A webhook tells your connection that something changed. Polling means your connection regularly asks whether anything changed. Neither is automatically the right choice for every business process.
Start with the delay you can tolerate. An overnight management report and a customer waiting for a booking confirmation do not need the same update speed.
What happens in each approach?
With polling, a connection might ask the order system every ten minutes for new orders. The delay depends partly on when the next check runs. Checking more often uses more requests and processing, even when nothing has changed.
With a webhook, the order system sends a notification to a receiving service. That can start work sooner, although “notification received” and “the right record was updated” remain separate outcomes.
Shopify's webhook overview explains this notification approach and its alternative to continuous polling. The available events and their contents depend on the provider. Some notifications contain useful record details, while others simply tell the receiver to fetch the current record.
Work backwards from the missed deadline
Consider a fictional equipment-hire business. Its daily revenue summary can wait until the next morning. A new reservation may need attention sooner so staff do not promise the same item to another customer.
Write three practical requirements for each process:
- Normal delay: how quickly an update should usually appear.
- Attention point: how old an unfinished update can become before a person needs to act.
- Fallback: what staff should do while the connection is unavailable.
These are business requirements to agree and test, not guarantees created by selecting a webhook option.
Ask about coverage and missed changes
A system might send notifications for new records but not for every edit or cancellation. A receiver may also be unavailable at the wrong moment.
Ask the developer to identify the events covered, how subscriptions remain active and how the integration checks for gaps. Google's Calendar push guide documents notification channels and their expiry. That illustrates why a connection can need maintenance even when the underlying business process has not changed.
A periodic comparison can complement notifications. It should compare meaningful records or changes, rather than repeatedly copying everything without considering volume and provider limits.
Specify what you want to see
For the hire business, a useful demonstration would include a normal booking, an amendment, a cancellation, two notifications for the same change and a temporary receiving outage.
Staff should be able to see the last successful comparison and any work still waiting. Ask whether an alert means the notification failed, processing failed or the final business record is missing. Those call for different responses.
Choose the simplest supported design that meets the timing requirement. “Real time” is a marketing phrase until someone defines the expected delay, measures it and shows how missed changes are recovered.