marketplace orders into one model, even when channel data and order states differ.
Record the order once and reserve its SKU’s warehouse stock in the same atomic operation. Two orders cannot reserve the ledger’s last unit.
Send the latest quantity through each channel’s own queue, at the pace its API allows. Zero-stock updates go before routine changes.
Zero-stock updates go ahead of routine changes in each channel’s queue.
The demo resolves it through another warehouse. An allowed backorder or cancellation are alternatives.
External APIs will not move in lockstep. Buffers, reconciliation and channel-specific connectors make those differences part of the product.
A stock buffer lets slower channels list a little less than the ledger holds.
Scheduled reconciliation compares channel listings with the ledger and corrects missed updates.
Connectors absorb each API’s rules. Sellers keep the same order statuses and actions.
One workspace connects the order, its deadline and the conversation around it. Operations and support work from the same context.
Shared statuses, dispatch deadlines, bulk actions and labels. Updates go back to the right marketplace.
Order details and past answers inform the draft. A support agent reviews, edits and sends the reply.
Each seller defines which returns qualify for an automatic refund. Other cases go to a person for review.
The example shows a return matching a seller’s rule, followed by a Stripe refund.
We integrated Stripe for the project’s payments, refunds, payouts and platform subscription billing.
Our development team built from inventory correctness outward: the ledger and connectors, then the workflows sellers use every day.
Inventory, marketplace integrations, order processing, buyer messaging, returns and payment tasks came together in the multi-tenant platform.
Java, Spring Boot and PostgreSQL underpin the platform, with Kafka and Redis. Docker and Kubernetes support the production deployment.
Results from the project account,
with seller outcomes in context.
Across all sellers on the platform.
Platform-wide volumeTime saved in processing and shipping an order.
Per seller; the upper reported figure.
Reported by some stores on the platform.
These figures describe the client project and reported seller outcomes.