Multi-channel inventory synchronisation

Overview

A technical reference architecture from Palmate Solutions detailing how to approach inventory synchronisation when a storefront, a warehouse tool, and a marketplace all need a shareable, reliable stock ledger.

Business problem

Stock was updated in three places: the website admin, a warehouse spreadsheet, and a marketplace seller central. Oversells appeared during promotions. Refunds and support load followed, but nobody could point to a single stock ledger.

Technical challenge

Each channel has a different idea of ‘available’: website available-to-sell, warehouse physical count, marketplace fulfilment lag. A naive two-way sync creates loops. Partial failures (marketplace rate limits, webhook retries) can resurrect old quantities.

Requirements

• One system is the stock authority for on-hand quantity
• Storefront and marketplace receive reserved/available figures, not a blind copy of warehouse count
• Order events decrement reservations immediately; receipts increment only after confirmation
• Operators can see sync lag and poisoned messages
• Idempotent webhooks so duplicate order events do not double-decrement

Solution

Introduce a small integration service that owns mappings and reservations. WooCommerce and the marketplace are clients of that service. The warehouse system remains the authority for physical stock. The service exposes an internal API for available-to-promise and consumes order/refund webhooks.

Architecture

Technology stack

• Node.js integration service
• PostgreSQL
• Redis or equivalent for short-lived reservations
• WooCommerce REST + webhooks
• Marketplace official APIs
• Docker Compose

Implementation approach

1. Map SKUs and packs (1 box = 6 units) before writing any sync code
2. Implement inbound order webhooks with idempotency keys
3. Ship a read-only available-to-promise endpoint and verify it against a week of historical orders
4. Turn on outbound marketplace updates for a subset of SKUs
5. Add alerting on lag and on persistent 4xx mapping errors

Outcome

In production, success looks like eliminating oversells, measurable sub-second sync lag, and an operator who can verify status without opening disparate vendor panels.