Designing for the webhook that never arrives
Any system that depends on another company's event delivery will eventually miss one. Design for that day and the outage becomes a non-event.
Webhooks are the right tool for order events: they arrive the instant something happens, and you don’t pay for polling. They also fail — network blips, deploys, upstream incidents, retries that exhaust. A system built on the assumption that every event arrives is a system that loses orders.
Three layers, not one
In the platform behind Factory Hub, event delivery has three layers: the webhook for immediacy, an interval re-sync that reconciles anything missing, and a manual sync with validation for the operator who notices something before we do. Each layer is cheap. Together they mean order data doesn’t silently vanish.
The question isn’t whether an integration will fail. It’s whether anyone will find out from a customer.
This is also what makes lean infrastructure defensible. You can run an MVP on modest tiers if the correctness of the data doesn’t depend on the infrastructure never hiccuping.