What to track, how to set it up, and when to alert — a practical guide for 2026.
Webhooks fail silently. Unlike API calls where your code receives an immediate error response, a failed webhook delivery means your system simply never gets the data. Three failure modes make monitoring essential:
Target: >99.5%
The percentage of webhook deliveries that return a 2xx status. This is your most important metric. Below 99% warrants immediate investigation. Between 99% and 99.5% indicates emerging issues. Track per-provider to isolate which integration is degrading.
Track: p50, p95, p99
Measure the time from when your endpoint receives the request to when it returns a response. Most providers time out at 5-30 seconds. If your p99 approaches the provider's timeout, you are one traffic spike away from mass delivery failures. Aim for p99 under 3 seconds.
Break down: timeout, auth failure, DNS, 4xx, 5xx
Not all errors are equal. Timeouts suggest load problems. Auth failures (401/403) mean rotated secrets. DNS errors indicate infrastructure issues. Categorizing errors tells you where to look first.
Watch for: spikes and drops
A sudden spike in event volume could mean an attack (replay or flood) or a bug in the sender's system. A sudden drop to zero often means the provider disabled your webhook or your DNS changed. Both deserve alerts.
Detect: new, removed, or changed fields
When a provider adds a new field, it is usually harmless. When they remove or rename a field your handler depends on, it breaks silently. Track the structure of incoming payloads over time and flag deviations from the expected schema.
A complete webhook monitoring setup covers three layers:
The goal is to alert on conditions that require action while avoiding alert fatigue. Follow these principles:
Webhooks fail silently. Unlike API calls where your code gets an immediate error, a failed webhook delivery means your system simply never receives the data. Without monitoring, you will not know about missed payments, unprocessed orders, or broken integrations until a customer complains.
Track five key metrics: delivery success rate (target above 99.5%), end-to-end latency (p50, p95, p99), error rate by type (timeout, auth, DNS), event volume trends to detect spikes, and schema drift to catch new or changed fields that could break your handlers.
A healthy webhook pipeline should maintain a delivery success rate above 99.5%. Anything below 99% warrants immediate investigation. Rates between 99% and 99.5% indicate emerging issues that should be addressed before they escalate.
Focus alerts on actionable conditions: sustained error rate above threshold (not individual failures), latency p99 exceeding SLA for more than 5 minutes, volume dropping to zero for a normally active provider, and schema changes. Use tiered severity — page for delivery rate drops, notify for latency, and log for schema drift.
Schema drift occurs when a webhook provider adds, removes, or changes fields in their payload without notice. New fields are usually harmless, but removed or renamed fields can break your handler logic. Monitoring schema drift lets you catch these changes early and update your code before it causes production failures.
Track delivery rates, latency, errors, and schema drift across all your providers — with alerts that actually matter.
Sign Up Free