KEY CHALLENGES IN TRANSFORMING SWIFT MT TO ISO 20022
Where the migration stands
The question facing institutions today is no longer whether they can send and receive ISO 20022 messages: on the cross-border network they have no alternative. The question is whether they can do anything useful with what those messages carry. The technical difficulty of the migration was never primarily about message syntax. It was about the depth of the data model, the state of the systems and data behind it, and the regulatory and commercial layers now building on top of the standard. The current phase is where the programme’s toughest challenges collide.
Sourcing structured data, not just parsing it
ISO 20022’s XML-based data model is far richer than MT’s fixed-format fields, and that richness now has a firm, imminent deadline attached. A hybrid address option was introduced with the November 2025 standards release, opening a one-year grace period in which structured, hybrid or unstructured addresses are all still accepted on cross-border payment messages. That grace period ends on 14 November 2026, when fully unstructured addresses will be rejected outright. Purpose codes and Legal Entity Identifiers have moved from optional enrichment to expected content. The difficulty is not building a parser for this data; it is sourcing it, at speed, from customer records that were often captured as unvalidated free text for decades.
Example: a corporate client’s address on file reads simply “123 Business Park, London.” Under the rules taking effect in November 2026, this must be decomposed into building number, street name, town, postcode and country as discrete, individually addressable elements — data the bank never validated at account opening and now has a few months left to retrieve or infer.
Integration debt relocated, not discharged
Financial institutions have had to enhance or replace core systems to accommodate ISO 20022, in coordination with counterparties, payment networks and service providers. At the market-infrastructure layer this has substantially happened: the Eurosystem’s T2 platform, CHAPS, FedNow, CHIPS and the Fedwire Funds Service have all completed native cutovers. Inside individual institutions, however, comparatively few reengineered their core banking systems. Most placed a translation gateway at the perimeter and left internal formats untouched.
Example: a bank receives a fully structured pacs.008 payment instruction at its gateway, translates it back into a legacy MT-shaped internal record to post to the core ledger, and in doing so discards the structured purpose code and remittance detail the sender took care to include.
Mapping and conversion risk, now internalised
While MT and ISO 20022 ran in parallel, the network itself absorbed format mismatches, rebuilding payment data centrally so a message could be delivered in whichever format the receiver could handle, but coexistence service has ended. Every remaining mapping decision now falls to the institution’s own edge, and a field dropped between a gateway and a core system produces no network-level exception, it simply disappears.
Example: a remittance information field longer than a legacy back-office system’s character limit is silently truncated at the gateway. The payment settles; the invoice references the beneficiary needed to reconcile it does not survive the trip.
Parallelism has moved, not ended
Coexistence for cross-border payments has closed, and MT is no longer accepted directly for that traffic. Two qualifications matter. First, SWIFT did not impose a hard technical block: institutions still sending MT can route it through a paid contingency conversion service, subject to escalating penalties, and MT remains in active use in domestic systems, closed user groups, and certain cash-management reporting messages that were given a longer runway than the main deadline. Second, and more importantly, parallel running has not ended for institutions themselves, it has simply moved behind the gateway, where it is cheaper to maintain, less scrutinised, and considerably harder to see.
Example: a bank’s payments team reports full ISO 20022 compliance while its reconciliation and sanctions-screening tools still operate on an MT-derived internal message format, maintained by a small team few people in the organisation know still exists.
Testing as a recurring obligation, not a project
Comprehensive testing was necessary to confirm systems processing ISO 20022 messages worked correctly with counterparties and market infrastructure. Shared standards repositories, readiness portals and in-flow validation matured considerably during the migration. What has proved more durable is that testing has become an annual operating obligation rather than a one-off migration cost: usage guidelines are refreshed with each yearly standards release, and any refresh can tighten a validation rule or change a field’s optionality.
Example: an institution that stood down its ISO 20022 testing team after go-live finds a batch of outbound payments rejected the following November, after a routine annual usage-guideline update quietly restricted a code list its interface still relies on.
A skills gap that has moved from syntax to semantics
Schema literacy, reading and producing valid XML, turned out to be the easier problem, largely solved by tooling. What remains scarce is semantic competence: knowing which of several plausible elements a given business fact belongs in, and how a market’s usage guideline narrows the base standard. This is data stewardship, not message engineering, and it sits awkwardly between technology, operations and compliance functions. At the same time, the generation fluent in MT is retiring precisely when residual MT expertise is still needed for domestic rails and archived traffic.
Example: two operations analysts at the same bank populate the same ISO 20022 field differently for an identical transaction type, each technically valid under the schema, producing inconsistent downstream reporting that only surfaces during an audit.
Regulation assuming structured data, not merely permitting it
This has been the fastest-moving area of the programme, and it is unambiguous: regulators now build requirements on the assumption that structured data exists. The Financial Action Task Force’s revised Recommendation 16 tightens originator and beneficiary information expectations in a manner aligned with ISO 20022 fields. The EU’s Instant Payments Regulation has driven a sequence of hard obligations on receiving, sending and verifying payees. In UK, enhanced-data requirements on CHAPS have been phased in for defined payment types.
Example: a payment that would have cleared without incident under MT is now held for manual review because a payee-verification service flags a mismatch between the structured beneficiary name and the name on file, a control that could not have existed before the data was structured.
A business case still waiting to be funded
The upfront premium for the migration that is system enhancements, testing, training and dual-running costs has now largely been paid. That makes the current period the first honest test of the business case, and results are uneven. Institutions that pushed structured data through to reconciliation, screening and liquidity forecasting have seen real gains in straight-through processing. Institutions that terminate the data at a gateway have a compliance certificate and little else. The benefits of ISO 20022 were never an automatic by-product of the migration; they are a separate, largely unfunded project.
Example: a bank’s compliance team can now screen structured sanctions fields with far fewer false positives, but the productivity gain never reaches the wider organisation because no budget was allocated to extend structured-data use beyond the screening engine itself.
Conclusion
Retiring MT was never the objective; it was the entry requirement. The technical challenges that remain sourcing rich data, retiring internal translation layers, sustaining annual testing discipline, closing the semantic gap between usage guidelines, and funding the second project of turning structured data into operational value are harder than the messaging change itself, and less visible, because most of them no longer generate the kind of network-level exception that used to force attention. Institutions that treat the current phase as complete because the network is single format are likely to discover otherwise at their own front door.
About the Author
Mr. Saravanakumar Dinakaran
Senior Managing Consultant – IBM
Mr. Saravanakumar Dinakaran is a Industry Consultant with expertise in various areas of Investment Banking and Securities Markets Industry.
Mr. Saravanakumar Dinakaran is holding 20 plus years of experience in variety of Information technology services and solution providers for global capital market participants such as Investment banks, Buy & Sell side firms, financial service providers and market regulator.
Mr. Saravanakumar Dinakaran is Interested to work in process improvement and regulatory compliance projects and assignments related to Derivatives (OTC & ETD) market operations.
Mr. Saravanakumar Dinakaran have strong experience in the Report Warehouse development project successfully launched in the Blockchain concept Distributed Leger Technology (DLT) for the Global leader in Depository and clearing services.
Mr. Saravanakumar Dinakaran is Bestowed with the following Certifications:
https://www.linkedin.com/in/
Mr. Saravanakumar Dinakaran can be contacted at:
