generated gpt
ISO 20022 and Compliance Screening: What Structured Data Really Changes in the Operational Contour
SWIFT's parallel MT/ISO 20022 period for CBPR+ ended in November 2025 — and with it, the era of the single unstructured text line in cross-border payments began to close. But does structured data automatically mean cleaner sanctions screening? Not quite. ISO 20022 creates the technical possibility for a 25–30% reduction in false positives — yet only if the fields are actually populated, counterparty reference data is clean, screening rules are recalibrated, and every bank in the chain is on board. This analysis breaks down what really changed, where the gaps remain, and the five steps compliance teams must take to turn format compliance into measurable screening efficiency.
Payment messages that banks exchange for cross-border transfers have been transmitted in an "unstructured" format for decades. Since March 2023, the global financial infrastructure has been migrating to ISO 20022. The key milestone was passed on 22 November 2025, when the parallel MT/ISO 20022 period for CBPR+ ended.
Immediately after the transition weekend, the community reached 97% adoption, and by August 2026 this figure exceeded 98%. It was the completion of the parallel period, not the current figure, that became the main event of the migration.
In the old format, all data about the payer, their address and country were recorded in a single continuous line. The screening system compared this line against sanctions lists. This generated numerous errors: according to SWIFT estimates, 99% of alerts were false positives.
The new format breaks information into separate fields: name, address, country, company identifier. The screening system can compare each field separately. This theoretically allows reducing false positives by 25–30%. In practice, however, the effect depends not on the format itself, but on how well and completely these fields are populated. If a bank does not transmit the country or LEI, there will be no improvement.
As of August 2026, more than 98% of payment instructions in the SWIFT network are transmitted in the new format.
The situation varies across countries:
Canada completed its transition to ISO 20022 in November 2025, tying it to the global Swift deadline. However, the national Lynx system implemented the standard as early as March 2023, and by October 2025 more than 98% of its messages were already in MX format. November 2025 is the date of the end of format coexistence.
The United Kingdom prepared in advance: the Bank of England had been supporting CHAPS participants' preparation since 2018, introduced strict control over field completion and made LEI a priority;
in the United States, key systems (CHIPS and Fedwire) have already migrated to the new format. CHIPS — in April 2024, Fedwire — in July 2025.
Initially, November 2026 was named as the final frontier, when unstructured addresses were to stop being accepted in the SWIFT network. However, in August 2026 SWIFT decided on a controlled extension of deadlines. The requirement remains, but no new date has been set — an update is expected no later than December.
What exactly changed in the data
Before the transition to ISO 20022, all data about the payer and payee were transmitted in a single text field. Company name, street, city and country were recorded in one line without any structure. This made automatic processing impossible. The computer could not understand where the name was and where the address was, so it compared the entire line as a whole against blacklists.
This generated a mass of errors. For example, in the old format a line could look like this: "Pierre Dupont, 15 Rue de Rivoli, Paris, France". If a person with the surname Dupont is on a sanctions list, the system would generate a false positive, even if this is a different Dupont living in another country and having no relation to restrictions.
The system could not distinguish one namesake from another because it had no other data besides the line. The new standard breaks information into separate fields. Now each element has its own place:
name;
address;
city;
country;
postal code.
Each element is transmitted separately. This allows the machine to understand what exactly it is comparing and against what. This reduces the number of false positives, since the system compares individual fields rather than the entire line.
However, a partial match does not automatically remove the restriction: the alert still goes to manual review, and the final decision is made by the compliance officer.
In addition, the message now has a clear separation of all parties to the payment. The following are now explicitly indicated:
payer;
payee;
sending bank;
receiving bank;
ultimate beneficiary;
the party on whose behalf the operation is performed.
This is critically important because each of these parties may have its own sanctions statuses and risks, and they must be checked separately. The format also introduced special codes for the purpose of payment and the ability to transmit legal entity identifiers such as LEI.
The key change: the system is now able to distinguish one party to a payment from another. This structuredness creates the basis for accurate screening and reducing the number of errors.
How this affects screening
The new data format directly affects the quality of sanctions screening. Theoretically, it can provide:
Reduction in the number of errors. The system matches individual fields. A random surname match does not trigger a false alert — verification is performed by address, country and identifier together.
Geographic accuracy: there is a separate field for country. If a payment comes from a high-risk jurisdiction, additional checks are applied to it.
Transparency of all participants. Each of them may be under sanctions. Working under the new rules, the system should better identify sanctioned persons.
New screening scenarios. For example, a structured field with the payment purpose code has appeared. If a company from a certain sector pays to a high-risk country, this is immediately noticeable.
Important: all these improvements work only provided that the fields are actually populated.
Why the format alone does not solve the task
The structured format creates the technical possibility for accurate screening, but it does not realize it by itself. There are practical problems.
Incomplete completion. Transmitting data in structured form and their actual availability are different things. A bank may send a message in ISO 20022 format but leave most fields empty or transfer the old text string into them "as is". According to SWIFT, as of March 2026 about 65% of payment messages still contained unstructured addresses. By April this figure had dropped to 61.2% — there is dynamics, but it remains uneven: significant parts of the industry across all regions are not completing the transition on time.
Low quality of source data. Company names are written differently in different countries; transliteration from local alphabets creates different variants of spelling for the same name. Many counterparties, especially in developing countries, do not have an LEI identifier. Without it, the system cannot unambiguously match the payer with an entry in the sanctions list, and screening accuracy remains low.
Configuration of screening rules. Transaction monitoring systems have been calibrated for decades to work with unstructured data. They are accustomed to searching for matches in continuous text and generating an alert at any hint of a match. The transition to structured fields requires complete recalibration: sensitivity thresholds, matching algorithms, and decision-making logic must be reconfigured. Without this, the system will either miss real risks or generate false positives.
Weak link in the chain. A cross-border payment passes through several banks. If one of the intermediaries does not support ISO 20022 in full or transmits data in a limited format, all structured information is lost. On the recipient's side, the payment arrives again as unstructured text, and all the advantages of the new format are nullified.
Thus, ISO 20022 is a necessary but insufficient condition for effective compliance operations. Without discipline in data completion, a clean counterparty database, recalibration of rules and readiness of all participants in the chain, the effect for the compliance function will remain minimal.
What needs to change beyond the format
It is necessary to put counterparty reference data in order. Every client must have a structured address (country, city, street, building) and, where possible, an LEI. If this data is missing, it must be collected at the stage of account opening or payment initiation. Transferring old text strings into new fields does not solve the problem.
Screening systems configured to work with unstructured data do not start working better automatically after the transition to the new format. Sensitivity thresholds, matching algorithms and decision-making logic require complete recalibration and testing on historical data.
When the alert profile changes, the structure of the manual review queue must also be revised. If there are fewer false alerts, operators will process, for example, not 100 alerts per day but 70. However, each of them will likely prove more complex and require careful analysis.
Agreements with correspondent banks on the completeness and format of transmitted details become even more important in the context of the deadline extension announced by SWIFT. While some participants in the chain are preparing, others may be postponing. If at least one intermediary does not transmit the country or LEI, the detail is lost — and the effect of one's own preparation is nullified.
The effect of all changes must be measured. Before the transition to the new rules, baseline metrics must be recorded: the share of false positives, alert processing time, the share of manual reviews and straight-through processing. After implementation, measurement is carried out again.
Only comparison of figures allows proving that the changes worked. Without them, any effect remains hypothetical.
It is important to understand: full address structuredness is not the only compliance option. Since November 2025, a hybrid format has been in effect that combines structured fields with unstructured address lines. The minimum requirement for a hybrid address is city and country. For a bank that is not managing to convert all reference data into a fully structured form, the hybrid format provides a minimally sufficient level of compliance.
| Criterion | Screening on an unstructured message | Screening on a structured message (partial completion) | Screening on a fully completed structured message |
|---|---|---|---|
| What is fed as input | A single text field (mixture of name, address, country) | Separate fields, but part of the data is empty or transferred as text | All fields completed: name, address, country, LEI, purpose code |
| Typical cause of a false positive | Match of a city name with a sanctioned toponym | Absence of country or LEI, transliteration of name | Minimal — matching by typed fields |
| Possibility of jurisdiction-based rules | Absent — country is extracted from text | Limited — if the country field is populated, the rule works | Full — country is known and structured |
| Transparency of payment parties | Parties are mixed in one text, identification errors | Partial — some parties are identified, some are not | Full — all parties are explicitly separated |
| Main limitation | High level of false positives (99% of all alerts) | The format exists, but the data does not provide its advantages | Requires recalibration of rules and data discipline |
Conclusions
The effectiveness of ISO 20022 for the compliance function depends on three groups of factors that must be checked before the start of migration.
The quality of one's own counterparty reference data and procedures for collecting data from clients. A structured address must be formed at the stage of account opening or payment initiation, and not generated at the output of the payment message.
Measurable metrics recorded before and after implementation: false positive rate, average alert processing time and the share of straight-through processing (STP) for payments undergoing screening. Without baseline values, the effect will remain unprovable.
Agreements with correspondents and clients on the completeness of transmitted details, primarily the structured address and LEI. The requirement for structured addresses remains — Swift merely postponed its introduction due to uneven industry readiness. No new date has been set, but preparation cannot be stopped: when the deadline is announced, the industry may not have time left for further development.
Frequently Asked Questions (FAQ)
Does the transition to the structured format automatically reduce the number of false positives?
No. According to SWIFT, the reduction potential is 25–30%. However, this effect is achieved only with full completion of structured fields and recalibration of screening rules. Without changes in processes and data, the advantages of the format are not realized.
What happens to data detail if one of the banks in the chain transmits a message in a limited format?
Detail is lost. If an intermediary bank does not transmit structured fields (for example, uses the old format or transfers data as text), the ultimate recipient does not receive structured information. This makes screening on the recipient's side just as inaccurate as before the migration of payment systems.
Which metrics allow proving the effect of migration for the compliance function?
False positive rate, average alert processing time, the share of payments undergoing straight-through processing (STP) after screening, and the number of manual reviews. Measurements must be carried out on a comparable sample before and after the implementation of structured data. SWIFT offers testing services to assess the impact of the new format on filter effectiveness.
