Should tags be written before or after being applied?
Before, whenever the process allows a choice. Writing a loose tag on a bench succeeds essentially every time, while writing one already fixed to a surface inherits every coupling problem that surface creates. The exceptions are real and narrower than people assume.
Why before wins
A loose tag on a flat surface, with the phone laid on top of it, is the best possible coupling situation. Nothing detunes it, the distance is minimal, and the phone can be held still without anyone reaching above their head.
The same tag on a machine housing, behind a cover or on a curved bottle inherits every problem that surface creates, and writing needs more margin than reading does.
The yield difference is the argument. A bench run completes without incident, and a wall run produces a tail of tags that would not take the write.
When after is unavoidable
Three genuine cases.
The content depends on the position, and the position is only known at installation. A tag whose address encodes which door it is on cannot be written before somebody chooses the door.
The tags are applied by a third party who will not handle content, such as a printer laminating inlays into cards or a manufacturer embedding them in a product.
And correction of an existing deployment, where the tag is already on the wall and replacing it is more expensive than writing it in place.
The hybrid that resolves most cases
Write a redirect address before application, with a unique identifier per tag, and decide what each identifier means afterwards.
The position dependent case disappears: the tag is written on the bench, applied to whichever door, and the mapping from identifier to door is recorded at installation and configured on the server later.
This turns a problem of writing into a problem of recording, which is far easier to do standing on a ladder.
Sequencing a real job
Write on a bench, in two piles, verifying as you go.
Apply, recording identifier and location as each tag goes on.
Walk the installation and tap every tag in place, which confirms both that the write survived and that the surface allows a read at all. This is the step that catches the tag applied to an unexpected metal panel.
Then configure destinations, and lock if the deployment warrants it, on a final pass once everything is confirmed.
Four passes sounds heavy and is faster than one pass plus rework, because rework happens on ladders and the first pass happens at a table.
Frequently asked questions
Does applying a tag damage what was written?
Not the content, which is electrically stored and unaffected by handling. What application can damage is the antenna, by creasing or folding, and that kills the tag rather than the data.
Can I write through a laminate?
If the laminate is thin and not metallic, usually yes, with less margin than a bare tag. Test one before assuming it for a production run.
What if the tag is on metal?
Then reading and writing will both fail unless it is a tag designed for metal. This is a specification problem rather than a sequencing one.
Should I verify before or after application?
Both. Before catches writing failures, after catches surface failures, and they are different problems with different fixes.