A pallet is not packaging. It is a capital asset that travels between companies, gets repaired, gets rented and occasionally disappears. Once that framing is accepted, the RFID design changes: the tag is an asset identity, and the data model is a move ledger rather than a stock quantity.

A pallet identity is an asset identity

Product codes describe what is on the pallet; you need something describing the pallet itself.

Use a serialised identifier per physical pallet, never a product or article number.
Keep the identifier stable across repairs, repainting and re-labelling.
Separate the pallet class (size, material, load rating) from the individual identity.
Assign ownership explicitly, because pooled pallets change hands more often than they change state.
Never reuse an identifier after scrapping; duplicate identities destroy the settlement trail.

Tag choice for wood, plastic and pooled pallets

The substrate decides the physics more than the chip does.

Wood with moisture content absorbs energy; a tag on a dry face reads differently in a rain-soaked yard.
Plastic pallets vary widely; a tag tuned for one moulding can read poorly on another.
Inset tags survive forklift impact far better than surface labels.
Nail-in or screw-in tags are durable but slow to apply and hard to move between pallets.
Test read performance on a loaded stack, not on a single empty pallet on a bench.

Read points on the pallet journey

Coverage is a chain, and the chain breaks at the point nobody instrumented.

Inbound receiving, where pallets enter and the owning pool is identified.
Dispatch, where custody transfers to a carrier and the settlement clock starts.
Customer sites, the hardest point, where the pallet is unloaded and often not scanned.
Return and repair depots, where condition and repairability are decided.
Scrap and disposal, where the identity must leave the ledger explicitly.

Rental settlement and the move ledger

Settlement is the reason the project exists, and it needs documents.

Model each pallet move as a document with sender, receiver, time and location, not as a counter update.
Match issues and returns by identifier, so a dispute can be resolved with one lookup.
Handle partial returns and mixed loads, which are the normal case in practice.
Reconcile with the pool operator monthly, using identifiers rather than counts.
Keep the settlement period's opening and closing balances, so recalculating is possible.

Loss, shrinkage and the deposit model

Loss is a financial number, so it needs a definition everyone accepts.

Define lost as unaccounted for beyond an agreed ageing window, not merely unseen today.
Distinguish loss from retention: a pallet sitting at a customer site is not lost, it is late.
Charge deposits per class, not per pallet identity, or administration outweighs the value.
Age every open move, and escalate by age band rather than by volume.
Publish loss rate by customer, which is the number that changes behaviour.

Repair, inspection and lifecycle state

A pallet's state matters more than its location for safety and cost.

Record inspection results at the depot: sound, repairable, scrap.
Change state on the pallet record, and keep the history rather than overwriting the current value.
Retire identifiers at scrap, and remove or destroy the tag at the same time.
Track repair cost per identifier, so the fleet's true cost becomes visible.
Keep safety-critical defects separate from cosmetic damage.

The sync model: pallet moves as documents

The integration is a ledger problem with a radio front end.

Every read produces an event; events become moves only after validation.
Validate that the pallet exists, is not scrapped, and is at a plausible location for the timestamp.
Make ingestion idempotent, because yard readers retry aggressively.
Keep corrections as new documents rather than editing history.
Expose balances to finance as of a date, not as a live counter.

Multi-party pools and data sharing

Pallets cross organisational boundaries, and so must the data.

Agree on the identifier scheme before the technical integration, not after.
Share move events, not full inventory positions, to limit commercial exposure.
Define who owns a disputed pallet while it is in transit.
Use a neutral reconciliation file format, exchanged on a schedule both parties accept.
Expect latency; a pool settlement that demands real-time visibility across companies will fail.

Reading pallets in stacks and on trucks

The physical conditions decide whether the data is usable.

A stack of fifteen pallets on a trailer is not fifteen clean reads; tags near metal and tags at the bottom fail.
Read at the dock door while the vehicle is stationary, not while it is moving.
Position antennas to see the pallet edge rather than the load face.
Accept that a percentage of moves will be captured manually and design for that entry path.
Validate read rates by lane and by carrier, because some routes will always be worse.

Keeping the pallet ledger separate from the goods ledger

Two assets move on the same truck, and they are not the same transaction.

A pallet leaving with goods transfers the pallet to the customer and the goods to the customer's stock.
Send pallet events to the pool operator; send goods events to the customer's order system.
Do not let one receipt document do both jobs, or the two ledgers will disagree on the first exception.
Reconcile the pallet ledger against the goods ledger monthly, because drift is guaranteed.
Make ownership explicit on every line, so a dispute can be routed to the right party immediately.

Metrics and rollout

Measure the settlement, not the read count.

Track read rate per read point, unaccounted moves, ageing of open moves and loss rate by customer.
Track the dispute rate with the pool operator, which is the clearest sign the data model works.
Start with one lane and one pool partner, where the reconciliation cycle is short.
Include the depot inspection step early, or the fleet's condition data will never be trustworthy.
Report cost per pallet per year, because that is the number the business can act on.