This is the first question in any automation, and the one people settle fastest when it deserves the most thought: what sets all this in motion?
The choice of trigger decides two things you will not easily change later: how fast your automation reacts, and what it costs you every month while doing nothing.
Definition
A trigger is the event that starts the execution of a Workflow. It is always the first step, and there is only one per workflow.
There are three families, and choosing between them is structural.
| Family | How it works | Responsiveness |
|---|---|---|
| Scheduled | Every hour, every Monday, on the 1st | As often as you set |
| Polling | The tool regularly asks "anything new?" | Delayed by the interval |
| Event-based | The service notifies you as it happens | Immediate |
Polling is the classic trap
A polling trigger queries a service every five minutes to see whether anything is new. It is simple to set up, which is why most beginners stop there.
Do the arithmetic: every five minutes is 288 checks a day, around 8,600 a month. If your event happens ten times a month, you have spent 8,590 calls on nothing. On tools billed per operation, the invoice gets unpleasant quickly.
Where the service offers it, always prefer an event-based trigger, which is to say a Webhook. You go from thousands of pointless calls to one call exactly when needed, and your automation reacts within the second instead of waiting for the next pass.
Based on ten real events per month, the rate at which something actually happens in a small business.
Best practices
Match the frequency to the real stakes. An accounting sync does not need to run every five minutes. Once a night is enough, and divides consumption by three hundred.
Beware of burst triggering. An import of a thousand rows can start a thousand simultaneous runs and blow your quotas. Batch processing exists for these cases.
Plan for double triggering. The same event can arrive twice, notably on webhooks that replay after a failure. Your workflow must be able to recognise that it already handled this event.
Keep a manual trigger. Being able to start a workflow by hand to test or catch up on a failed run saves a lot of contortion.
Frequently asked questions
Can you have several triggers?
A workflow has one entry point, but nothing stops you building several workflows that call the same processing. That is cleaner than piling up conditions at the start.
What if the service is down at trigger time?
It depends on the family. A polling trigger catches up on the next pass. A webhook can be lost if your tool does not respond, unless the sender replays it, which serious services do. That is worth checking in their documentation.
Does a scheduled trigger cost anything when it finds nothing?
On most tools, yes: every run counts, even an empty one. This is exactly why the frequency deserves care rather than being set as short as possible out of convenience.
How do you choose the right one in a real case?
By starting from the delay you can accept, then taking the cheapest option that respects it. Our n8n course works through that reasoning on concrete cases, with the consumption maths to back it.