What Is a Webhook? Making Your CRM Talk to Other Systems
A webhook guide: how they work, use cases, security (signature verification), retries and error handling.
Two systems can talk in one of two ways. The first is asking: checking every five minutes whether anything new has happened. The second is waiting: letting the other side tell you when something happens. A webhook is the second — and it is almost always better.
How it works
A webhook is an HTTP request sent automatically to an address you choose when an event occurs. It has three parts:
- The event: what triggers it? A new ticket, a stage change, a quote approval.
- The target URL: where the request goes — the endpoint in your system that receives it.
- The payload: the data sent, usually JSON, with the record’s identifier and relevant fields.
The receiver takes the request, processes it and returns a response meaning “received”. That is the whole mechanism.
Where it helps
- Accounting/ERP notification: telling the other system when an order is approved.
- Shipping: creating a consignment and writing the tracking number back.
- Custom reporting: streaming records into a data warehouse.
- Your own application: giving something that happens in the CRM an effect in your software.
- Connector platforms: automation tools that link to hundreds of services without code are fed by webhooks.
The rule: if it needs to be immediate and is a one-way notification, a webhook is the right tool. If you need two-way, query-based communication, an API call fits better.
Six conditions for reliability
Signature verification
Confirm the request really came from your CRM. Signing with a shared secret eliminates forged calls. Do not rely on the URL being secret.
Respond quickly
Take the request, queue it, respond immediately. Doing heavy work inside the request produces timeouts and needless retries.
Idempotency
The same event can arrive twice. Every event should carry a unique id, and an id already processed must not be processed again.
Retries
If the receiver does not respond, the request should be repeated — at increasing intervals and with an upper limit.
No ordering guarantee
Events may not arrive in the order they were sent. Verifying the record’s current state at the source is more robust than trusting sequence.
Visibility
Failed calls must be listed somewhere. A webhook that fails silently is worse than no webhook at all.
The most common mistake
A webhook is set up, works, and is forgotten. Months later the receiving address changes or an update breaks the endpoint — and nobody notices. Data starts arriving incomplete, and the problem only surfaces when a report does not add up. That is why failure alerting is a mandatory part of the setup, not an optional one.
Testing
- Start with a public test endpoint to see the shape of the payload.
- Write your own endpoint and add signature verification from the beginning — it does not get added later.
- Return an error deliberately and confirm retries work.
- Send the same event twice to test idempotency.
- Define who receives failure alerts before going live.
Frequently asked questions
What is the difference between a webhook and an API?
With an API you ask; with a webhook the other side tells you. They are not competitors: the webhook announces the event and an API call then fetches the detail. That pair is the most common and most robust pattern.
What data should the payload carry?
An identifier and the event type are enough; keeping sensitive data out of the payload is safer. The receiver pulls the detail with an authenticated API call.
I have no technical team — can I still use them?
Writing an endpoint is developer work, but ready-made automation platforms can receive a webhook and connect it to hundreds of services without code. That is a practical route for small teams.
Next step: to see how the webhook action is defined inside automation rules, request a demo.