Taking EDI Live with an Automotive OEM: A Project Checklist That Survives Reality
Every EDI project plan looks the same on a slide: requirements, mapping, testing, go-live. Real projects fail in the gaps between those boxes - the guideline revision that arrived mid-mapping, the label printer that prints one digit short, the test file that passed because the tester was the same person who wrote the mapping. This checklist is organised around what actually has to be true at each gate, with explicit sign-off evidence, so a status meeting can ask “show me” instead of “are we done?”.
Gate 0 - Scope freeze (week 0)
- Deliverable: the plant-specific message implementation guides for every message in scope, with version and date on the cover page, plus a written statement of which release each UNH-2 must carry (D:04A, D:07A, D:20B - they are not interchangeable).
- Evidence to demand: PDF filenames or portal links, logged in a document register. If the guide is “coming by email next week”, the project has not started.
- Common failure: mapping against the generic VDA recommendation while the OEM's own subset sits in a portal nobody was given access to.
Gate 1 - Codes and master data (week 1-2)
- Deliverable: the full code inventory in writing - your supplier/interchange IDs (UNB), NAD party codes for BY/SE/SU/SF/ST/DP, LOC unload-point and dock codes, contract references (RFF+CT), order-number ranges, and label number allocation rules.
- Evidence: a filled-in code sheet countersigned by the customer's EDI contact. Verbal codes are how trucks end up booked to plant 0011 instead of 0010.
- Also here: transport decided - OFTP2 direct (needs an OFTP server, certificates, a fixed IP or provider) or EDI service provider/VAN (needs the account and the routing table). This is the long-lead item of the whole project; start it in week 1 regardless of mapping progress.
Gate 2 - Mapping complete, self-tested (week 2-4)
- Deliverable: generator (4987 outbound) and consumer (4984 inbound) mappings, plus a structural validation pass on every produced file: envelope integrity, required segments, date formats, code-list values, control counts.
- Evidence: validation reports for at least three realistic files - one single-pallet shipment, one multi-CPS mixed load, one DELFOR with firm and forecast zones. Keep the files; they become the regression suite.
- Common failure: testing only the happy path from one ERP document. Real projects meet partial deliveries (QTY+11 splits), batches (PAC with batch numbers), and multiple ship-to locations in one interchange.
Gate 3 - Customer test phase (week 4-6)
- Deliverable: completed official test cycles on the customer's environment: test DESADVs matched against real delivery notes, test DELFORs consumed end-to-end into MRP, and - often forgotten - a physical label scan test: print the VDA 4902 goods tags from your label system and have the plant scan them against the GIN numbers in your test DESADV.
- Evidence: the customer's written test approval per message type. An oral “looks good” is not sign-off; the go-live blocker six weeks later will be the missing formal approval.
- Escalation trigger: if the customer's validator rejects a file that your local structural check passes, diff the two rule sets in writing - that gap list is your mapping errata for the rest of the project.
Gate 4 - Pilot (week 6-8)
- Deliverable: production traffic for one plant, one material group, with daily monitoring: every interchange tracked from generation to acknowledgement (CONTRL/APERAK where used), every rejection root-caused within 24 hours.
- Evidence: a two-week clean window - no structural rejections, no booking mismatches, cumulative figures reconciled daily.
- Common failure: declaring victory after three quiet days. The fourth day is when the weekly DELFOR with its long forecast horizon exposes the SCC double-counting bug that daily JIT files never triggered.
Gate 5 - Steady state and change management
- Deliverable: runbook (who watches rejections, who owns guideline updates), a regression suite of golden files re-validated after every ERP or mapping change, and a subscription to the customer's guideline-change notifications.
- Rule of thumb: OEMs revise guidelines quietly. The day your messages start failing with no code change on your side, the first question is “did their guideline change?” - and the second is “where is our versioned copy of the old one?” Keep both, forever.
The three timelines that actually slip
- Certificates and connectivity (OFTP2, AS2, provider onboarding) - external dependencies you cannot hurry by working weekends.
- Label printing alignment - the label system, not the EDI mapping, owns the GIN numbers; it is a different team, often with its own vendor and lead time.
- Test-file turnaround - each customer test cycle takes days of queue time on their side; build two extra cycles into the plan, because you will need them.
None of this is exotic - it is simply the discipline of evidence at every gate. The projects that slip are the ones where “done” meant “the mapper thinks it works”. The ones that ship on time can answer, for every gate: show me the file, show me the validation report, show me the written approval.
A note for small suppliers
If you are a tier-2 or tier-3 without a dedicated EDI team, compress - do not skip - the gates. Gates 0 and 1 become one phone call with your EDI provider who has done this OEM before and owns the code sheet. Gate 2 stays yours: run three realistic files through a structural validator and archive the reports. Gates 3 and 4 are unavoidable regardless of company size, but your provider usually fronts the customer test cycle, which removes most of the turnaround pain. The budget version of this checklist is still: pinned guideline version, written codes, three validated golden files, written customer approval, two clean pilot weeks. Everything else is optimisation.