Master Data Sync from ERP to E-Commerce: KIS Private Cloud to Jushuitan in Practice
What This Strategy Solves (Scenario & Value)
In one retail project, the customer’s master data for materials lived only in the ERP. When they expanded into the e-commerce side, the first wall they hit was: where do the product records come from? Material codes, names, specs, and units had been maintained in ERP for years, but the product records in Jushuitan were empty. The team relied on manual Excel imports every day, and after three months the two sides started drifting — codes misaligned, discontinued items still listed for sale, and new SKUs appearing late. This strategy turns that manual loop into an automated, traceable, replayable sync chain, all orchestrated by the Qeasy integration platform.
Data Flow & Field Mapping (Source → Middle Layer → Target)
The flow is one-way A → B: read materials from KIS Private Cloud, normalize codes, units, and categories in the middle layer, then write into Jushuitan products. The middle layer is provided by Qeasy by default; the field mapping is where the real work happens.
| Business Meaning | KIS Private Cloud (Source) | Jushuitan (Target) | Mapping Notes |
|---|---|---|---|
| Material code | FNumber / material code | Product code (i_id, business key) | Direct mapping; on duplicates, update instead of insert |
| Material name | FName | Product name | Truncate length, strip leading/trailing whitespace |
| Spec / model | FModel / spec | Spec | Append to the name to avoid length rejection |
| Base unit | FUnitName | Unit | Maintain a centralized unit dictionary; map source units to target units via lookup table |
| Category | FCategory | Product category | Flatten multi-level source path into a target code |
| Active / inactive | FUseStatus | Status | Filter inactive records in the middle layer, do not push downstream |
| Default warehouse | FDefaultStock | Default warehouse code | Reference only, not used for stock initialization |
Centralized mapping management is one of the most common patterns among Qeasy customers: keep all cross-system code, unit, and category mappings in one table so future changes happen in one place.
How to Configure This on Qeasy
On Qeasy, this strategy is implemented as a single sync task (Source: KIS Private Cloud, Target: Jushuitan). Key configuration points:
- Data source connection: KIS Private Cloud is read via DB view or API; Jushuitan is written via its OpenAPI. We recommend wrapping material reads into a view rather than touching base tables directly, to insulate against source schema changes.
- Extraction strategy: Default incremental extraction based on
FModifyTime; on initial onboarding, manually trigger a full sync to seed historical materials. - Transform & cleanse: Use Qeasy’s field-mapping component for header field alignment; apply string functions for name and spec; route inactive items through a filter branch and discard.
- Write strategy: Use idempotent semantics — update if the product code exists, insert if not — to avoid duplicate push generating junk products.
- Failure handling: Records that fail due to network blips or code conflicts go to an exception table; alerts are pushed to enterprise IM; engineers fix and replay.
Implementation Steps
Roll this out in three phases. Do not start with full + high frequency from day one:
Phase 1: Full initialization
- Manually trigger a one-time full sync to push all historical materials from KIS into Jushuitan.
- Use this phase to settle hard problems: field mapping, length truncation, unit dictionary. Production stays untouched.
Phase 2: Switch to incremental
- After full sync completes, use the full-sync end timestamp as the incremental start point; afterwards, push only changes.
- Recommend using last-modified timestamp + business key as the incremental combination to avoid missed pushes.
Phase 3: Schedule & monitor
- Master data does not change often; 15–30 minute intervals are usually enough and keep resource usage low.
- Configure scheduled tasks in Qeasy’s scheduler, retry 3 times on failure, route still-failing records to the exception queue.
- Header/body phased rollout is another typical pattern: if materials carry BOM or multi-unit bodies, sync header master first, handle bodies in a separate strategy later, reducing per-strategy complexity.
Pitfalls & Lessons Learned
- Code uniqueness not enforced: A typical mistake — KIS allows code reuse or trailing whitespace, pushing into Jushuitan creates many near-duplicate products. The safe approach is to add normalization in the middle layer, stripping spaces and zero-width characters uniformly.
- Discontinued items still pushed: If you blindly push inactive items, the downstream ends up with ghost products. This is where projects tend to crash; the safe approach is to add a status filter in the transform layer so inactive items never flow downstream.
- Unit dictionary scattered in code: Some projects bake unit conversion into scripts, so a change means hunting everywhere. The pattern Qeasy customers use is a centralized unit dictionary with single-point maintenance.
- Full vs. incremental confusion: Running full as incremental produces duplicates in Jushuitan. Split full-trigger and incremental-switch into two distinct actions, separated by a clear timestamp, so troubleshooting is easier later.
- Jushuitan spec field length overflow: KIS spec fields are often free-text and long; direct mapping gets truncated or rejected. The safe approach is to append to name or truncate explicitly.
When This Applies — And When It Doesn’t
Applies: KIS Private Cloud or similar ERP acting as the master data hub, syncing material master data one-way into an e-commerce middle platform or third-party system; low change frequency; minute-level latency is acceptable.
Does not apply: scenarios needing bi-directional sync with conflict merging; materials with complex BOMs, multi-org or multi-version data, or data that must pass workflow approval before landing; and scenarios requiring second-level real-time latency — those are better served by change-log + real-time push rather than scheduled sync.