The Correspondent Account Statement Is a Treasury Control Artifact

Treasury Desk Brief
Source retrieval 2026-08-19

A payment can move quickly while the account record still tells a more complicated story. Treasury still needs to know what is booked, what the account servicer explicitly reports as available, what has not yet appeared in the booked record, and what is still unresolved in reconciliation.

A fast rail can tell treasury that a payment event happened. That still does not tell a reviewer what the correspondent account actually shows at a given time.

That matters because a treasury review rarely ends with “the payment was sent.” The reviewer still has to rebuild the account picture: which entries are booked, which balance is being cited, when that balance applies, whether the source reports an availability balance, how the account entry links back to the payment, and what is still unmatched.

That is why the correspondent account statement is a treasury control artifact: reviewer-visible evidence that helps reconstruct account state. It does not, on its own, prove that a control is effective, that funds are legally or operationally usable, or that liquidity is sufficient.

Start with the reporting object, not the word “balance”

In correspondent banking, the two banks can describe the same account relationship differently. The respondent or customer bank may call its account at another bank a Nostro account; the servicing bank sees the corresponding account from the Vostro/Loro perspective.

For account reporting, ISO 20022 separates three important objects:

  • camt.052 — Bank-to-Customer Account Report, used for intraday account reporting;
  • camt.053 — Bank-to-Customer Statement, the statement object;
  • camt.054 — Bank-to-Customer Debit/Credit Notification, an advice or notification object that does not itself carry account balances.

There is also a version-layer distinction worth making once and then getting out of the way. As reviewed on 19 August 2026, the live ISO repository lists version 14 of these message definitions, while the current Swift CBPR+/CGI materials reviewed for this article use version 8. “Current camt version” is therefore incomplete unless the layer or market profile is named.

Legacy MT reporting needs the same caution. MT940/942 and related category-9 messages still matter in defined scopes and coexistence arrangements, even while Swift’s migration direction is toward the camt reporting family. Calling them “obsolete everywhere” would be too strong.

The practical rule is simple: before interpreting any field, identify the reporting object, profile, account servicer and coverage period behind it.

Booked, available, pending-like and unmatched are not four peer statement fields

A treasury review may need all four concepts, but they do not all come from the statement itself.

Booked state can come directly from the reporting source. In the Swift PMPG interbank camt.053 practice, entries are booked (BOOK). Reporting also distinguishes booked balance types such as closing booked, opening booked and interim booked balances. For the reviewer, that means the entry has been posted to the account servicer’s books under the cited practice. It still does not establish legal finality, accounting recognition or unrestricted use.

Explicitly reported available state can also come from the source — but only when the report actually supplies an availability balance code and its context. CGI reporting practice defines available-family balance codes such as closing available, opening available, interim available and forward available. Read those codes with the timestamp and the account-servicer or market-practice definition behind them. Do not quietly turn “available” into “cash you can use now.”

A pending-like or not-yet-booked state needs a different source. The reviewed material does not establish one universal pending field across camt.052, camt.053, camt.054 and all banks. If an intermediate payment state matters, point to the payment, event or service-specific evidence that actually supplies it. A missing statement entry is not enough to manufacture a PENDING status.

Unmatched state sits on the review side, not the statement side. The reviewed reporting standards provide references and transaction data that can support matching, but they do not establish a universal statement field called unmatched. “Matched”, “unmatched” or “ambiguous” is a reviewer-derived reconciliation classification, so keep the match rule, reason and evidence pointers with it.

That separation is the point of the control artifact: source state on one side, review state on the other.

Time is part of the evidence

An account-state review can still be wrong when every amount is right if all the timestamps are collapsed into one “as of” label.

At minimum, the reviewer may need to distinguish:

  • message or report creation time;
  • reporting-period start and end;
  • the timestamp attached to a particular balance;
  • booking date or time for an entry;
  • value date or time;
  • the reviewer’s own evidence-freeze time.

These are not interchangeable.

Each clock answers a different question. Creation time says when the reporting object was created. The reporting period says what window it covers. A balance timestamp says when that balance applies. Booking date records when the entry was posted to the account-servicer books under the cited practice. Value date has its own account-availability timing meaning in the PMPG interbank practice.

So if a reviewer receives a report at 10:05, 10:05 should not automatically become the balance cutoff, booking time, value time and business cutoff. Keep the clocks separate.

Link the account entry back to the payment event

A statement becomes much more useful when the reviewer can trace an entry back to the payment event behind it.

Swift PMPG interbank guidance points to references such as UETR, Instruction ID, End-to-End ID and Account Servicer Reference as useful linkage fields when they are present. Other transaction identifiers, amount, currency, date and party/remittance information may help corroborate a match.

The catch is that there is no universal join key to assume. In particular, Entry Reference should not be treated as a guaranteed cross-message key across camt.052, camt.054 and camt.053.

In practice, the reviewer can trace the evidence in this order:

  1. payment or settlement event;
  2. debit/credit advice where used;
  3. intraday account report;
  4. booked statement;
  5. reconciliation record linking the exact evidence objects;
  6. exception and closure evidence when the link is incomplete or later reversed.

That is a linkage pattern, not an ISO or bank standard.

The statement does not complete reconciliation by itself

A bank statement can give the reviewer booked entries, balances, timestamps, references and transaction codes. It cannot decide whether the organization matched the expected payment correctly or whether an exception is actually resolved.

That decision belongs in a separate reconciliation record.

The derived fields can stay simple: match state, rule or review method, reason, confidence where relevant, exception ID, owner and next evidence required. If the match is ambiguous, leave it ambiguous instead of forcing a link from amount and date alone.

Returns and reversals make the separation even more important. PMPG practice provides return/reversal indicators and original/return references that can help connect a new account movement to the original payment. The reviewer should preserve both events. A return does not mean the original payment “never happened,” and a reversal should not overwrite the original evidence.

Failure cases the evidence file should make visible

This is where the control artifact becomes most useful: when the evidence is incomplete or conflicting.

A stale report should be marked stale from its coverage, creation time or balance time — not carried forward as “current.”

A missing report or sequence/page gap should open an evidence-gap exception. No report does not mean no account activity.

With a delayed posting, keep the payment/event state separate from the booked account state. Do not invent a statement-level pending status while you wait for posting.

Treat a duplicate delivery as a duplicate evidence object, not as a second account movement.

For a value-date mismatch, preserve booking date and value date separately rather than forcing them to agree.

A missing reference should lower match confidence and make the secondary matching attributes visible. Amount and date alone should not become a guaranteed join.

Keep a service or channel delay classified as a freshness problem. By itself, it is not evidence of a liquidity shortage.

Finally, a bank-specific availability definition should stay bank- or service-specific. The same balance code family does not create a universal promise about overdrafts, credit lines, collateral, earmarks or executable funds.

Intraday Liquidity Evidence Map

Treasury Desk author synthesis — reviewer evidence map.

This map records account-state and reconciliation evidence; it does not calculate complete intraday liquidity or prove funds are executable.

Reading note: Source fields remain source fields; source-derived normalized fields and reviewer-derived fields are Treasury Desk synthesis. The map is not an ISO or bank standard.
Field Provenance Reviewer use Boundary
Evidence object type (camt.052, camt.053, camt.054, legacy or separately identified event) [Source field] Establish the semantic scope of the record Do not merge different objects into one generic “statement”
Message definition / profile version [Source field] Distinguish ISO definition from community/bank profile Do not call a market-profile version the latest ISO version
Source owner / account servicer [Source-derived normalized field] Keep bank/service-specific semantics attached to the record Public examples should use sourced institutions or Bank A
Account reference pseudonym [Reviewer-derived field] Group evidence without exposing an account number Use synthetic labels such as NOSTRO-USD-01 only
Currency [Source field] Establish balance/entry context Synthetic examples only
Report/statement ID [Source field] Identify the reporting object Preserve missing values rather than inventing one
Creation time [Source field] Show when the object was generated Not the same as coverage or booking time
Reporting period from/to [Source field] Show the state window Preserve timezone/offset
Sequence/page/last-page indicators [Source field] Support completeness checks Do not assume a partial page set is complete
Balance code, amount and balance time [Source field] Distinguish booked vs explicitly reported available balance Never collapse to a generic “available cash” field
Entry amount and debit/credit direction [Source field] Identify the movement Synthetic values in public illustrations
Entry status [Source field] Preserve an explicit source status such as BOOK Do not create PENDING when absent
Booking date/time [Source field] Show posting time Not legal or accounting finality
Value date/time [Source field] Preserve source-defined availability timing Do not turn it into accounting-recognition date
UETR / Instruction ID / End-to-End ID / Account Servicer Reference [Source field] Link payment/event evidence to the account entry when present No single identifier is guaranteed universally
Source state [Source-derived normalized field] Normalize only what the source actually supports: booked, explicitly reported available, notification seen, not evidenced Keep the original source pointer
Match state [Reviewer-derived field] matched, unmatched or ambiguous under a stated rule/review Not an ISO/bank field
Match rule/version and reason [Reviewer-derived field] Make reconciliation reproducible Do not hide a manual/heuristic match behind a generic status
Exception owner and next evidence [Reviewer-derived field] Keep unresolved items actionable for review Use role labels, not real names, in public examples
Source reference and retrieval time [Reviewer-derived field] Preserve lineage and freshness Retrieval time is not account observation time

A public example should stay synthetic. It can still be useful without a real statement, account number, UETR, customer name, remittance narrative or bank credential.

What this does not prove

This evidence model does not prove that a specific bank balance is currently accurate, legally usable or executable by a particular entity. It does not prove the availability of overdrafts, intraday credit, collateral or other funding resources. It does not prove payment, settlement, accounting or legal finality.

It does not determine accounting recognition, journal-entry timing, compliance sufficiency, audit assurance or control effectiveness. It does not make pending or unmatched universal ISO 20022 fields, and it does not assume every bank uses the same camt version, cadence, balance codes or reference population.

A statement or report is one account-state evidence object within a broader treasury review. BCBS intraday-liquidity guidance explicitly treats liquidity resources as broader than a single account balance, including credit capacity, collateral and other balances.

Treasury Desk’s public method is read-only and reviewer-first. It does not imply live bank access, statement ingestion, reconciliation posting, cash movement or execution.

A reviewer question worth testing

Where does a public or synthetic reconciliation case get hard to reconstruct: report timing, value date, reference mapping, or exception ownership? If you know a public implementation guide that clarifies one of those boundaries, that is the most useful feedback for this map.

Reader-facing sources

Sources below were reviewed for the Source / Evidence Pack on 19 August 2026. Dynamic standards/profile and bank-service pages require a publication-day refresh before release.


Leave a Comment