Insights Contact

Testing Playbook — Payments

Payments test levelsEach level gates the next; nothing reaches production without a passing parallel run1Unit & Component• Validation rules• Fee and FX math• Message mappingEnv: DevOwner: Engineering2API & Contract• Schemas and error codes• Idempotency keys• Auth and rate limitsEnv: Dev / QA with mocksOwner: QE3System Integration• Hub to fraud and sanctions• Ledger and network gateways• Return and reject codesEnv: QA + network simulatorsOwner: QE4End-to-End & UAT• Full customer journeys• Ops screens and reports• Cut-offs, holidays, month endEnv: UAT, production-like dataOwner: Business + QE5Parallel Run / Pilot• Old vs new give same result• Reconcile to the cent• Peak-volume performanceEnv: Pre-prod or limited prodOwner: Ops + Product6Go-live Gate• 0 open Sev 1 / Sev 2 defects• Compliance and AML sign-off• Rollback and kill switch readyEnv: ProductionOwner: All owners sign offTested at every levelSanctions · AML · Fraud · Idempotency · Reconciliation · Security · Performance · Audit trailPayments test levelsEach level gates the next; nothing reachesproduction without a passing parallel run1Unit & Component• Validation rules• Fee and FX math• Message mappingEnv: DevOwner: Engineering2API & Contract• Schemas and error codes• Idempotency keys• Auth and rate limitsEnv: Dev / QA with mocksOwner: QE3System Integration• Hub to fraud and sanctions• Ledger and network gateways• Return and reject codesEnv: QA + network simulatorsOwner: QE4End-to-End & UAT• Full customer journeys• Ops screens and reports• Cut-offs, holidays, month endEnv: UAT, production-like dataOwner: Business + QE5Parallel Run / Pilot• Old vs new give same result• Reconcile to the cent• Peak-volume performanceEnv: Pre-prod or limited prodOwner: Ops + Product6Go-live Gate• 0 open Sev 1 / Sev 2 defects• Compliance and AML sign-off• Rollback and kill switch readyEnv: ProductionOwner: All owners sign offTested at every levelSanctions · AML · Fraud · IdempotencyReconciliation · Security · PerformanceAudit trail
Free download
All 273 test cases in Excel
Priority, preconditions, test data and expected results, plus a defect log, status tracking and auto-calculated KPIs.
Download Excel (.xlsx)

Pick a topic from the panel to see the details.

Overview

Why a payments test playbook?

WhyIn payments a defect means money lost, duplicated or sent to the wrong party, plus regulatory risk. A standard playbook makes every release tested the same proven way.
BenefitsFewer production incidents · faster, repeatable releases · audit-ready evidence · new team members productive quickly.
WhenEvery new payment product, new rail, major release, vendor change or ISO 20022 migration.
WhoQE lead (owner) · engineering · product · operations · compliance.
HowRun the six levels in order; each level has entry and exit criteria; reuse the ready-made scenarios and add product-specific cases.
Green flags
  • Every level has written exit criteria
  • Negative and edge cases are first-class
  • Results tracked and reported each release
Red flags
  • Testing squeezed at the end of the project
  • Only happy paths tested
  • Sign-offs given without evidence
Reference

Purpose and scope

Rails covered: ACH (Nacha), wire (Fedwire, CHIPS, SWIFT MT/MX), cards (auth, capture, refund, chargeback), real-time (RTP, FedNow) and recurring/subscription payments.

Systems in scope: channels (web, mobile, API), payment hub/orchestration, fraud and sanctions screening, ledger/core banking, network gateways, notifications, reconciliation and reporting.

Out of scope unless agreed: third-party network certification (run by the network), and infrastructure DR drills owned by platform teams.

Reference

Test strategy

Test in five levels, each gating the next. Nothing reaches production without a passing end-to-end run on the network simulator and a signed-off parallel run.

LevelGoalEnvironmentOwner
Unit and componentValidation rules, fee and FX math, message mappingDevEngineering
API and contractRequest/response schema, error codes, idempotency keysDev/QA with mocksQE
System integration (SIT)Hub to fraud, sanctions, ledger, gatewayQA with network simulatorsQE
End-to-end and UATFull customer journeys, ops screens, reportsUAT with production-like dataBusiness + QE
Parallel run / pilotNew vs old system give the same resultPre-prod or limited productionOps + Product

Test data rules

  • Use synthetic or masked data only; never real PANs or account numbers outside PCI scope.
  • Keep a fixed set of test accounts: funded, zero balance, closed, frozen, dormant, joint, business.
  • Include sanctions test names (exact, fuzzy, alias) and high-risk countries.
  • Cover cut-off times, weekends, US bank holidays and month/quarter end.

Entry criteria: requirements signed off, environment stable, test data loaded, simulators connected.

Exit criteria: 100% of critical cases passed, zero open Sev 1/Sev 2 defects, reconciliation balanced to the cent, compliance sign-off received.

Level 1

Unit & Component

WhyCatches logic bugs where they are cheapest to fix.
BenefitsFast feedback; safe changes to money math.
WhenOn every commit and pull request.
WhoEngineers.
HowAutomated tests in CI for validation rules, fees, FX rounding and message mapping.
Green flags
  • Runs in CI on every pull request
  • Rounding and boundary values covered
Red flags
  • Low coverage on money calculations
  • Tests skipped to hit a deadline
Level 2

API & Contract

WhyCustomers and partners depend on stable APIs.
BenefitsNo breaking changes; clear, documented errors.
WhenEvery build, before release to sandbox.
WhoQE and API developers.
HowSchema and contract tests, idempotency replay, auth and rate-limit negatives.
Green flags
  • Duplicate requests create exactly one payment
  • All error codes documented
Red flags
  • Breaking schema change without versioning
  • Generic errors with no reason code
Level 3

System Integration

WhyMost payment defects hide between systems.
BenefitsIntegration issues found before business testing.
WhenAfter components are stable, every release.
WhoQE, integration developers, platform team.
HowNetwork simulators, sanctions test lists, every return and reject code, ledger checks.
Green flags
  • Every return and reject code tested
  • Simulators match the network specification
Red flags
  • Mocks only, no simulators
  • Sanctions screening stubbed out
Level 4

End-to-End & UAT

WhyConfirms the business and operations can actually run it.
BenefitsUser acceptance and operations readiness.
WhenAfter system integration exit criteria are met.
WhoBusiness users, operations, QE.
HowReal customer journeys, cut-off and holiday calendar, reports and ops screens.
Green flags
  • Business signs off on real scenarios
  • Ops team trained on new screens
Red flags
  • UAT done by developers only
  • No production-like data
Level 5

Parallel Run / Pilot

WhyProves the new system matches the old one with real volume.
BenefitsConfidence to switch; reconciliation and performance proof.
WhenBefore moving production traffic.
WhoOperations, product, finance / reconciliation.
HowRun both systems, compare outputs daily, reconcile to the cent, include a peak-volume day.
Green flags
  • Zero unexplained differences over the agreed period
  • Peak day included
Red flags
  • Differences waved away
  • Run cut short to meet a date
Level 6

Go-live Gate

WhyOne clear, accountable go / no-go decision.
BenefitsLaunch with every owner aligned and rollback ready.
WhenJust before production cut-over.
WhoProduct, engineering, QE, compliance, operations.
HowChecklist review meeting with recorded sign-offs from each owner.
Green flags
  • All checklist items signed
  • Rollback and kill switch rehearsed
Red flags
  • Open Sev 1 / Sev 2 accepted “to fix later”
  • Missing compliance or AML sign-off
Focus area

Cloud testing

WhyPayments in the cloud must survive zone failures, scale for peaks and stay secure. These risks do not show up in functional tests.
BenefitsProven failover and recovery · confidence at peak volume · security misconfigurations caught before go-live.
WhenSystem integration and parallel-run levels, plus every infrastructure change.
WhoQE, cloud / platform engineers, security, operations.
HowInfrastructure-as-code checks · load tests at peak volume · chaos tests (kill a node, zone or dependency) · failover and backup restore tests · cloud security scans (IAM, encryption, open ports) · cost checks at scale.
Green flags
  • Failover tested and measured against recovery targets
  • Load test at peak volume passed
  • Security scan clean before go-live
Red flags
  • DR plan exists only on paper
  • Test environment differs from production
  • Over-broad IAM roles or public storage buckets
Focus area

AI Agent testing

WhyAI is used two ways: to speed up testing, and inside payment products (fraud scoring, AML alerts, support agents). Both must be tested so AI never moves money wrongly.
BenefitsMore test cases in less time · better edge-case coverage · AI features proven safe, fair and explainable before launch.
WhenUse AI to generate tests at every level. Test AI features in system integration, UAT and parallel run.
WhoQE, data science / ML team, compliance and model risk, product.
HowAgents generate scenarios, test data and negative cases; humans review. For AI features: golden test sets, accuracy and false-positive targets, prompt-injection and misuse tests, guardrails on actions, explainability checks, monitoring for drift.
Green flags
  • AI cannot release or approve a payment without a human
  • Golden test set with clear pass thresholds
  • Every AI decision logged and explainable
Red flags
  • AI-generated tests run without review
  • No limits on what an agent can do in production
  • Model changes deployed without re-testing
Focus area

Notification testing

WhyStatus messages drive customer actions. A wrong, late or duplicate notification can trigger a duplicate payment or a complaint.
BenefitsCorrect, timely status for every payment · fewer support calls · no data leaks in messages.
WhenAPI and contract, system integration and end-to-end levels.
WhoQE, developers, product, compliance (message content).
HowTest every status (initiated, sent, settled, returned, failed) on every channel: webhooks, email, SMS, push, ops alerts. Check signature validation, retries, duplicates, out-of-order delivery, timing and content.
Green flags
  • Each status sent exactly once, in order
  • Webhook retries and signature checks pass
  • No PII or full account numbers in messages
Red flags
  • Notifications tested only on the happy path
  • Duplicate events after retries
  • Messages with no reason for a failure
Reference

Ready-made test scenarios

Each rail lists the cases that catch most production defects. Run the positive case first, then the negatives.

ACH (Nacha)

IDScenarioExpected result
ACH-01Same-day credit submitted before cut-offFile built, settles same day, ledger posted
ACH-02Credit submitted after cut-offRolls to next business day
ACH-03Debit with insufficient fundsR01 return processed, customer notified
ACH-04Account closed / invalid accountR02 / R03 / R04 returns mapped correctly
ACH-05Unauthorized debit disputeR10 handled within Nacha time limit
ACH-06NOC (notification of change) receivedAccount details updated, no duplicate post
ACH-07PrenoteZero-dollar entry sent, live entries held as configured
ACH-08File formatBatch headers, control totals and hash valid

Wire (Fedwire, CHIPS, SWIFT)

IDScenarioExpected result
WIR-01Domestic Fedwire, valid ABAISO 20022 pacs.008 built, IMAD/OMAD captured
WIR-02International SWIFT MX with intermediary bankCorrect routing chain, charges (OUR/SHA/BEN) applied
WIR-03FX wireRate locked, quote expiry enforced, amounts match
WIR-04Sanctions hit on beneficiaryPayment held for review, not released
WIR-05Wire after Fedwire cut-offQueued for next business day
WIR-06Recall / return (camt.056, pacs.004)Status updated, funds returned to account
WIR-07Dual approval above limitSingle user cannot release
WIR-08Tracking (gpi / UETR)End-to-end status visible to customer

Cards

IDScenarioExpected result
CRD-01Authorization approvedHold placed, available balance reduced
CRD-02Decline codes (insufficient funds, expired, CVV fail)Correct response code and customer message
CRD-03Partial and multiple capturesCaptured total never exceeds auth
CRD-04Full and partial refundCredit posted, original transaction linked
CRD-05Reversal / void before settlementHold released
CRD-06Chargeback and representmentCase created, deadlines tracked
CRD-073-D Secure challengeLiability shift flagged
CRD-08Tokenized wallet payment (Apple Pay, Google Pay)Token mapped to correct card

Real-time (RTP, FedNow)

IDScenarioExpected result
RTP-01Credit transfer 24/7, including weekendAccepted or rejected within network timeout
RTP-02Amount above network or bank limitRejected with correct reason code
RTP-03Receiver timeoutNo duplicate debit, status resolved
RTP-04Request for PaymentPayer can accept or decline
RTP-05IrrevocabilityNo reversal path; only a return request

Recurring and subscription

IDScenarioExpected result
REC-01Schedule on 29th, 30th or 31stRuns on correct date in short months
REC-02Schedule falls on weekend or holidayMoves per business rule
REC-03Failed payment retryRetries per policy, then stops and notifies
REC-04Customer cancels or pausesNo further charges
REC-05Card expired / updatedAccount updater applied, schedule continues
REC-06Amount change mid-cycleProration or next-cycle rule applied
Test scenarios

Holidays and cut-offs

Calendar and timing cases. Cut-offs are evaluated in Eastern Time unless the product says otherwise.

IDScenarioExpected result
CAL-01ACH or wire dated on a US Federal Reserve holidayNot sent that day; value date moves to next business day; customer shown the new date before confirming
CAL-02Payment initiated on a weekendQueued and released when the next business-day window opens
CAL-03Recurring payment falls on a holidayMoves to prior or next business day per product rule, the same way every time
CAL-04Submitted one minute after Same Day ACH cut-offGoes to next window or next day; customer told before final confirmation
CAL-05Wire submitted after Fedwire customer cut-offQueued for next business day, or released only with an approved ops exception
CAL-06Customer in another time zone submits near cut-offCut-off checked in Eastern Time; result consistent across channels
CAL-07RTP / FedNow payment on a holidaySettles immediately (24/7/365); ledger posting and statement dates follow the bank’s rule
CAL-08Daylight saving time change dayScheduled jobs and payments run once, not skipped or doubled
CAL-09New holiday added to the calendarFuture-dated payments re-dated and affected customers notified
CAL-10Month-end and quarter-end peakCut-offs honored and processing SLA met at peak volume
Test scenarios

System failures (FX and dependencies)

What must happen when a service the payment depends on fails. The rule: never lose, duplicate or release an unchecked payment.

IDScenarioExpected result
FXF-01FX rate service down when customer starts a cross-border paymentNo stale rate used; customer sees “rate unavailable, try again”; no debit taken
FXF-02FX quote expires before customer confirmsCustomer must accept a new quote
FXF-03FX booking fails after account debit, before sendingPayment held in suspense and retried; if still failing, debit reversed and customer notified; never debited twice
FXF-04Rate moves beyond tolerance between quote and bookingRe-quote or reject per policy; customer never charged an unagreed rate
FXF-05Currency not supported or restrictedRejected at validation with a clear reason
FXF-06Sanctions screening service times outPayment held (fail closed), not released unscreened
FXF-07Core ledger unavailablePayments queued; no partial postings; processed in order when ledger returns
FXF-08Network gateway times out after sendingStatus set to “pending confirmation” and resolved by inquiry; payment not resent blindly
FXF-09Fraud scoring service downFallback rules applied (for example lower limits or manual review); event logged
Test scenarios

Internal transfers and notifications

On-us (book) transfers inside the bank and the notifications around them.

IDScenarioExpected result
INT-01Transfer between the customer’s own accountsDebit and credit posted together instantly; both balances updated; no network fee
INT-02Transfer to another customer at the same bankPayee validated; posted on-us; payer and payee each notified
INT-03Internal transfer with insufficient fundsRejected before posting with a clear message
INT-04Success notification contentSent once per party; correct amount and date; account number masked
INT-05Failed transfer notificationReason given; no “success” message ever sent
INT-06Notification service down during transferTransfer still completes; messages queued and sent once when service returns
INT-07Customer opted out of SMSOnly the channels the customer allows are used
INT-08Webhook to a corporate clientSigned event; retried on errors; client receives it exactly once (idempotent)
INT-09Ops alert for stuck transferAlert raised after SLA breach and routed to the on-call team
Test scenarios

High-value payments

Controls that protect large payments.

IDScenarioExpected result
HVP-01Amount above the user’s single-payment limitNeeds a second approver (maker-checker)
HVP-02Amount above the company daily limitBlocked or sent for limit-increase approval
HVP-03User tries to approve their own paymentBlocked
HVP-04High value triggers call-back verificationHeld until verification is recorded
HVP-05Approval not completed before cut-offRolls to next day or expires per rule; initiator notified
HVP-06High value to a newly added beneficiaryExtra fraud check or cooling-off period applied
HVP-07Wire of $3,000 or moreTravel Rule originator and beneficiary details captured and sent
HVP-08Large FX amountPriced by treasury / dealer desk, not the standard retail rate
Test scenarios

ACH to wire upgrade

Moving a payment from one rail to another without duplicating it.

IDScenarioExpected result
RTE-01Same-day payment above the Same Day ACH per-payment limitCustomer offered a wire instead; wire fee shown before confirming
RTE-02Customer upgrades an ACH payment to wire before the ACH file is sentACH entry cancelled; exactly one wire sent; customer charged once
RTE-03Upgrade requested after the ACH file was already sentUpgrade blocked; customer told to wait for the ACH result or request a reversal
RTE-04Wire fee higher than ACH feeCustomer sees and accepts the new fee before conversion
RTE-05Beneficiary bank’s wire ABA differs from its ACH routing numberWire routing number validated separately
RTE-06Wire fails; fallback to ACHOnly with customer consent; new value date shown
RTE-07Receiving bank not enabled for RTP or FedNowRouted to an available rail per rules; customer told of the timing change
Test scenarios

Account types and status

Payments from and to different account types and states.

IDScenarioExpected result
ACC-01Business vs consumer accountCorrect ACH SEC code (CCD vs PPD), limits and fees applied
ACC-02Savings account debitBank’s savings transfer-limit policy applied
ACC-03Joint accountInitiation follows the account mandate; owners notified per their preferences
ACC-04Inactive / dormant accountOutgoing payment blocked; customer asked to reactivate
ACC-05Closed accountRejected; ACH return R02 when received
ACC-06Frozen account (legal hold or garnishment)Debits blocked; ops alerted
ACC-07Account holder deceasedOutgoing blocked; ACH returns R14 / R15 handled
ACC-08Insufficient fundsRejected, or overdraft used only if enabled; NSF fee per policy
ACC-09Funds on hold (uncollected deposit)Only the available balance can be used
ACC-10High-risk account (enhanced due diligence)Payments above threshold held for compliance review
ACC-11Custodial or minor’s accountCustodian approval rules enforced
ACC-12Foreign currency accountDebited in account currency; FX applied only when currencies differ
Test scenarios

Risk and compliance

Sanctions, AML and fraud cases. Customer messages stay neutral so no one is tipped off.

IDScenarioExpected result
RSK-01Beneficiary in a comprehensively sanctioned countryPayment blocked or held; compliance alerted; neutral customer message
RSK-02Beneficiary in a high-risk (non-sanctioned) countryExtra review; purpose of payment required
RSK-03Fuzzy name match on the OFAC SDN listHeld; analyst clears or escalates with full audit trail
RSK-04Several payments just under a reporting threshold (structuring)AML alert raised
RSK-05Sudden volume or amount spike vs customer profileAlert or step-up authentication
RSK-06Money in then quickly out (mule pattern)AML alert; outgoing held per rule
RSK-07New device + new payee + large amountStep-up authentication or block
RSK-08Politically exposed person (PEP) as beneficiaryEnhanced due diligence review
RSK-09Cross-border payment missing purpose code or remittance infoRejected or more information requested
RSK-10Account under compliance holdNo payments released until hold is cleared
Test scenarios

Special characters and data quality

Each rail allows a different character set: SWIFT MT uses the restricted FIN “X” set, ISO 20022 allows more but needs XML escaping, and Nacha files are fixed-width ASCII. Test every free-text field: names, addresses, references and remittance info.

IDScenarioExpected result
CHR-01Accented letters in beneficiary name (José Muñoz, Müller)Transliterated per scheme rule (for example Jose Munoz, Mueller) or accepted where allowed; screening runs on both forms
CHR-02Non-Latin script name (Chinese, Arabic, Cyrillic)Rejected at entry with a clear message, or transliterated per policy; never sent garbled
CHR-03Ampersand in name (Smith & Sons LLC)Escaped as & in ISO 20022 XML; replaced (for example “and”) for SWIFT MT
CHR-04Apostrophe or hyphen (O’Brien, Smith-Jones)Accepted and preserved
CHR-05Characters outside the SWIFT X set (such as @ # ! _ and semicolon) in an MT fieldReplaced or rejected before sending; no NAK from the network
CHR-06Reference starting or ending with “/” or containing “//”Rejected or corrected at validation (not allowed in SWIFT reference fields)
CHR-07XML-special characters (< > “) in remittance infoEscaped correctly; message stays valid against the XSD schema
CHR-08Emoji or control characters pasted into a memo fieldStripped or rejected; no file or message corruption
CHR-09Name longer than the field limit (35 per MT line, 140 for ISO 20022 name)Clear error at entry, or truncated only by an agreed rule; screening uses the full name
CHR-10Lowercase or mixed case in Nacha file fieldsConverted to the case the bank’s Nacha spec requires; file passes validation
CHR-11Leading / trailing spaces and double spacesTrimmed; no false sanctions mismatch
CHR-12Injection-style input (‘ OR 1=1, <script>)Treated as plain text; no error leak; security event logged
Test scenarios

Address validation (ISO 20022)

Since SWIFT Standards Release 2026 (14 November 2026), CBPR+ cross-border messages accept only fully structured or hybrid postal addresses, with Town Name and Country mandatory. Fully unstructured addresses are rejected by the network.

IDScenarioExpected result
ADR-01Fully structured address (street, building, postal code, town, country)Accepted; each part in its own XML tag; no AdrLine
ADR-02Hybrid address (TwnNm + Ctry structured, rest in AdrLine)Accepted; AdrLine within the allowed limit (max 2 lines of 70 characters)
ADR-03Fully unstructured address (AdrLine only)Blocked before sending; customer or ops asked to complete town and country; no network rejection
ADR-04Town name missingValidation error at entry; payment not created
ADR-05Invalid country code (USA, UK, XX)Rejected; only ISO 3166 two-letter codes (US, GB) accepted
ADR-06Placeholder text (“NOTPROVIDED”, “N/A”, “UNKNOWN”)Rejected; real values required
ADR-07Postal code format wrong for the country (US ZIP not 5 or 9 digits)Warning or rejection per policy
ADR-08Postal code and country do not matchFlagged for correction
ADR-09Hybrid address where structured town differs from town in AdrLineStructured data takes precedence; mismatch flagged for data-quality review
ADR-10Legacy beneficiary record with free-text addressAddress service structures it (town + country) and asks the user to confirm before first use
ADR-11Address in a sanctioned region inside a non-sanctioned countryScreening checks town and subdivision, not just country; payment held
ADR-12SWIFT MT option 50F / 59F structured partyLine 3 carries country code and town in the correct format; maps cleanly to ISO 20022
ADR-13PO Box only addressAccepted or rejected per policy; town and country still required
Test scenarios

SWIFT MT and ISO 20022 messages

Test each message type end to end and its MT-to-ISO 20022 mapping where both still exist in your flow (for example corporate MT101 in, ISO 20022 out). Fedwire and SWIFT cross-border now run on ISO 20022.

IDScenarioExpected result
MSG-01MT101 request for transfer from a corporate (single and multiple transactions)Each transaction validated and executed; maps to pain.001 / pacs.008; status returned to the corporate
MSG-02MT101 with one invalid transaction in a batchOnly that transaction rejected (or whole batch, per agreement); clear status per transaction
MSG-03MT102 multiple customer credit transferEach transaction split into its own single-transaction pacs.008, as CBPR+ requires; totals reconcile
MSG-04MT103 single customer credit transferMaps to pacs.008; fields 50 / 59 / 70 / 71A carried correctly; no data truncated
MSG-05Charges option OUR / SHA / BEN (71A, ChrgBr)Correct fees deducted or billed; amounts on both sides match
MSG-06MT202 / MT202 COV bank-to-bank cover paymentMaps to pacs.009 / pacs.009 COV; underlying customer details travel with the cover for screening
MSG-07pacs.002 status report (accepted, rejected, pending)Status shown to customer and ops; reason codes mapped to clear messages
MSG-08Return (pacs.004)Funds returned to the original account; return reason shown; fees per policy
MSG-09Recall request (camt.056) and answer (camt.029)Recall tracked to resolution; funds credited back only on positive answer
MSG-10Debit / credit confirmations (MT900 / MT910, camt.054)Booked once; matched to the original payment
MSG-11Statements (MT940 / MT950, camt.053)Every payment appears; balances reconcile to the ledger
MSG-12Free-format messages (MT199 / MT299)Routed to ops queue; never auto-processed as payments
MSG-13UETR on every paymentSame UETR kept end to end, including returns and recalls; gpi tracking works
MSG-14Duplicate message received (same reference and UETR)Detected and not booked twice
MSG-15Schema-invalid ISO 20022 messageRejected before sending with the failing element named
Test scenarios

Payment files and mapping

Files and data flowing in from upstream systems (ERP, corporate clients, channels) and out to downstream systems (networks, ledger, reconciliation, reporting). Every field must map correctly and every file must be accounted for.

IDScenarioExpected result
FIL-01Upstream bulk file intake (pain.001, CSV, Nacha, MT101)File accepted, parsed and each payment created once; acknowledgement returned
FIL-02Duplicate file received (same name, hash or control totals)Detected and rejected or held; no payment created twice
FIL-03Header / trailer and control totals mismatchWhole file rejected per rule; clear error returned to sender
FIL-04One bad record in a valid fileBad record rejected (or whole file, per client agreement); others processed; status per record
FIL-05Empty, truncated or corrupted fileRejected safely; no partial processing; ops alerted
FIL-06File encoding and line endings (UTF-8 with BOM, CRLF vs LF)Parsed correctly; special characters preserved or transliterated per rule
FIL-07File arrives after cut-offProcessed for next business day or held per rule; sender notified
FIL-08Encryption, signature and secure transfer (PGP, SFTP)Invalid signature or key rejected; file not processed
FIL-09Large file (peak volume)Processed within SLA; no timeouts or memory errors
FIL-10Field-by-field mapping upstream to internal to downstreamEvery field lands in the right target field; mapping sheet signed off
FIL-11Code-list mapping (purpose codes, SEC codes, charge bearer, return codes)Each source code maps to the correct target code; unknown codes go to an error queue
FIL-12Mandatory target field missing in sourceDefaulted only by the documented rule, otherwise rejected; never sent blank
FIL-13References preserved end to end (EndToEndId, UETR, client reference)Same references appear in network message, status reports, statements and ERP
FIL-14Status and reports returned upstream (pain.002, camt.054, ACK / NACK files)Client receives status for every payment in the agreed format and time
FIL-15Downstream ledger and GL posting mappingCorrect GL accounts, debit / credit sides, fees and FX booked; ledger balances
FIL-16Downstream reconciliation and reporting files (camt.053, BAI2, data warehouse, regulatory)All payments present; totals match the ledger and network settlement
FIL-17Mapping version change (new field or format release)Regression suite passes; old and new versions handled during transition
Test scenarios

Currency logic

ISO 4217 currency codes, minor units, rounding and FX direction. US domestic rails (ACH, RTP, FedNow, Fedwire) settle in USD only.

IDScenarioExpected result
CUR-01Currency with 0 decimals (JPY) and 3 decimals (KWD, BHD)Amounts stored and sent with the correct minor units; no extra or missing decimals
CUR-02Amount with more decimals than the currency allows (USD 10.005)Rejected at entry, or rounded only by the documented rule
CUR-03Rounding rule on FX conversion (half-up vs half-even)Result matches the bank’s documented rounding rule at every step
CUR-04Buy vs sell rate and inverse rateCorrect side of the quote used; inverse rate gives the same result within tolerance
CUR-05Cross-currency via USD (triangulation, e.g. GBP to JPY)Both legs use the agreed rates; final amount matches the quoted amount
CUR-06FX rate precision (6 or more decimals)Rate not truncated in storage, messages or statements
CUR-07Payment currency differs from account currencyFX applied once; customer sees both amounts and the rate before confirming
CUR-08Non-USD amount on a USD-only rail (ACH, RTP, FedNow, Fedwire)Rejected or routed to a cross-border rail
CUR-09Invalid or lowercase currency code (usd, US$, XXX)Rejected; only valid upper-case ISO 4217 codes accepted
CUR-10Amount display by locale (1,234.56 vs 1.234,56)Display follows locale; stored and sent values are unchanged
CUR-11Zero, negative or blank amountRejected at entry with a clear message
CUR-12Amount above the rail’s maximum field lengthRejected before sending; no truncation
Test scenarios

Calculation logic

Fees, totals, balances and limits must be correct to the cent, every time.

IDScenarioExpected result
CLC-01Flat fee, percentage fee and tiered feeFee matches the fee schedule for each tier boundary
CLC-02Percentage fee with minimum and maximum capFee never below the minimum or above the cap
CLC-03Fee rounding to the centRounded once, by the documented rule; no sub-cent residue in the ledger
CLC-04Fee waiver (premium client, promo, internal transfer)Fee set to zero only when the waiver rule applies
CLC-05FX margin / markupCustomer rate = market rate adjusted by the configured margin; margin booked to revenue
CLC-06Total debit = amount + fees (+ FX cost)Customer shown and debited the same total
CLC-07Amount received under charge options OUR / SHA / BENBeneficiary amount = sent amount minus the right deductions
CLC-08Available balance = ledger balance – holds + overdraft linePayment allowed only up to the available balance
CLC-09Daily limit summed across channels (web, mobile, API)Cumulative total enforced; resets at the documented time
CLC-10Fees reversed when a payment fails or is returnedFee refunded per policy; no fee charged twice
CLC-11Partial refunds and split paymentsSum of parts never exceeds the original; totals reconcile
CLC-12Nacha batch and file control totalsEntry count, total debits/credits and entry hash match the entries
CLC-13Large amounts and decimal precisionNo floating-point error (for example 0.1 + 0.2); exact decimal arithmetic
CLC-14Interest or proration on recurring and scheduled paymentsCalculated amount matches the formula and day-count rule
Test scenarios

Payment status

Every payment follows a defined lifecycle: Draft, Pending approval, Approved, Scheduled, Sent, Accepted, Settled / Completed, with exits to Rejected, Returned, Cancelled, Failed and On hold.

IDScenarioExpected result
STS-01Happy-path lifecycle (Draft to Settled)Each status reached in order with a timestamp and user / system actor
STS-02Invalid transition (Settled to Cancelled, Rejected to Sent)Blocked by the system; error logged
STS-03Cancel before releaseStatus Cancelled; no message sent; no debit or debit reversed
STS-04Cancel after releaseCancel refused; customer offered recall where the rail allows
STS-05Network status codes mapped (pacs.002 ACCP / ACSC / RJCT / PDNG)Each code maps to the right internal and customer status
STS-06Completed payment later returned (ACH R-code, pacs.004)Status moves to Returned; funds and fees handled; customer notified
STS-07Compliance hold and releaseStatus On hold, then resumes to Sent or ends Rejected; history kept
STS-08Same status event received twiceProcessed once; no duplicate notification or posting
STS-09Events arrive out of order (Settled before Accepted)Final status is correct; older event does not overwrite a newer one
STS-10Payment stuck in one status past SLAAlert raised to ops; customer sees an accurate pending message
STS-11Status consistent everywhereUI, API, webhook, notification and statement show the same status
STS-12Approval expires before actionStatus Expired / Cancelled per rule; initiator notified
STS-13Future-dated paymentShows Scheduled until the value date, then follows the normal lifecycle
STS-14Customer-facing status labelsPlain-language labels with a reason for Rejected, Returned and On hold (neutral wording for compliance holds)
Reference

Cross-cutting tests

Run these on every rail. Most costly payment incidents come from these areas, not from the happy path.

AreaWhat to testPass condition
Sanctions screening (OFAC)Exact, fuzzy and alias name matches; country and BIC hitsTrue hits held; false-positive rate within agreed target
AML monitoringStructuring, velocity, rapid in-and-out, high-risk geographyAlerts generated with correct rule and risk score
FraudAccount takeover, new payee plus large amount, device changeStep-up or block triggered
Idempotency and duplicatesSame request sent twice, retry after timeoutExactly one payment, one ledger entry
ReconciliationHub vs ledger vs network settlement fileBalanced to the cent; breaks reported
Limits and entitlementsPer-transaction, daily and user-role limitsOver-limit rejected; maker-checker enforced
SecurityAuth tokens, PII masking, PCI scope, API rate limitsNo data leak; OWASP top 10 clean
PerformancePeak volume (for example payroll day, month end)Throughput and latency meet SLA
ResilienceGateway down, ledger slow, network reject stormPayments queue safely and resume; no loss or duplicate
Audit trailEvery create, approve, release, cancelUser, time and before/after values logged
NotificationsEmail, SMS, push, webhooks per statusCorrect status, sent once, no PII in plain text
Reference

Defects, metrics and go-live

Defect severity

SeverityPayments meaningExample
Sev 1Money lost, duplicated or sent to the wrong party; compliance breachDuplicate wire released; sanctions hit not held
Sev 2Payment blocked or wrong amount, no workaroundFX rate applied wrongly
Sev 3Workaround existsStatus shows late in UI
Sev 4CosmeticLabel typo

Metrics to report

  • Test cases planned, executed and passed, by rail
  • Open defects by severity, and defect leakage to UAT and production
  • Requirement coverage, including every return and reject code
  • Reconciliation breaks found and fixed
  • Automation coverage for regression

Go-live checklist

  • ☐ All critical and regression cases passed
  • ☐ Zero open Sev 1 / Sev 2 defects
  • ☐ End-to-end run on network simulator or certification passed
  • ☐ Parallel run matched, reconciliation balanced
  • ☐ Compliance and AML sign-off received
  • ☐ Performance test met peak-volume SLA
  • ☐ Rollback plan tested and kill switch ready
  • ☐ Ops runbook, monitoring alerts and support contacts in place
  • ☐ Business and product owner sign-off
Test scenarios

Automation testing

What to automate so every release is checked the same way, fast. Aim to automate stable, high-risk, repeatable checks; keep exploratory and usability testing manual.

IDScenarioExpected result
AUT-01Regression suite runs in CI on every buildBuild blocked when a P1 test fails; results published to the team
AUT-02API and contract tests automated (schemas, error codes, idempotency)Breaking API change caught before merge
AUT-03Test data created and reset automaticallyEach run starts from a known state; no dependency on leftover data
AUT-04Network simulators / service virtualization for ACH, Fedwire, SWIFT, RTPTests run without real networks; every return and reject code can be triggered
AUT-05File-based tests automated (generate, submit, compare output)Generated Nacha / pain.001 files and resulting messages compared field by field to expected
AUT-06Smoke test after every deploymentCore flows (create, approve, send, status) pass in each environment within minutes
AUT-07Automated reconciliation checkNightly job compares hub, ledger and settlement totals and flags breaks
AUT-08Flaky test trackingFlaky tests quarantined and fixed; flake rate reported
AUT-09Results linked to test management and defectsEach automated run updates test case status and links failures to defect IDs
AUT-10Automation coverage measuredCoverage of P1 cases and API endpoints reported each release
Test scenarios

Performance testing

Prove the platform handles real payment volumes and peaks within SLA, and fails safely when overloaded.

IDScenarioExpected result
PRF-01Load test at expected peak (payroll day, month end)Throughput and latency within SLA; zero lost or duplicate payments
PRF-02Stress test beyond peak to find the breaking pointSystem degrades gracefully (queues, throttles); recovers without data loss
PRF-03Soak / endurance test (24 hours or more)No memory leaks, growing queues or slowdowns over time
PRF-04Spike test (sudden burst)Auto-scaling or queuing absorbs the spike; SLA recovers quickly
PRF-05Batch window (ACH file build, end-of-day processing)Batch completes before the cut-off with margin
PRF-06API latency percentiles (p50, p95, p99)Within agreed targets under load
PRF-07Instant payments end-to-end time (RTP / FedNow)Response well inside the network timeout under load
PRF-08Concurrent payments on the same accountNo race conditions; balance and limits stay correct
PRF-09Slow downstream dependency under load (screening, FX, ledger)Timeouts and retries behave as designed; no pile-up or duplicates
PRF-10Failover under loadTraffic switches to standby; in-flight payments complete or are safely retried
PRF-11Database and storage growthQuery times stay within target with 12 months of data
Test scenarios

Security testing

Payments are a prime fraud and attack target. Test that only the right people can move the right money, and that data stays protected.

IDScenarioExpected result
SEC-01Authentication and MFA / step-up for high-risk actionsNew payee, large amount and limit changes require step-up
SEC-02Authorization: access another customer’s payment or account (IDOR)Access denied; attempt logged and alerted
SEC-03Maker-checker bypass attemptsSame user cannot create and approve; role changes take effect immediately
SEC-04Payment tampering between approval and sendingAmount or beneficiary change after approval forces re-approval; message integrity checked
SEC-05Injection and input attacks (SQL, XML, script)Inputs treated as data; no execution or error leaks
SEC-06API security (token expiry, scopes, rate limits, mTLS)Expired or wrong-scope tokens rejected; bursts throttled
SEC-07Replay attack (same signed request or webhook sent again)Rejected using nonce / timestamp / idempotency key
SEC-08Webhook and file signature verificationUnsigned or wrongly signed payloads rejected
SEC-09Encryption in transit and at restTLS 1.2+ only; sensitive data encrypted in databases, files and backups
SEC-10PII and card data masking (UI, logs, notifications, test data)Full account and card numbers never appear in logs or messages
SEC-11Session management (timeout, concurrent sessions, logout)Idle sessions expire; tokens invalidated on logout
SEC-12Audit log integrityLogs cannot be edited or deleted by application users; every action recorded
SEC-13Vulnerability scanning and penetration testNo open critical or high findings at go-live
SEC-14Secrets managementNo keys or passwords in code, config or logs; rotation works
Test scenarios

A/B testing

Experiments on payment journeys (screens, copy, flows) to improve conversion and experience. The money logic must be identical in every variant; only the experience changes.

IDScenarioExpected result
ABT-01Same payment, both variantsAmount, fees, FX rate, routing and status identical in A and B
ABT-02Stable variant assignmentA user stays in the same variant across sessions and devices
ABT-03Required disclosures present in every variantFees, FX rate, delivery time and other required disclosures shown in A and B
ABT-04Feature flag kill switchVariant can be switched off instantly; users fall back to control safely
ABT-05Metrics captured correctlyConversion, drop-off, errors and time to complete tracked per variant
ABT-06Guardrail metrics monitoredFraud rate, payment failures and support contacts do not worsen in the test variant
ABT-07No sensitive data sent to analytics toolsAccount numbers, names and amounts masked or excluded from analytics events
ABT-08High-risk controls excluded from experimentsSanctions, AML, limits and approvals never vary by variant
ABT-09Sample size and significance before decisionDecision made only after the planned sample size and significance level
ABT-10Rollout and cleanupWinning variant rolled out gradually; losing code and flags removed
Test scenarios

DR testing

Disaster recovery: prove payments can continue from the recovery site within the agreed RTO (time to recover) and RPO (data you can afford to lose), with no lost or duplicate payments.

IDScenarioExpected result
DRT-01Planned failover of the payment platform to the DR siteService restored within RTO; all configuration, keys and certificates work at DR
DRT-02Unplanned failover during live processingIn-flight payments completed or safely retried; zero duplicates or losses
DRT-03Data loss within RPOLast committed payments present at DR; gap (if any) within RPO and reconciled
DRT-04Network connectivity from DR (Fedwire, ACH operator, SWIFT, RTP / FedNow)All network connections, certificates and IP allowlists work from DR
DRT-05Files and schedules after failoverNo file sent twice or missed; scheduled and recurring payments run once
DRT-06Cut-off management during an outageUrgent payments prioritized; customers and ops informed of delays
DRT-07Reconciliation after failoverHub, ledger and network totals match after recovery
DRT-08Failback to primaryReturn to primary without data loss or duplicates; DR changes synced back
DRT-09Backup restoreBackups restore cleanly and within target time; restored data verified
DRT-10Runbook and communicationRunbook followed step by step; roles, contacts and customer messages work
Reference

UAT vs end-to-end (E2E) testing

Both run on full journeys, but they answer different questions: E2E asks “do all the systems work together?”; UAT asks “does this do the job the business needs?”.

E2E testingUAT
GoalProve systems integrate correctly from channel to network to ledgerProve the product meets business needs and is ready to use
Who runs itQE / test engineersBusiness users, operations, compliance, sometimes clients
Written inTechnical steps, messages, data checksBusiness language and real tasks
DataSynthetic, covering every path and codeProduction-like, realistic business scenarios
FocusFlows, mappings, statuses, reconciliation, negative pathsUsability, workflow fit, reports, wording, sign-off
AutomationMostly automatedMostly manual, guided by scripts
OutputPass / fail evidence and defectsFormal business sign-off (go / no-go input)
ExampleWire created in UI produces correct pacs.008, status and ledger entriesTreasury user pays a supplier and confirms the confirmation and statement make sense
Test scenarios

UAT test cases

Business-facing acceptance tests run by business users and operations, in business language, on production-like data. They confirm the product does the job; E2E tests confirm the systems work together.

IDScenarioExpected result
UAT-01Treasury user pays a supplier by wire using a saved beneficiaryPayment completes; confirmation and statement entry are correct and understandable
UAT-02Payroll file uploaded and approved by two approversAll employees paid on payday; totals match payroll report
UAT-03Customer sets up, changes and cancels a monthly paymentSchedule works as expected; no payment after cancellation
UAT-04Operations investigates and repairs a rejected paymentOps can find the reason, repair and resend within the ops screens
UAT-05Compliance analyst reviews and clears a sanctions holdCase shows all needed information; decision recorded with audit trail
UAT-06Customer service answers “where is my payment?”Agent sees status, tracking and next step in one screen
UAT-07Finance reconciles the dayReports match ledger and bank statements without manual workarounds
UAT-08Corporate client receives status reports and statements in their ERPClient confirms format and content are usable
UAT-09Fees and FX shown before confirming match what is chargedBusiness confirms disclosure and charges are correct
UAT-10Business sign-off on reports and notifications wordingWording clear, branded and compliant
Reference

Shift Left, AI First, UI Next — and Data First, AI Next

Two simple ideas shape how this playbook is used. Read the full article: Shift Left, AI First, UI Next.

IdeaWhat it means for payments testing
Shift LeftTest from day one: validate ISO 20022 mapping, sanctions data and fee rules in design and API tests, not after go-live.
AI FirstAI drafts test cases, test data and defect triage notes; testers review and approve. High-risk decisions stay with people.
UI NextTest the capability through the API first, then each channel (screens, chat, files) as a thin layer on top.
Data FirstGet the data right before adding AI: clean reference data, structured addresses, correct mappings, masked and realistic test data.
AI NextOnce data is trusted, use AI for anomaly detection, smart test selection, log triage and predicting risky releases.

Guardrail: AI suggestions are never the final approver for money movement, compliance decisions or go-live sign-off.

Test management

Defect triage process

Triage turns a raw defect into a clear decision: how bad it is, who fixes it, and by when.

Steps

  1. Log with steps, expected vs actual, test case ID, evidence and payment reference.
  2. Check for duplicates and confirm it is a real defect (not data or environment).
  3. Classify severity (business impact) and priority (fix order).
  4. Assign an owner and target date.
  5. Decide: fix now, fix next release, accept risk, not a defect, or duplicate.
  6. Retest the fix and run the related regression cases.
  7. Close with root cause recorded.
SeverityMeaningFix targetGo-live
Sev 1Money lost, duplicated or misrouted; compliance breach; no workaroundSame dayBlocker
Sev 2Payment blocked or wrong, workaround is hard1–2 daysBlocker
Sev 3Workaround exists; limited impactNext sprintAllowed with sign-off
Sev 4CosmeticBacklogAllowed

Cadence: daily triage call during test cycles (QE lead, dev lead, product owner, ops / compliance when needed). AI first: AI can suggest duplicates, severity and likely root cause from logs; the triage team decides.

Test management

Testing KPIs

Track a small set of KPIs every day of a test cycle. The Excel download calculates these automatically.

KPIFormulaExample target
Execution progressExecuted / planned (excluding N/A)100% before go-live
Pass ratePassed / executed95% or higher
P1 pass rateP1 passed / P1 executed100%
Open Sev 1 / Sev 2Count of open defects0 at go-live
Defect densityDefects per 100 executed testsTrending down
Defect leakageDefects found after release / total defectsBelow 5%
Retest pass rateRetests passed / retests run90% or higher
Rejected defect %Not-a-defect / total loggedBelow 10%
Automation coverageAutomated P1 cases / total P1 casesRising each release
Mean time to fix (Sev 1/2)Average hours from logged to fixedWithin SLA
Test management

Sample reports

Sample report with illustrative figures, to show the format.

1. Daily test status report (Shift Left)

Planned
273
Executed
214
Passed
196
Failed
12
Blocked
6
AreaExecutedPass rateOpen defectsRAG
ACH40 / 4295%2 (Sev 3)Green
Wire / SWIFT33 / 3888%1 Sev 2, 2 Sev 3Amber
Files and mapping12 / 1783%1 Sev 2Amber
Risk and compliance10 / 10100%0Green

Top risks: FX booking retry not yet fixed (Sev 2). Next 24h: retest FIL-11 and WIR-06; start DR exercise.

2. AI-assisted defect triage summary (AI First)

DefectAI suggestionHuman decision
D-118 pacs.008 missing purpose codeDuplicate of D-104 (92% match)Confirmed duplicate, closed
D-121 fee charged twice on returnSev 1, root cause: fee reversal jobAgreed Sev 1, fix today
D-125 label typo on status screenSev 4Agreed, backlog

AI drafts; the triage team approves every severity and decision.

3. Data quality report (Data First)

Structured addresses
97.8%
Mapping fields covered
100%
Test data masked
100%
Reference data current
Yes

Gaps: 2.2% of legacy beneficiaries still need town and country completed before the ISO 20022 address rule applies to them.

4. Go / no-go summary

CriterionStatus
All P1 cases passedMet
Open Sev 1 / Sev 2 = 0Met
Parallel run reconciled to the centMet
Performance at peak within SLAMet
DR exercise passedMet
Business, compliance and ops sign-off2 of 3 received

Recommendation: Go, subject to the final operations sign-off.