VDA EDI Tools

From Fixed-Length Records to EDIFACT: The Evolution of the VDA 49xx Messages

VDA EDI Tools Blog · 2026-11-13

Understanding why VDA messages look the way they do - and why the same business document exists in two wildly different syntaxes - explains half the confusion in any automotive EDI project. The story has three acts: the fixed-length era, the EDIFACT migration, and the continuous-revision present.

Act 1 - the classic VDA record format (1970s-1990s)

Long before XML and EDIFACT, the VDA (Verband der Automobilindustrie) standardised German automotive data exchange with fixed-length record formats: each message a numbered sequence of record types, each field at a fixed byte offset. Suppliers of a certain age still speak this language fluently:

The format was robust for its era - punch-card discipline, deterministic offsets, no ambiguity - but rigid. Every change meant republishing offset tables, and interop beyond Germany was hopeless. EDIFACT was coming.

Act 2 - the EDIFACT migration and the 49xx family (1990s-2000s)

As UN/EDIFACT became the international standard, the VDA re-expressed its business documents as profiles of EDIFACT messages under new recommendation numbers in the 49xx range. The mapping was deliberately parallel so veterans could translate mentally:

Classic (fixed-length)EDIFACT-basedBusiness document
VDA 4913 (Abruf)VDA 4984 (DELFOR)Delivery forecast / call-off schedule
- (new capability)VDA 4985 (DELJIT)JIT sequenced call-off
VDA 4905 (Lieferschein)VDA 4987 (DESADV)Despatch advice / ASN
- (new capability)VDA 4988 (RECADV)Receiving advice
VDA 4906 (Rechnung)VDA 4938 (INVOIC, self-billing)Invoice - direction reversed: the customer bills the supplier
VDA 4911 (Gutschrift)4906 in EDIFACT formCredit note variants
-VDA 4915 (ORDERS)Purchase order

Two design decisions of that era still shape daily work. First, self-billing (4938): German OEMs moved invoice creation to their own side, so suppliers stop sending 4906 and instead confirm customer-generated invoices - an entire workflow that exists almost nowhere outside this industry. Second, the global variants: VDA 4984 exists in a “Global DELFOR” flavour and OEM-specific subsets, which is why two “VDA 4987” files can legitimately differ.

Act 3 - the revision treadmill (2000s-today)

Each 49xx recommendation is a profile of a specific EDIFACT directory release, and the releases march on: D:96A, D:01B, D:04A, D:07A, D:09B, D:16A, D:20B. The VDA recommendations and the OEM guidelines that implement them advance at different speeds, producing the familiar landscape:

What this history means for your project

  1. The version is part of the standard. “We use VDA 4987” is incomplete; the answer you need is “VDA 4987, release D:20B, per the plant guideline dated…”. Always pin the UNH-2 composite in your mapping contract.
  2. Mixed-syntax sites are normal. A plant can demand EDIFACT DESADV for ASN, classic 4902 for labels, and 4905 for the paper that rides in the truck driver's cab. Budget for all three.
  3. Direction matters. The migration inverted the invoice direction (4906 supplier-issued vs 4938 customer-issued). Confirm which side creates the invoice before designing anything around it.
  4. Guidelines age. An OEM guideline is a snapshot; check its release date against the current VDA recommendation and ask explicitly whether a newer revision is in pilot.

The 49xx family is not a museum - it is a living stack where three decades of decisions (fixed offsets, self-billing, global variants, release pinning) still surface in today's test files. Knowing which era a requirement comes from is often the fastest way to understand why it exists at all.

How the directory releases differ in practice

The lettered releases are not marketing labels; they change element positions, add segments and retire code values. Three examples suppliers hit constantly:

The practical rule: never ask a customer “which VDA message?” and stop. Ask “which VDA recommendation, which EDIFACT release, which guideline version and date?” - and pin all three in your mapping specification. That single line in the spec prevents more test-phase churn than any amount of clever code.

← All articles