Resources / Practical guide
An integration checklist for retries, payments and customer messages
An integration is ready when it handles both success and uncertain outcomes. Define which system owns each record, how retries avoid duplicates, how operators recover unfinished work and how users know whether an action really completed.
Woro Global editorial · Published 4 October 2026
Choose an authoritative record
For each entity, name the source of truth and its identifier. Map required fields, access scopes and retention rules. Avoid making an email address or display name the only identity when the provider offers stable identifiers. Keep tenant boundaries explicit in storage and job execution.
Treat acknowledgements carefully
A request timing out does not prove the provider did nothing. Record the attempt before a paid or irreversible action, use a provider idempotency key where supported, and reconcile uncertain outcomes. Do not tell a user no charge occurred simply because the response was lost.
Design retries as part of the workflow
Use bounded retries and backoff for transient errors. Permanent permission refusals need a visible repair path, not an endless retry loop. A durable job should preserve its phase across a process restart and prevent a stale worker from repeating completed work.
Make payment and delivery states honest
A generated message is not a delivered message. A submitted payment is not a settled payment. Use distinct states and explain when a person must check the provider. Keep refunds, captures and repeated callbacks idempotent and audit their results.
Test with deliberate interruption
Test duplicate events, out-of-order events, a lost acknowledgement, an unavailable queue and a worker restarting during an action. Use realistic permissions: a local superuser can hide a problem that appears under the production database role. Preserve failed-run logs rather than treating a rerun as a diagnosis.
Give operators useful evidence
Log a safe correlation identifier, the workflow stage and the provider outcome without printing tokens or customer content unnecessarily. Alert on sustained failures and stuck work. Define who resolves an exception and show the user whether retry, correction or manual intervention is required.
Discuss your workflow
Explore the service or prepare a project brief. Scope and verification depend on your systems and requirements.