The fixed asset register is a document that is true on the day it is signed and gradually becomes a work of fiction. An organisation with more than a few hundred assets typically drifts between five and fifteen per cent within two years — not because anyone is dishonest, but because transfers happen by conversation, disposals happen quietly, and the annual count is done in a hurry by people who have other jobs.

RFID does not fix that on its own. Tags are the easy part. What fixes it is a design in which counting stops being an annual project and becomes a by-product of work that people were doing anyway.

Why asset registers drift, and why barcodes did not stop it

The causes are mundane and remarkably consistent across sectors.

Transfers between rooms, floors and sites are recorded days later, or not at all, because recording them costs more effort than the move itself.
Disposals — scrapped, sold, lost — rarely generate a document at the moment they happen.
Shared equipment has no single custodian, so nobody feels responsible for its record.
Sub-registers live in spreadsheets owned by individual departments and are reconciled once a year, badly.
The count itself is scheduled at the worst possible time, staffed with temporary labour, and produces a single number rather than a corrected record.

Barcodes were supposed to solve this. They did not, because a barcode still requires a person to decide to scan it, point at it, and find it — and the finding is the expensive part, not the scanning. RFID changes the economics of finding.

What passive UHF actually changes

Passive UHF reading at 3 to 10 metres, hundreds of tags per second, without deliberate human action, changes three things at once.

A room can be counted by walking through it, not by opening every cupboard.
A doorway can record movement without anyone stopping to scan.
A shelf or store room can be re-verified weekly at a cost low enough that nobody argues about it.

The strategic consequence is not faster counting. It is that the register can be kept current, because verification no longer costs a project. That is the difference between a digital transformation and a faster annual ritual.

Tag selection: the asset class decides the tag

Most failed asset projects fail at the tag, not at the reader. The mapping below is the one that survives contact with real equipment.

Metal assets (machines, tools, racks, medical devices): on-metal tags with a ferrite backing or a foam spacer. A standard label applied directly to metal reads at a fraction of its rated range, and sometimes not at all.
IT and office equipment (laptops, monitors, printers): thin flexible labels on plastic surfaces, or hard tags where the item is handled roughly.
High-value portable items (instruments, test equipment, tools): hard tags with tamper evidence, mounted where removal would be obvious.
Consumable-looking but durable items (cables, adapters, furniture): low-cost labels, accepted as semi-disposable.
Items that are rented out or leave the site: dual-technology tags or tags with a printed barcode, so a partner without RFID can still identify them.

Mounting discipline matters more than chip choice. Keep at least the manufacturer's specified stand-off from metal, avoid curved surfaces under tension, and test ten units in the real environment before committing to a thousand.

The count becomes a by-product

The operational design that makes RFID pay is exception-first.

Weekly zone sweeps with a handheld: ten minutes per zone, compare against expected, produce a short exception list.
Doorway or room-level readers for high-value stores: movement is recorded as it happens, and the weekly sweep only confirms.
No annual wall-to-wall shutdown. The annual audit becomes a sample-and-verify exercise on a register that is already current.
Exceptions are resolved within days, with a named owner, not accumulated for the year-end.

The metric to watch is not read rate. It is register accuracy measured continuously: the percentage of assets whose recorded location and custodian match physical reality on any given day.

Integration: where the asset record actually lives

The tag is an identifier, not a record. The record has to live in the ERP or asset system, and the integration has to respect its business logic.

Maintain an explicit mapping between the EPC and the asset number, with the EPC stored on the asset record itself.
Post transfers as transfer documents, not as direct field edits. The document is what the auditor will ask for.
For Odoo, use account.asset for the asset and a stock or internal transfer for the physical move; depreciation runs on the asset record, and an unexplained location change is exactly what triggers audit questions.
Keep the custodian and the location as separate fields. Conflating them is the single most common modelling error in asset systems.
Disposal must deactivate the tag in the reader configuration too, or the next count will report a ghost.

Process redesign: custody, transfer, disposal

Technology amplifies whatever process you already have. Three workflows need to be settled before tagging begins.

Custody: every asset has a named person or role. Issuing a laptop means scanning it to a person, not writing a note in a chat.
Transfer: request, approve, then move. The scan confirms the move; it does not authorise it. Approval-first is what prevents the register from becoming a log of things that already happened.
Disposal: a disposal certificate, the tag killed or quarantined, and the asset record closed on the same day. Tags that survive their assets are worse than no tags, because they keep producing plausible data.

Rollout sequence that survives contact with finance

Baseline first. Measure the current error rate on 200 to 500 assets before any tag is applied. Without a baseline there is no business case, only an assertion.
Pilot one department for six to eight weeks, including the awkward assets — the ones in the ceiling, the ones in the store room, the ones nobody has seen since 2019.
Tag the unknowns as well as the knowns. An unidentified asset that is now visible is the clearest proof of value a pilot can produce.
Scale by site, not by asset class. Each site has its own building physics and its own habits; a rollout that assumes uniformity fails at the second site.

What to measure, and how the case is made

Register accuracy before and after, measured by sampled physical verification.
Hours spent on counting and locating, before and after. This is usually the largest single line in the case.
Write-offs avoided. Every asset found is an asset that does not have to be re-purchased or expensed.
Utilisation of shared equipment, which is often invisible until it is measured.
Audit findings per cycle, which is the number the CFO actually reads.

If the case depends on tag price falling further, it is not a case. The economics of UHF asset tracking already work at typical tag prices; what fails is process design, tag selection on metal, and integration that writes into the database instead of into the business document.