Building a posting crosswalk when the source rule engine is not publicly documented

A disciplined discovery framework for translating QAD's operational-event-to-ledger behaviour into D365 posting profiles, built on tenant extracts and transaction replay rather than on documentation that does not exist.

What you will be able to do

Introduction

Every other module on this path can tell you what to build. This one mostly tells you what to find out, and that difference is deliberate.

The ten source transaction families

The adapter's posting rules define ten families. Four are recorded as gaps with low or medium confidence; six are recorded as mapped but every one carries a pending client-validation flag.

The event census

Before any mapping, build a census. This is the artefact the whole module rests on.

The evidence each family needs

The census tells you what exists. This table tells you what to demand for each family before design starts.

The four families that need most attention

These are one problem, not two, and the problem is the accrual between them. The target models product receipt and vendor invoice as distinct accounting events with a clearing account holding the received-not-invoiced position in between.

Mapping to the target constructs

D365 splits posting derivation across several constructs, and knowing which one owns which decision prevents a great deal of misdirected design.

Proving it with replay

The playbook specifies golden scenarios, and they are the correct replay set: purchase receipt and invoice, sales shipment and invoice, production issue, completion and variance, quarantine, scrap and release, intercompany, recall and traceability drill-back, and EDI shipment acknowledgement.

What to do while waiting for extracts

Extracts take time to obtain, and the programme cannot idle. Four activities are productive before they arrive.

Knowledge check

Summary

The QAD posting engine is not publicly documented, and this module does not pretend otherwise. The translation is a discovery exercise with named deliverables: a configuration extract and a posted-transaction sample per family, an event census that separates real behaviour from configured possibility, and a replay test that compares…