How it works
A routine combines a trigger, optional conditions and actions. The trigger starts evaluation, conditions decide whether to continue, and actions request changes from devices. A time, button or sensor event can be the trigger. A saved scene is different: it describes desired device states and may be one action within a larger routine.
Important differences and limits
Execution depends on the controller and the connections each action needs. Some routines run locally, others depend on an online service. Multiple rules can issue conflicting commands, and a successful trigger does not prove every requested action succeeded. Manual changes may also be overwritten by a later rule.
A useful practical check
Write the intended behaviour in plain language before configuring it. Identify the trigger, exceptions, target devices and a manual alternative. Test normal use and edge cases such as a controller restart or unavailable device. Review activity logs and give related rules clear names so an unexpected action can be traced to its source.
An example in practice
For example, a motion event might turn on a hallway light only after sunset, then turn it off after a quiet period. Test repeated movement and someone using the switch manually. Otherwise a simple timer may extinguish the light while the room is still occupied or undo a deliberate manual choice.