From VDA 4905 to VDA 4984: Migrating Off Legacy Fixed-Length Messages
Some automotive suppliers still run EDI flows on the classic fixed-length VDA messages: VDA 4905 (delivery note / despatch advice), VDA 4906 (invoice) and VDA 4913 (call-off / delivery schedule request). They work, they are battle-tested, and every old-timer on the shop floor can read one in a text editor. But when a new OEM onboarding project lands on your desk and the specification says "VDA 4984 DELFOR, UN D.04A or newer, per our implementation guideline" - or when an existing customer announces the switch to the Global messages - you are migrating. This guide walks the path from 4905-era fixed formats to the EDIFACT-based 49xx world, with the errors you will actually meet in test files.
Why OEMs push the Global messages
The 49xx "Global" family replaces dozens of national fixed-length dialects with one EDIFACT-based definition per business document: VDA 4984 for delivery forecasts and call-offs (DELFOR), VDA 4987 for despatch advices (DESADV), VDA 4938 for self-billing invoices (INVOIC). One mapping specification works for a plant in Wolfsburg and a plant in Puebla; validation tooling is standard; and the newer constructs (multi-level packaging hierarchies, cumulative quantities, schedule-line statuses) simply do not exist in the fixed-length world. For the supplier, the practical consequence is that your 4905-era habits - field offsets, record types, implied decimal places - become bugs in the new format.
The five syntax differences that bite
- Offsets become tags. In 4905 a delivery note number lives at fixed character positions in a record; in 4984/4987 the same fact is
BGM+351+<number>. Nothing stops you from writingBGM+241+...(the DELFOR code) into a DESADV by copy-paste - the file looks plausible and fails only at the receiving mapping. Our VDA 4987 validator flags a wrong BGM document code as an error with the exact segment position. - The UNA segment rules the alphabet. EDIFACT files open with
UNA:+.? 'declaring the component, element and segment separators for the whole file. A generator that hardcodes defaults and then emits a message containing the separator character inside a part description will silently corrupt every following segment. Always re-read UNA before parsing - or let a parser do it, like the one behind our tools. - Counts must match reality.
UNTdeclares the segment count of its message,UNZthe message count of the interchange, and the interchange control reference in UNB must equal the one in UNZ. Fixed-length files had no equivalent, so hand-edited test files almost always break here: fix a typo, forget to recomputeUNT+1, and the OEM parser rejects the file at the envelope before any content check runs. - Dates change shape. The old VDA world used DDMMYY-style positions; EDIFACT DTM with code
102is eight digitsCCYYMMDD, and203is twelve digitsCCYYMMDDHHMM. The single most common migration defect we see isDTM+11:13.11.2026:102- a German dotted date pasted into a 102 field. The validators treat it as a hard error because downstream MRP systems will, too. - Numbers become plain. Fixed-length formats padded quantities with leading zeros and encoded implied decimals; EDIFACT wants the plain numeric value -
QTY+113:1250:PCE, notQTY+113:1.250:PCE(that parses as one point two five) and not000001250unless your partner's guideline explicitly says otherwise. Clarify the rule per field in your mapping specification; do not assume.
Field mapping cheat sheet: 4905 concepts to 49xx segments
| 4905-era concept | Global message equivalent | Notes |
|---|---|---|
| Delivery note / advice number | BGM+351 + RFF+CRN (despatch advice number) | 351 is the DESADV document code; wrong code = hard error |
| Supplier / customer numbers | NAD+SE / NAD+BY / NAD+ST with GLN or assigned codes | Qualifiers and code lists differ per OEM guideline |
| Despatch / arrival dates | DTM+11: (shipped) and DTM+132: (arrival), code 102 or 203 | Eight or twelve digits, no separators |
| Gross / net weights | MEA+AAX+AAD:…:KGM and MEA+AAX+AAL:…:KGM | Both are mandatory in most OEM guidelines - see the 4987 common errors write-up |
| Item rows and quantities | LIN loop with QTY+12 (despatched) and QTY+11 (split) | Plain numerics only |
| Schedule/call-off logic (4913 role) | SCC status (1/2/3/4) + QTY + DTM in VDA 4984 DELFOR | Commitment strength must survive the mapping; do not flatten everything to SCC+4 forecast |
A migration test strategy that survives the audit
- Dual-run one full week. Keep the legacy 4905 flow alive while the new 4984/4987 flow runs in parallel; reconcile quantities per material number, not per file. Differences here reveal mapping bugs that structural validation cannot.
- Structural-validate every generated test file before sending. Paste outbound DESADV into the VDA 4987 validator and outbound/inbound DELFOR into the VDA 4984 validator; fix every error, review every warning against your partner's guideline. Both tools run 100% locally - production part numbers never leave your browser.
- Test the unhappy paths. A truncated file (missing
UNZ), a hand-edited quantity, an Umlaut in UTF-8 where the partner expects ISO 8859-1, an empty file at end-of-day. Write down what your system does in each case before the OEM's test team finds it for you. - Pin the versions. Record, per partner: VDA recommendation number, EDIFACT directory release (D:04A / D:07A / D:20B…), guideline version and date. The 49xx evolution history explains why this one line prevents more test-phase churn than any tooling.
The 4905 era ended gradually and then suddenly: first one OEM project requires the Global message, then another, and eventually the legacy interface is the exception. Migrating early, with validated files and pinned guideline versions, turns that surprise into a checklist.
Update (2026-11-14): the full guide edition is live - VDA 4905 to VDA 4984 Migration Guide: Field Differences, Parallel-Run Errors, Checklist, including the structure comparison table and six-step migration checklist.