
Reconciliation
Part of Checkout performance
Monitoring errors after a checkout release
Watch browser, API, payment-event and order errors after a release, then investigate exceptions without mislabelling payment outcomes.
After a checkout release, monitor whether buyers can submit payment and whether attempts reach the right order state. Watch browser failures, failed store requests and payment-event handling together. An error-free page view does not prove an order was paid; one failed attempt does not prove the whole checkout is unavailable.
Establish a baseline
Record the release version, affected routes and a comparable period before rollout. Separate new failures from ordinary declines and expected pending payments. Assign someone who can investigate each layer: storefront, store API, payment integration and order updates.
| Signal | Investigation |
|---|---|
| Browser script error or unhandled rejection | Did a control fail before a payment request was sent? |
| Failed delivery, pricing or payment request | Which operation failed, and what did the buyer see? |
| Payment event received but order unchanged | Was the event authenticated, accepted and applied to the right order? |
| Order pending longer than expected | Is the provider still processing, or is the store missing an outcome? |
Browser error and unhandledrejection events reveal some uncaught failures, not every application error. When the application uses fetch(), it must also check HTTP error responses explicitly. Record a safe error type, route, release version and correlation reference where available, without unnecessary customer or payment details.
Join payment and order timelines
Keep the store order reference, provider payment reference and event status together. Some integrations provide an immediate result and a later webhook. Adyen's guidance says webhook messages help keep systems synchronised with events on Adyen's side, and that event types can inform business logic.
Verify that each Adyen webhook message is genuine and has not been modified before processing it.
For an exception, establish whether an attempt was submitted, what the provider currently reports, whether the store received an event and whether the order applied it once. If the outcome remains unknown, keep it unresolved rather than calling a missing browser confirmation a decline.
Respond to a new pattern
Agree in advance when to pause a rollout or restore a previous version. A repeatable inability to pay, incorrect totals or a paid order without the required payment outcome merit immediate investigation.
Use counts with rates so a small route remains visible without making a thin sample look decisive. Compare the affected version and journey with the baseline before attributing the problem to the release.
When payment events are missing, inspect the provider's delivery tools. Adyen documents a retry queue and troubleshooting view for its configured webhook endpoints; those facilities are provider-specific. Repair the underlying fault and review affected orders before clearing an alert.
After a fix, repeat the affected journey in a permitted test mode and review later live signals. Record orders still awaiting investigation. The customer-facing result, store record and dependable payment outcome should agree.
Key Metrics After Checkout Release
- Error rate increase
- Monitor for spikes above baseline
- Unprocessed webhooks
- Check Adyen’s webhook troubleshooting view
- Pending order duration
- Flag if exceeding expected time
- Failed fetch requests
- Verify HTTP status codes explicitly



