Integration Overview: Xiaoman OKKICRM and Guanjia Po ERP — BOM Materials to CRM Products
Scenario and Value
A manufacturer had Guanjia Po ERP running production and inventory, while its sales team used Xiaoman OKKICRM for quotations and contracts. The pain appeared quickly: sales built one set of product codes in Xiaoman while procurement maintained another in Guanjia Po. When a BOM parent item's spec changed, Xiaoman didn't notice until the next month-end stocktake — quotations were running on stale data, eroding contract margins.
The goal of this integration is straightforward: let CRM-side product master data have exactly one source of truth, and that source is Guanjia Po's BOM materials. BOM parent items (finished goods and semi-finished goods) flow incrementally into Xiaoman as sellable products. Xiaoman's product groups and Guanjia Po's product archives serve as two query caches, providing group_id lookups and code validation during sync.
After rollout: sales quotations align with production on spec, unit, and price; new or discontinued items take effect in CRM the same day; code mismatches that used to cause rework get caught earlier.
Integration Architecture and Data Flow
The whole integration runs on a privately deployed middleware layer that encapsulates scheduling, transformation, code mapping, incremental watermarks, exception retry, and cross-source lookup.
┌─────────────────────────────────────────────────────────────────────────────┐
│ Xiaoman OKKICRM │
│ Products | Product Groups │
└─────────────────────────────────────────────────────────────────────────────┘
▲ │
│ Product master (write) │ Product groups (query)
│ │
┌─────────────────────────────────────────────────────────────────────────────┐
│ Middleware (Qeasy Integration Platform) │
│ Schedule | Transform | Code Mapping | Watermark | Retry | Lookup │
└─────────────────────────────────────────────────────────────────────────────┘
▲ ▲
│ BOM parents (read) │ Product archive (query)
│ │
┌─────────────────────────────────────────────────────────────────────────────┐
│ Guanjia Po ERP │
│ BOM Materials (parents) | Product / Inventory Archive │
└─────────────────────────────────────────────────────────────────────────────┘
Data flow has two stages. Stage one runs two query strategies in parallel: it pulls Xiaoman product groups and Guanjia Po products into the middleware as lookup caches. Stage two runs the BOM→product sync strategy serially after stage one, depending on the group mapping and code validation those queries provide. Only BOM parents are taken (finished/semi-finished); child components and raw materials are filtered out to avoid polluting the CRM-side sellable product pool.
We use Qeasy Data Integration Platform (轻易云) as the middleware. Its appeal in a private deployment is that code mapping, watermarks, and exception retry come out of the box — no need to build an ETL framework from scratch.
Interface Inventory
| Strategy ID | Data Object | Direction | Notes |
|---|---|---|---|
| Q-01 | Xiaoman product groups (id, name, parent_id) | Xiaoman → Middleware | Query only; supplies group_id lookup |
| Q-02 | Guanjia Po products (product_code, product_name, category, etc.) | Guanjia Po → Middleware | Query only; supplies code validation and model/unit/price enrichment |
| S-01 | Guanjia Po BOM materials (pusercode, pfullname, bomtype) → Xiaoman products (product_no, name, cn_name, model, group_id, unit, price) | Guanjia Po → Xiaoman | Master data sync; parents only |
Implementation Notes
Layered scheduling. The two query strategies (Q-01, Q-02) run in parallel: Q-01 once per day, Q-02 every 15 minutes, keeping lookup caches fresh. The sync strategy (S-01) runs every 10 minutes, serially after stage one.
Incremental fields. On the Xiaoman side, pull by start_time/end_time window. On the Guanjia Po side, use modification timestamp or the API's incremental parameter. Each incremental window should overshoot the previous watermark by a small margin (a typical choice is 5 minutes extra) to avoid missing boundary records.
Full-load fallback. Every strategy supports manual full-load trigger for initialization or post-repair reconciliation. When triggering full load, pause the subsequent incremental runs, finish one pass, then resume — otherwise two sides will overwrite each other.
Code mapping. BOM material pusercode maps to Xiaoman product_no (key-to-key). Fields missing from the BOM — model, unit, price — get enriched via lookup against Guanjia Po's product master. bomtype looks up Xiaoman's product group name to resolve group_id.
Category mapping. Guanjia Po's category system and Xiaoman's product groups are not the same thing. A classic mistake is writing the category string directly as the group name. The robust approach is to maintain a category→group-name mapping table in the middleware, or fall back to a default group.
Exception retry. Transient errors (network blips, rate limits) retry with exponential backoff. Business errors (field overflow, missing required values) go to a dead-letter queue and wait for human review.
Privacy. Connection strings, accounts, and tokens live in environment variables — never in solution docs or script comments.
Best Practices and Pitfalls
-
Centralize code mapping; don't scatter it across strategies. All three strategies need the code cross-reference. If each strategy maintains its own copy, a single change forces a global search. In Qeasy we put the mapping table in the middleware so every strategy reads the same source — one edit, full propagation.
-
bomtype→group-name mapping needs fuzzy matching. In Guanjia Po the same category appears as "成品", "产成品", "finished goods", but Xiaoman's group is just "成品". Exact-match equality drops a batch; the symptom is some products sync with emptygroup_id. Configure a small alias table in the middleware. -
Be explicit about BOM parent judgment. A BOM parent in Guanjia Po can be a finished good or a semi-finished good. Semi-finished goods typically aren't sold directly. Recommendation: only sync parents whose
bomtypeis clearly marked as finished; either skip semi-finished or route them into a separate group. -
Stagger query schedules away from the sync run. If the two queries and the sync fire at the same instant, the lookup cache may serve stale data. Offset by 5 minutes and the conflict largely disappears.
When to Use Qeasy
When CRM and ERP come from different vendors, product master data must stay consistent across both, and the deployment is private — Qeasy Data Integration Platform (轻易云) fits the middleware role. It packages code mapping, watermarks, retry, and scheduling into ready capabilities. A small 3-strategy integration can be orchestrated and live within a day, and later mapping changes don't require rewriting scripts.