What is RFID, and when does it actually pay off?

RFID stands for Radio Frequency Identification. In plain language: a small label with a chip in it states its own number when a device comes close enough to ask. Nobody has to aim, open anything or pick anything up.

That sounds simple, and at heart it is. The hard part is not the technology but the question of where in your work such a read belongs, and whether the answer solves anything there.

Where you already run into it

Chances are you used a tag a few times today without giving it a thought.

• The badge that gets you into the building.
• A car key that unlocks the door while it sits in your pocket.
• The chip in the neck of a dog or a cat.
• The hard plastic tag on a garment in a shop, and increasingly a thin label inside the price ticket.
• The ski pass that opens the gate while you keep walking.

All the same idea: a chip that states its number when a device asks for it. The difference is the frequency, and with it the distance. A door badge has to be held against a panel deliberately, because you do not want the door opening every time someone walks down the corridor. A ski gate has to work at half a metre while you keep moving.

That is exactly why RFID feels different on a work floor than at home. At home you hold one thing against one device. In a warehouse you want hundreds of things at once, without anyone touching them. The same technology, a completely different design question.

What happens during one read

A read always consists of the same three steps, whether it covers one box or a full pallet.

The reader transmits

The reader sends a radio signal into the room through an antenna. That signal fills an area. How large that area is, is something you decide when setting it up.

The tag wakes up and answers

The label picks up part of that signal. That is enough energy to switch the chip on briefly. The chip then sends back its unique number. So there is no battery in it that anyone has to replace.

Software turns it into an event

The reader delivers a list of numbers that answered, often the same ones many times per second. That is not information yet. Software decides what it means: this pallet has left the building, this tool has been issued, this item is still there.

That third step is the one most often underestimated. A reader does not see a box or a cart. It sees numbers in a field, and nothing more.

What is actually on the tag

This is the misunderstanding that causes the most surprises. An RFID tag holds no information about the object. It holds a number.

That number is unique, and that is all it is. It is a key. Your own systems keep track of what that key means: which item it is, from which order, in which batch, who took it and when. If the status of the object changes, it changes in your system and not on the label.

Technically you can put more on a tag. A chip has some memory, enough for an item number, a date or a short piece of text. In practice that is rarely wise. You end up with two sets of records that can drift apart, and the one on the label is precisely the one you cannot correct without visiting every object.

There is one situation where it does make sense: an object that spends part of its life outside your own systems and still has to be recognised along the way. A carrier sitting at a customer for weeks, for example. Even then the rule of thumb holds that the number leads and the rest is at most a copy.

The number gets there in one of two ways. It is already on the chip when it is manufactured, or you write it in at the moment you print and apply the label. The second is the usual route. The printer writes and prints in one action, so what you see on the label and what sits inside it belong together.

What RFID is not

Three things get confused with it regularly. Knowing the difference saves a wrong expectation.

That last line is the expectation that fails most often. RFID tells you where something was last seen, and helps you search for it directly. It is not a continuous position.

Barcode and RFID are not mutually exclusive

Almost everyone frames this as a choice: do we move to RFID, yes or no. In practice the two often run side by side, and that is not a half-hearted compromise but usually the cheapest answer.

The reason is straightforward. RFID earns its keep where there is volume and a fixed point things pass. Where it is one object at a time, a barcode does the same job for a fraction of the cost. A handheld that reads both makes that distinction invisible to the person using it.

What you do have to keep in one place is the identity. If a box carries a barcode and an RFID label with a different number, you have two truths about the same object. The same number on both, or a clear translation between them, saves a great deal of untangling later.

The frequencies, briefly

RFID is an umbrella term. There are three families, and they are not interchangeable: a reader for one does not read the others.

• LF, low frequency. Very short range. Think animal identification.
• HF, 13.56 MHz. NFC sits here. Strong when someone deliberately touches one thing.
• UHF, around 860 to 960 MHz. This is the family for logistics and manufacturing: many objects at once, metres away.

For warehousing, manufacturing and carrier management it is almost always UHF. How far that reads, what material does to it and the difference between active and passive is covered in the separate explainer UHF RFID explained. This page stays on the main line.

Why it sometimes does not read

RFID has one property that explains nearly every surprise. It works with radio waves, and radio waves take no notice of what was agreed on paper. Five causes come up again and again.

Metal reflects. An ordinary label on a steel cart or a metal cabinet does not read, or reads ten centimetres. The fix is a label that holds the chip a few millimetres off the surface. That label is larger and costs more, which is why the material matters before the unit price does.

Liquid absorbs. Water swallows the signal. A tag on a full jerrycan or on the far side of a filled tank is difficult to impossible. That is why reading usually happens before filling.

Orientation counts. An antenna has a direction. A label sitting square to the field answers worse than the same label turned a quarter turn. If your boxes always lie the same way up, you will never notice this. If they lie at random, this is the first cause to investigate.

Shielding. A tag in the middle of a fully stacked pallet sits behind everything around it. If that load is full of metal or moisture, that one tag becomes the gap in your count.

Volume takes time. Tags answer in turn. Fifty labels are read in a fraction of a second, five thousand are not. At a gate someone drives through, the question is therefore not only whether it reads, but whether it reads within the time the action takes.

All five can be measured, and all five are cheaper to find out beforehand than after a rollout.

The words you will run into

Quotes and data sheets use ten words for about four things. Here is the short translation.

When RFID pays off

The technology works almost anywhere. Whether it pays is a different question. There are three conditions, and all three have to hold.

There is volume

For one object at a time a barcode is cheaper and often faster. The difference appears as soon as someone has to work through tens or hundreds of things. That turns a counting action into a reading action.

There is already a moment where the object passes

This is the important one. If there is a point in your process where the object passes anyway, you can read there without adding work. If that point does not exist, you are adding an action instead of removing one, and the gain disappears again.

The error costs money today

Searching, recounting, over-ordering, an order that leaves wrong. If there is no error that hurts, there is nothing to earn back either.

An example: a pallet coming in

Round numbers to make the picture clear, not a measurement.

The way it works now: a pallet arrives with forty boxes. Someone picks up each box, finds the barcode, aims the scanner and puts the box back. Allow ten seconds a box if it goes well, and more when a label is underneath or damaged. That is a good six minutes. At the end you know what was scanned, not what was in the truck.

With a read point at the door, the same pallet drives through it. Within a few seconds the labels have answered and there is a list. That also shifts the question you ask: no longer “what did I scan”, but “what is missing against the packing slip”.

The real gain is in that second sentence. The time saved counts, but what sticks is that you see a discrepancy while the pallet is still on the floor, and not two days later during a count. A box missing at that moment can still be found.

And the honest downside: a gate like that also reads the pallet parked next to it, if you do not bound the field. Which is why the first question at a read point is not how far it reads, but where it has to stop.

Five questions for a conversation with a supplier

Anyone new to this is talking about brands and read ranges within ten minutes. These five questions pull that conversation back to what it has to deliver.

1. Which action disappears? If the answer is “we get more insight”, there is no business case yet. There has to be an action that costs time now and will not afterwards.
2. Where is the read event? Point at the place where the object passes anyway. If there is no such place, you are adding work instead of removing it.
3. Has it been tested on our own material? Not on a demo box, but on the cart, the bin or the product as it looks at your site, in the space where it has to work.
4. What goes to our ERP or WMS, and when? Passing raw read data straight through only produces faster mess. What should come out is an event: this has arrived, this has left, this is missing.
5. When do we stop? Agree up front on the read rate or the cycle time at which it is a no-go. Without that line, every setback becomes a reason to order something else.

A supplier who gives a concrete answer to all five has done it before. A supplier who deflects question three to a data sheet has not.

From explainer to a system

The moment you want more than understanding, four parts come into play: the labels on your objects, the readers that observe them, the software where the reads land, and a connection to the system you already work with. Those four are much the same with every supplier.

The difference is in the implementation: which label your material can carry, where the read event sits, and who measures that before anything is ordered. What those parts are, how an implementation runs and what the first step costs is on what is an RFID system.

Frequently asked questions

In closing: the question is not whether it works

RFID works. That has not been in question for years, and the parts are much the same everywhere.

What remains is where in your process a read acquires meaning, and whether the error it removes is large enough to earn back the labels and the equipment. That is not a technology question but a process question, and it is answered by measuring one action before anything is rolled out.