Which NFC chip should you choose, and on what criteria?
Four criteria in order. The NFC Forum tag type, because it decides which phones respond. Then memory, because content that does not fit is a hard failure. Then features such as password protection or mirroring, which cannot be added later. Price comes last and endurance almost never matters.
The four criteria in order
Tag type, memory, features, price. The order matters because a failure at each level is progressively cheaper to fix, and only the first is a compatibility problem you cannot work around.
Tag type comes first
The NFC Forum tag type determines whether an ordinary phone reads the tag with no special software. Everything else is negotiable and this is not.
Type 2 is the safe answer for anything the public will tap. It is the most widely supported, the cheapest and the least surprising. The NFC Forum publishes the specifications for all five types.
Choosing another type needs a reason: reading at distance with dedicated hardware points at Type 5, and card emulation scenarios point at Type 4. Choosing one by accident, because a supplier had stock, is how a deployment fails for a subset of visitors.
Memory second
Content that does not fit is a hard failure discovered at the worst moment, usually after the tags are printed.
Size the memory against the free space, not the advertised capacity, and leave margin. Content grows: an address gains a parameter, a card gains a second phone number.
The rough hierarchy is that a short web address fits anywhere, a contact card or Wi-Fi credentials need noticeably more, and anything beyond that belongs on a server rather than on a tag.
Features third, and they are permanent
Password protection, a read counter, and identifier mirroring are the three that change what a deployment can do.
Mirroring is the one most often needed and least often anticipated. It lets a thousand identically written tags deliver a thousand different addresses, which removes the entire per tag encoding step. Discovering you need it after buying tags without it is an expensive lesson.
None of these can be retrofitted. They are properties of the silicon, so they belong in the purchase order.
What to ignore
Write endurance, in almost every case. The published figures are far beyond what any normal deployment reaches, and treating them as a differentiator is a distraction.
The brand family, in itself. Two chips from different families with the same tag type and the same free space behave identically for your purposes.
And read distance claims attached to the chip, which are attached to the wrong component entirely. Distance follows antenna size, and the antenna is part of the tag rather than the chip.
Frequently asked questions
Should I just buy the most capable chip available?
It costs more per unit for capability most deployments never use. The disciplined approach is to identify which features you genuinely need and buy the cheapest chip offering them.
How do I find out the tag type of a product?
Ask the supplier, or read one sample with a tool that reports it. A listing that omits the tag type is a listing that has not answered the most important question.
Does the chip determine the tag's physical size?
No. The same chip appears in stickers, cards, keyfobs and discs, with very different antennas. Chip and form factor are independent choices.
Can I mix chips in one deployment?
You can, and it creates inconsistent behaviour that is hard to diagnose later. Where a deployment must grow over time, recording exactly which chip was used matters as much as choosing it.