LEARN. PRACTICE. GROW.Build a timeline
The user requests account A verification through Q1. While waiting, they change the draft to B and initiate Q2. Q2 returns NMTC; Q1 later returns MTCH. An interface choosing the last received response falsely confirms B. Drawing the timeline before writing code makes the cause visible. The model signature includes revision, environment, mode, account, and party. Only a response bound to the current snapshot is applicable. This is an original conservative DR policy, not a universal EPC rule for interface revisions.
Preserve identity and context
The local registry accepts repeating a key with the same snapshot and rejects reuse for different data. That property is limited to the in-memory dictionary. It proves neither exactly-once financial execution nor a remote service’s idempotence. DR-Q1 is a teaching label; it does not meet real headers’ UUID requirement, which the program does not validate. In a project, request identity policy, cross-environment scope, retention, and collision handling. Retaining only MTCH loses the link to checked data. Cancelling an old request can improve experience but does not replace checking binding when a response arrives.
Define duration before calculating it
The VOP inter-PSP interval has a maximum of five seconds from request dispatch to response receipt at the requester. Do not use another payment flow’s limit. In the fixture, 1000 to 6000 milliseconds meets the boundary; 6001 exceeds it. These numbers are synthetic, not performance measurements. For actual instrumentation, define correlated boundaries and a clock appropriate for local interval measurement. Subtracting timestamps from different machines without knowing their relationship can produce negative results. A monotonic clock helps measure local elapsed time, while CPU time excludes the entire network wait.
Choose denominators and preserve purpose
The fictional set contains one hundred requests: eighty MTCH, ten NMTC, five NOAP, and five API errors. Saying one hundred responses represent one hundred positive checks confuses receipt with outcome. Report eighty percent MTCH across all requests and keep other categories visible. A rate based only on completed comparisons needs another explicit denominator. The PM should be able to explain both to the sponsor. Responses also do not provide general authorization to build a commercial account-holder directory. Define access, purpose, and retention with accountable owners; the lab uses invented names only and establishes no bank retention policy.
Prepare the decision in English
In the forty-five-minute workshop below, the group receives a response race, a negative measurement, and a request for final acceptance. Prepare an English summary covering observation, impact, action, owner, and next update. For example, saying the old response does not belong to the current draft is more precise than announcing failure of every check. Attach neither credentials nor personal data. Identify conclusions supported by the model and those requiring the supplier or an authorized environment. Operational handover should enable another person to reconstruct the case from sanitized inputs and expected outcomes.
Distinguish exercise and acceptance
Run the lab, change one signature dimension, and predict which responses cease to apply. Then repeat the scenario where Q2 arrives before Q1. The forty checks cover parsing, bounded contract handling, binding, and arithmetic boundaries. The program contains no automated browser, HTTP call, or network measurement. The described interface is a decision case. Record workshop attendance only if it takes place; individual practice is self-assessment. For actual acceptance, list untested dependencies and agree evidence with their owners. Completing this course grants no EPC certification and demonstrates no operational readiness of a banking integration.
DR original workshop / Oficina original DR: 45 minutes
Synthetic fixtures only. No requests, real accounts, credentials or payments.
Apenas fixtures sintéticas. Sem pedidos, contas reais, credenciais ou pagamentos.
0-10 min: Predict / Prever
Q1: revision=1, account=ACCOUNT-A. Q2: revision=2, account=ACCOUNT-B.
Q2 returns NMTC. Q1 later returns MTCH. Current revision remains 2.
Q2 devolve NMTC. Q1 devolve MTCH depois. A revisão atual continua a ser 2.
Expected: Q1 is stale for the current draft; retain Q2's applicable outcome.
Esperado: Q1 é antigo para o rascunho atual; manter o resultado aplicável de Q2.
10-20 min: Diagnose / Diagnosticar
100 requests = 80 MTCH + 10 NMTC + 5 NOAP + 5 API errors.
Report 80% MTCH over all requests; preserve every other category.
Reportar 80% MTCH sobre todos os pedidos; conservar as restantes categorias.
Synthetic timing: sent=1000ms, received=6001ms; elapsed=5001ms.
Tempo sintético: envio=1000ms, receção=6001ms; duração=5001ms.
Negative cross-clock duration cannot establish compliance.
Uma duração negativa entre relógios não demonstra cumprimento.
20-35 min: English decision note / Resumo de decisão em inglês
Observed: a late response overwrites the current draft's result.
Impact: the interface may display verification for different data.
Action: bind observations to immutable request snapshots and draft revisions.
Owner: nominate the adapter and frontend owners; agree a specific update time.
Evidence: sanitized timeline, contract versions, expected and actual outcomes.
Acceptance: local fixture evidence only; external dependencies remain open.
Não incluir tokens ou dados pessoais. Não afirmar autorização de pagamento.
Do not include tokens or personal data. Do not claim payment authorization.
35-45 min: Handover / Passagem à operação
Another participant reruns the model and explains the remaining evidence.
Outro participante repete o modelo e explica a evidência ainda em falta.
Record self-review for solo practice, not independent operational acceptance.
Na prática individual, registar autoavaliação, não aceitação operacional independente.
Deliver code hash, fixtures, predictions, actual results, owners and open questions.
Entregar hash, fixtures, previsões, resultados, responsáveis e questões abertas.
Case: a late MTCH from Q1 replaces Q2’s NMTC after account editing. Retain Q1 in its own history and keep the observation applicable to the current revision.
Common pitfalls
Using arrival order as freshness; reusing identities; interpreting negative duration as speed; declaring external qualification from local tests.
Related topics: Timing, confirmations, and late outcomes · Mandates, Core, and B2B · Changes and operational readiness
The decision depends on an observation bound to the right request and defined measurements; reporting should retain uncertainty and missing evidence.
Reference: EPC Verification Of Payee rulebook 1.1 · BigSavant Payments and SEPA professional assessment2026.10