PAYMENTS SIGNAL REFERENCE ARCHITECTURE · SYNTHETIC / TRAINING ONLY

Instant-payment processing, settlement, timeout, and 24×7 operations architecture

Makes the service clock, bounded controls, funded settlement, receiver response, late-status ordering, and 24×7 recovery responsibilities explicit.

Payments Signal reference architecture v1 · reviewed 2026-07-23

Reference architecture—not a scheme mandate. The blueprint uses one time-budgeted route to explain common instant-payment responsibilities. Exact execution clocks, response codes, prefunding, finality, return rights, maintenance, and participant obligations differ by rail and current rulebook.

Audience and purpose

Instant-payment architects, fraud and sanctions teams, operations, SRE, product, and business analysts

Components

Payer channel and confirmation

Captures an always-available payment intent and tells the payer what is known within a short time budget.

Kind
channel
Owner
Digital channel
Responsibilities
  • Authenticate and confirm the payer's intent
  • Show a truthful pending, rejected, or completed outcome
Inputs
  • Payer, payee, amount, account or proxy, confirmation
Outputs
  • Idempotent instant-payment request
Controls
  • Authentication
  • Entitlement
  • Confirmation
  • Client idempotency key
Failure modes
  • Repeated tap
  • Channel timeout
  • Stale payee verification
Recovery
  • Enquire using the same reference
  • Do not create a second payment while outcome is unknown
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Sending instant-payment engine

Completes validation, controls, posting intent, routing, and submission inside the service time budget.

Kind
application
Owner
Sending payment product and technology
Responsibilities
  • Validate and route quickly
  • Coordinate screening and fraud decisions
  • Reserve or debit under policy
Inputs
  • Confirmed instant-payment request
  • Reference and risk data
Outputs
  • Network instruction and durable pending state
Controls
  • Duplicate detection
  • Account and funds check
  • Time-budget enforcement
Failure modes
  • Control exceeds time budget
  • Unknown network outcome
  • Internal posting lag
Recovery
  • Stop or time out under the rail rules
  • Reconcile before retry
  • Release reservations from proven outcomes
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Real-time fraud, sanctions, and payee controls

Returns a clear, reasoned outcome without letting a control timeout become an accidental approval.

Kind
control
Owner
Fraud and financial crime teams
Responsibilities
  • Assess transaction and party risk
  • Apply payee-verification outcome where the service uses it
Inputs
  • Payment, customer, device, party, list, and behavioural data
Outputs
  • Clear, stop, challenge, or controlled timeout
Controls
  • Fail-safe policy
  • List freshness
  • Model/rule version
  • Decision audit
Failure modes
  • Control timeout
  • False positive
  • Stale list or model
Recovery
  • Apply the documented fail-safe outcome
  • Investigate without hiding the payment state
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Instant clearing and routing service

Validates and routes the payment or settlement instruction continuously under the rail’s timeout rules.

Kind
external-network
Owner
Instant-payment infrastructure operator
Responsibilities
  • Confirm participant reachability
  • Apply rail validation and duplicate controls
  • Route or settle within the service clock
Inputs
  • Instant-payment message
Outputs
  • Acceptance, rejection, timeout, or settlement outcome
Controls
  • Participant status
  • Duplicate detection
  • Service clock
  • Availability
Failure modes
  • Unavailable receiver
  • Network timeout
  • Duplicate or invalid message
Recovery
  • Return the rule-defined outcome
  • Support status enquiry and exception handling
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Instant settlement accounts

Moves central-bank or designated settlement asset per transaction or under the rail’s funded model.

Kind
ledger
Owner
Central bank or instant-settlement operator
Responsibilities
  • Check funded capacity
  • Post the settlement movement
  • Return finality evidence
Inputs
  • Eligible instant settlement instruction
Outputs
  • Settlement debit and credit
Controls
  • Available funds
  • Atomic posting
  • Finality evidence
Failure modes
  • Insufficient prefunding
  • Settlement service unavailable
  • Posting uncertainty
Recovery
  • Reject or queue only where rules permit
  • Reconcile the authoritative ledger before reopening
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Receiving instant-payment engine

Validates the beneficiary and posts funds once the rail’s conditions and settlement evidence are satisfied.

Kind
application
Owner
Receiving payment product and technology
Responsibilities
  • Validate the destination account
  • Respond within the rail clock
  • Post once and report availability
Inputs
  • Instant payment
  • Settlement outcome
Outputs
  • Beneficiary posting and response
Controls
  • Account status
  • Duplicate posting protection
  • Settlement-state check
Failure modes
  • Invalid account
  • Receiver response timeout
  • Core ledger unavailable
Recovery
  • Return the rule-defined response
  • Repair posting idempotently after confirmed settlement
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Status, exception, and reconciliation service

Connects customer status, network outcome, settlement, participant posting, return, and investigation evidence.

Kind
operations
Owner
Instant-payment operations
Responsibilities
  • Answer status enquiries from authoritative state
  • Reconcile timeouts and late outcomes
  • Control returns and investigations
Inputs
  • Network status
  • Settlement evidence
  • Participant postings
  • Customer enquiry
Outputs
  • Truthful customer status
  • Resolved timeout
  • Return or investigation
Controls
  • Original-reference integrity
  • Late-status ordering
  • Four-eyes funds action
Failure modes
  • Late positive response after customer timeout
  • Duplicate return
  • Conflicting participant states
Recovery
  • Establish settlement first
  • Correct customer status without duplicating value
  • Use the rail's exception path
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

24×7 service operations and continuity

Maintains round-the-clock observability, capacity, failover, change, and incident ownership.

Kind
operations
Owner
Participant and infrastructure service operations
Responsibilities
  • Operate without a daily recovery window
  • Coordinate participant and infrastructure incidents
  • Protect the service clock during failover
Inputs
  • Latency, queue, error, liquidity, and dependency telemetry
Outputs
  • Incident action, capacity action, continuity decision
Controls
  • Error budget
  • Change freeze and rollback
  • Failover test
  • On-call ownership
Failure modes
  • Cascading dependency latency
  • Capacity exhaustion
  • Uncoordinated maintenance
Recovery
  • Shed or stop safely under the rules
  • Fail over without duplicate value
  • Reconcile in-flight payments
Non-functional requirements
  • Durable correlation through stable business and technical identifiers
  • Capacity and availability matched to the service-level objective
  • Auditable state changes, configuration, and operator actions

Interfaces

Confirmed instant request

instant-payer → instant-sender

Start one idempotent payment and its end-to-end time budget.

Contract
api · synchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
A channel timeout preserves the client reference for enquiry.

Bounded control request

instant-sender → instant-controls

Obtain explicit fraud, sanctions, and payee-control outcomes.

Contract
control · synchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Unavailable or late controls follow a documented fail-safe outcome.

Instant payment submission

instant-sender → instant-network

Submit the payment once participant checks and controls pass.

Contract
message · asynchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
An ambiguous send is enquired before any retry.

Per-payment settlement

instant-network → instant-settlement

Move funded settlement asset under the rail's model.

Contract
settlement · synchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Insufficient funds or unavailable settlement returns the rule-defined outcome.

Routed payment and response

instant-network → instant-receiver

Deliver the payment and receive the beneficiary validation result.

Contract
message · asynchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
A missing receiver response becomes a precise timeout state, not an automatic success.

Settlement evidence

instant-settlement → instant-receiver

Prove the settlement state used by the receiver's posting policy.

Contract
event · asynchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Delayed evidence is reconciled before duplicate or reversal action.

Beneficiary outcome

instant-receiver → instant-reporting

Record the receiving decision and posting state.

Contract
event · asynchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Late or conflicting outcomes enter ordered reconciliation.

Customer outcome

instant-reporting → instant-payer

Tell the payer only the outcome supported by current evidence.

Contract
api · synchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Unknown remains pending or timed out; it is not changed to failed merely to close the screen.

Service-clock telemetry

instant-network → instant-resilience

Expose availability, latency, error, and participant health.

Contract
operator · asynchronous
Controls
  • Authentication and authorisation
  • Integrity and replay protection
  • Correlation and audit evidence
Failure treatment
Telemetry loss raises an incident and does not fabricate payment status.

Authored traces

One funded instant credit transfer

Follow one payment through payer confirmation, bounded controls, network routing, settlement, beneficiary posting, and customer status.

  1. Payer confirms once — Customer confirmed. The channel creates one stable reference and starts the service clock.
  2. Sender validates and funds — Participant accepted. Account, funds, duplicates, route, and time budget pass.
  3. Controls clear — Controls cleared. Fraud, sanctions, and applicable payee checks return an explicit outcome.
  4. Network admits the payment — Network accepted. The infrastructure validates participant, message, duplicate, and service timing.
  5. Settlement moves — Settled final. Funded settlement asset moves under the rail's model.
  6. Beneficiary posts — Beneficiary credited. The receiving participant validates and posts once.
  7. Outcome reconciles — End-to-end confirmed. Network, settlement, and participant posting evidence agree.
  8. Payer sees the proven result — Customer informed. The channel reports completed only when the authoritative state supports it.

Stress cases

Timeout with a late positive outcome

The payer-facing timer expires while settlement or receiver response is still in flight.

Last confirmed state
Network accepted; final settlement and beneficiary posting are not yet confirmed to the sender.
Settlement
Unknown until the settlement service returns authoritative evidence.
Funds and entries
The payer's debit or reservation stays under pending-outcome control; no second payment is created.
Next owner
Sending participant instant-payment operations
Safe action
Show a truthful pending or timed-out status, enquire with the original reference, and accept a late final outcome in state order.
Evidence required
  • End-to-end reference
  • Network timestamp
  • Settlement result
  • Receiver response
Recovery
  • Reconcile late outcome
  • Update the customer once
  • Release or retain postings according to final state

Insufficient prefunding

The sender lacks the settlement asset required by the instant service.

Last confirmed state
Payment admitted to the service; settlement not completed.
Settlement
Not settled unless the rail explicitly permits queueing and later confirms it.
Funds and entries
Receiver has no settled funds; sender must apply the rail-defined rejection or liquidity treatment.
Next owner
Participant treasury and instant-payment operations
Safe action
Apply the documented liquidity response and do not report completion.
Evidence required
  • Settlement balance
  • Liquidity threshold
  • Rail response
  • Customer posting state
Recovery
  • Fund the account for future payments
  • Close or retry only if the rail and customer contract permit

Settlement final, receiver core unavailable

The instant settlement movement is final but the receiving participant cannot post the beneficiary account.

Last confirmed state
Settlement final; beneficiary posting pending.
Settlement
Final under the rail's settlement rules.
Funds and entries
The receiving participant holds settled funds and owes controlled resolution to the beneficiary.
Next owner
Receiving participant payment and core-ledger operations
Safe action
Repair the beneficiary posting idempotently; do not resend or reverse the interbank payment without the formal exception path.
Evidence required
  • Settlement reference
  • Receiving account result
  • Posting journal
  • Customer-status history
Recovery
  • Restore core service
  • Post once
  • Report and reconcile the delay

Design decisions

What should the payer see when the service clock expires?

  • Failed — Simple screen, but dangerous when network or settlement outcome is unknown.
  • Pending or timed out — Truthfully preserves uncertainty and requires enquiry and later status update.

Use a state that represents evidence. A user-interface timer does not reverse network acceptance or settlement.

How should a real-time control outage behave?

  • Fail closed — Stops risk but reduces availability and may breach customer expectations.
  • Fail open — Preserves speed but creates explicit financial-crime or fraud exposure.
  • Risk-tiered fallback — Uses bounded, pre-approved criteria and requires measurable governance.

The choice is a governed risk decision. Return an explicit unavailable outcome and never let a timeout look like a clean approval.

Sources and disclosed synthesis