Can a tag trigger a whole scene?

The tag asks by name

A scene is a named group of actions defined in your home system: these lights at this level, that thermostat setting, those blinds.

The tag does not contain the scene. It asks for it, and the system performs it.

So adding a step to the scene requires no change to the tag, and the tag can stay on the wall for years while the scene evolves.

That separation is the whole value, and it is worth preserving even when it would be quicker to put the steps in a phone automation.

Why that is the right design

The system executes it, so the scene runs the same way whoever asked and from whatever device.

It survives a new phone, since nothing about the scene lived there.

It can be tested from the system's own application before a tag is anywhere near it, which separates a scene problem from a tag problem.

And it can be edited by anyone in the household who has access to the system, rather than only by whoever built the phone automation.

What breaks it

Renaming the scene, which is the commonest cause by a distance. A tag asking for a name that no longer exists does nothing and says nothing.

Deleting a device that the scene refers to, which can make a scene fail partway with no obvious sign.

A platform change, which is a larger problem covered separately.

None of these is a tag failure, and all of them present as one. Read the tag and try the scene from the application before suspecting anything physical.

Scenes worth a tag

The ones with several steps you would otherwise do one at a time, since that is where the saving is real.

The ones tied to a place: leaving, arriving, going to bed, starting work.

The ones somebody else in the household should be able to trigger without learning your system.

And not the ones that already happen automatically. A scene triggered by time or by presence does not need a tag, and adding one gives you two things to debug instead of one.