Syncing Online Sales Outbound Orders from Jushuitan to Kingdee Cloud Galaxy: A Single-Strategy Deep Dive
What This Strategy Solves
Pushing online sales outbound orders from an e-commerce ERP (Jushuitan) into a financial ERP (Kingdee Cloud Galaxy) looks like a simple document handover, but in one real retail project we saw the classic failure: three months in, the outbound quantities on both sides no longer matched, and the finance team ended up doing manual journal adjustments at month-end. The core problem this strategy addresses is landing the e-commerce-side outbound result into the finance side with a stable, well-defined contract, and keeping document status, inventory deductions, and accounting dimensions consistent across both systems. We use the Qeasy Data Integration Platform as the carrier, and its role here is intentionally narrow: a single-strategy, one-way sync that is observable and replayable.
Data Flow and Field Mapping
The overall flow is one-way: Jushuitan·Qimen (source) → Qeasy middle layer → Kingdee Cloud Galaxy (target). The middle layer is not just a pipe; it does three things: protocol adaptation (from Qimen API to Kingdee open platform), field reshaping, and code mapping.
Key field mapping (only the fields that actually cause trouble are listed):
| Business meaning | Jushuitan·Qimen source | Qeasy middle layer | Kingdee Cloud Galaxy target | Notes |
|---|---|---|---|---|
| Document number | io_id / io_no | strategy_biz_no (unified business key) | FBillNo | One unique key across systems |
| Store / channel | shop_id | channel_code | FHeadEntity.FShopNo | Build a channel dictionary in the middle layer |
| Outbound warehouse | warehouse | wh_code | FStockOrgId / FStockId | Branches into local vs. cloud warehouses |
| Material code | sku_id | master_sku_code | FMaterialId.FNumber | Strongly recommend unified code mapping |
| Quantity | qty | qty (4 decimals kept) | FQty | Do not round in the middle layer |
| Warehouse type branch | wms_type | local / cloud | Drives the local-warehouse vs. cloud-warehouse branch | This case is the cloud-warehouse distribution branch |
Note: the source side has multiple parallel strategies for local warehouses and cloud warehouses. If code mapping is scattered across each strategy, it will drift within three months. On Qeasy we centralize code mapping into a shared mapping table used by every sales outbound strategy.
How to Configure It on Qeasy
When configuring on the Qeasy Data Integration Platform, we usually build the strategy in four blocks:
- Datasource registration: register the Jushuitan·Qimen open interface and the Kingdee Cloud Galaxy open platform separately. In an on-premise deployment, pay attention to outbound direction and the callback allowlist.
- Strategy orchestration: create one SYNC strategy, with the source pointing to the Qimen outbound-order interface and the target pointing to the Kingdee sales-outbound save interface. Set sequence to A and flow_direction to A_TO_B.
- Field mapping and transformers: handle the header and the body in stages. The header does document-level transformations (store, warehouse, document date), and the body loops over material lines for code mapping and unit conversion.
- Exception handling: enable retry (3 attempts with exponential backoff is a sensible default), and persist the source document number, target document number, error code, and raw payload into logs so failed documents can be tracked down later.
Implementation Steps
The rollout rhythm we have battle-tested is staged:
- T-7: Centralize code mapping. On Qeasy, build and freeze the four master-data mappings (material, warehouse, store, customer), and disable ad-hoc mapping edits inside individual strategies.
- T-3: Confirm the incremental starting point. Align with the business on the cutoff date for incremental sync, and backfill everything before that cutoff with a one-off full load.
- T-1: Gray run. Run the single strategy on one store for three days, and reconcile quantities and amounts between both systems.
- T0: Go live on schedule. Poll the Qimen incremental interface every 5 minutes, and trigger only on document statuses such as shipped or outbound-confirmed. Do not push drafts to Kingdee.
- T+7: Full reconciliation. Run a full validation script that compares document numbers and quantities between both systems, and freeze it as the baseline.
The dual-track pattern (incremental for freshness, full reconciliation to fix drift) is a common playbook among Qeasy customers.
Lessons Learned
- Pitfall 1: scattered code mapping. The typical mistake is to maintain a mapping table inside every sales strategy. In one retail account, this produced inconsistent encodings for the same SKU under "Tmall warehouse" and "cloud distribution warehouse". The safe approach is to use Qeasy's centralized mapping table: every outbound strategy only references, never edits.
- Pitfall 2: submitting header and body as a single payload. Kingdee sales outbound orders have a header-plus-body structure; if the whole JSON is pushed at once and the field order is wrong, the entire document is rejected. Header-then-body staged processing is the more robust pattern.
- Pitfall 3: pushing unapproved documents. E-commerce-side outbound orders have intermediate statuses; pushing draft status to the finance side creates dangling documents. Add a status filter inside the strategy.
- Pitfall 4: ignoring time zones and date formats. Qimen returns timestamps in seconds; Kingdee expects a string date. The middle layer must format it explicitly.
- Pitfall 5: retries causing duplicate submissions on already-successful documents. Without an idempotency key, duplicates happen. Writing the Kingdee document number back into the middle layer and using it as the idempotency key is the safe way.
When This Applies and When It Does Not
Applies: one-way sync of e-commerce sales outbounds into a finance ERP, with stable document contracts, daily volumes from a few thousand to tens of thousands, and multiple warehouse branches (local and cloud). Does not apply: scenarios that require two-way collaboration, frequent document modifications and reversals, or source-side status rollbacks. Those are better served by an event-driven, bidirectional integration design.