Integration
How a purchase order travels, and where it gets stuck
The purchase order is green in the monitor. The supplier still cannot find it.
In short
A purchase order crosses five hand-offs before a supplier sees it, and the confirmation crosses them again on the way back. When someone says "the supplier never got the PO", don't open five consoles at once. Find the last place the document was seen, then look only at the next hand-off. With that habit, a two-day email chain becomes a ten-minute check.
The situation
Go-live week. A buyer releases a PO in the ERP, the supplier says nothing arrived, and three teams each say their part is fine. All three are often right, because the problem sits in the gap between their parts.
The five hand-offs
| # | Hand-off | Who usually owns it | Where to look |
|---|---|---|---|
| 1 | ERP decides to send the PO | Buyer functional team | The output / message status on the PO itself |
| 2 | ERP add-on sends it out | Basis + integration | The ERP's web-service or message monitor |
| 3 | Gateway maps and validates it | Integration team | The gateway's transaction tracking view |
| 4 | Network routes it to the supplier | Network admin (buyer side) | The buyer account's document status |
| 5 | Supplier sees it and replies | Supplier + enablement lead | The supplier's inbox, then the return path back through 4 → 3 → 2 → 1 |
What usually goes wrong
1. The PO never left the ERP. It's the most common cause and the least suspected. The output wasn't triggered for that supplier or document type. Often a new supplier was never flagged for network sending, or a condition record is missing. Everyone checks the gateway and finds nothing, because there was nothing to find.
2. It left, but the connection failed. An expired certificate, a proxy change, a locked technical user, the wrong endpoint after a system copy. These rarely affect one PO. They affect everything from that moment, so check the time of the first failure.
3. The gateway rejected it. It was mapped but failed validation: a unit of measure with no agreed code, a missing mandatory date, or a supplier ID the network doesn't recognise. The error text is usually precise. Read it slowly before escalating.
4. The network didn't route it. The trading relationship isn't active yet, or the supplier isn't set up to receive this document type, or the PO went to the test account when you meant production (or the other way round).
5. It arrived, but the reply bounced. The supplier confirmed, but the confirmation broke a buyer rule (a quantity above tolerance, a changed price, a line that doesn't exist) or failed on the way back into the ERP. Most teams test the outbound direction thoroughly and the return direction barely.
What we do
- Trace one document with one ID. Agree that the PO number plus its version is the tracking key on every screen and in every message. "PO 4500012345 v2 is at hand-off 3" ends the meeting. For hard cases, build a small correlation record: the ERP message ID, the gateway transaction ID and the network document ID for that one PO version.
- Don't trust one green light. A successful status at the gateway proves hand-off 3, not that the supplier received anything. Each hand-off needs its own evidence.
- Work outwards from the ERP. Start at hand-off 1 and move forward until the document disappears. Never start in the middle.
- Keep a tracer sheet during test and hypercare: one row per document, one column per hand-off, a timestamp in each cell. It doubles as your test evidence.
- Test changes and cancellations, not just creation. Every change makes a new version, and that is where mappings, tolerances and supplier habits break. Plan at least one change, one cancellation and one partial confirmation per scenario.
- Test the return path on day one. Send a confirmation and a ship notice back into the ERP early, before every outbound edge case is polished.
- Write down who owns each hand-off. A named person per row of the table above, with a backup. Unowned hand-offs are where documents wait for days.
Check it worked
- Pick three real POs from the last test cycle: one created, one changed, one cancelled.
- For each, you should be able to point to a timestamp at all five hand-offs and at the return trip.
- If any cell in your tracer sheet is "probably", it isn't tested yet.
Watch out
- Copies of systems. After an ERP or gateway system copy, endpoints and certificates often point at the wrong place. Put a connectivity check in the post-copy checklist.
- Test vs. production accounts. Make the account name visible in every test report screenshot, so nobody can mix them up.
- Batch jobs. Some hand-offs run on a schedule, not in real time. "Not there yet" may only mean "the job runs every 15 minutes". Find out the schedule before you escalate.
- Names and screens change by release. The gateway is now formally SAP Integration Suite, managed gateway for spend management and SAP Business Network, but many teams still call it CIG. Check SAP's current documentation for your landscape.