ISO 20022 Address Parser
Turning messy party-address text into structured ISO 20022 PstlAdr output.
A technical preview that turns free-text party addresses into structured ISO 20022 PstlAdr output. It uses a hybrid workflow with confidence bands, reason codes, and strict PstlAdr validation to make structured-address readiness for CBPR+ reviewable rather than guessed.
Introduction
Payment messages are moving to ISO 20022, and party address data is one of the hardest parts to get right. The ISO 20022 Address Parser is a technical preview that turns messy, free-text party-address input into structured ISO 20022 PstlAdr output that is constrained, reviewable, and operationally useful. It comes out of about nine years at Swift, much of it spent on SwiftRef, the BIC directory, and bank reference data.
The Problem
From November 2026, CBPR+ usage guidelines begin retiring fully unstructured postal addresses in cross-border payments, pushing banks toward structured and hybrid ISO 20022 addresses. The catch is that most real address data still lives as free text in beneficiary files, customer master data, ERP payment records, and onboarding systems. Splitting that text into the correct PstlAdr elements (street name, building number, town, post code, and country) at scale, without silently guessing, is genuinely hard, and getting it wrong shows up as repairs, investigations, and stalled payments.
The Build
The parser uses a hybrid workflow rather than a single model call. Input is routed and cleaned, enriched with libpostal hints, then passed to a fine-tuned 4B model served from a private, token-gated inference endpoint. Post-model rules and strict PstlAdr validation run on top, and every field comes back with confidence bands and reason codes so a reviewer can see what the parser was sure about and what it was not. My own V3 qualification eval reached 85.87% micro field accuracy on 2,000 cases held out and hash-segregated from training, and every one of those cases produced well-formed final XML. Those are my own evals, not an external benchmark or a promise on anyone else's data.
Why It Matters
Banks already carry the cost of weak reference data through manual repairs and opaque exception handling, and the ISO 20022 migration raises the stakes for address quality specifically. The useful tool here is not a black box that quietly rewrites addresses. It is a constrained parser that structures party addresses into PstlAdr, shows its confidence, and keeps a human in the loop, so structured-address readiness for CBPR+ becomes something a team can review and trust rather than guess at. The aim is a repeatable way to take any ISO 20022 address format and produce a structured address ISO 20022 output that holds up to review.
Tags: ISO 20022, CBPR+, Structured Address, PstlAdr, Payments, Reference Data, SWIFT, Data Quality, Migration