Most Windchill to ERP integrations work on the day they go live. The difference shows a year later: some run unnoticed, others need someone checking error lists every morning. In our experience the difference comes down to five decisions made early.
1. Decide who owns each attribute
Write down, attribute by attribute, which system is the master. Engineering data such as numbers, descriptions and BOM structure usually belongs to Windchill; plant, cost and procurement data belongs to ERP. When ownership is unclear, both systems end up editing the same field and the integration becomes a referee.
2. Transfer on release, not on save
Send data when a change is released, not every time an engineer saves. ERP should only ever see approved, consistent product definitions, and the release event is the natural boundary.
3. Make every message safe to repeat
Networks fail and messages get resent. If processing the same message twice creates a duplicate material or a second BOM, the integration will eventually break. Designing every operation so that repeating it changes nothing turns retries from a risk into a routine.
4. Plan for failure you can see
Some transfers will fail: a missing plant view, a locked material, a value ERP rejects. Failed messages should wait in a queue with a readable reason, and someone should be able to fix and resend them without a developer.
5. Reconcile on a schedule
Even a good integration drifts when people make manual corrections. A nightly job that compares released BOMs on both sides and reports differences finds problems while they are small, long before they reach the shop floor.
Connector or custom?
PTC’s standard connectors, an integration platform or direct web-service calls can all work. The right choice depends on what your company already runs and supports. The five decisions above matter more than the tool.