HospitalityRoom keys, staff credentials and the pieces around them.
Retail and loyaltyGift, loyalty and membership cards with variable data.
Access controlCards and fobs matched to the reader you installed.
Clubs and associationsOne chip for the turnstile, the locker and the desk.
Events and festivalsWristbands, badges and lanyards for each edition.A golden sample protects an order only when its record says which characteristics it approves: colour, dimensions, chip and finish, each with a tolerance.
A reference sample is useful only when everyone knows which characteristics it controls.
A golden sample is the card both sides agree to compare against: one physical unit, produced at a known moment, held by the buyer and by the supplier, pulled out of its sleeve when a later batch looks wrong. It settles arguments quickly, and it settles the wrong ones just as quickly, because a card on a desk looks like proof of everything and proves very little until somebody writes down what it was approved for.
The fix is a scope, and a scope is a document, not a card. The Kaway reference sample record is the single page that travels with every golden sample: the job and revision it belongs to, the characteristics it approves, the characteristics it explicitly does not approve, the measured values with the method behind each one, the named approvers on each side with the date, and the location of the two matched copies. The card is the exhibit. The record is the approval.
A golden sample approves named characteristics, not a whole card. Most of what goes wrong with reference samples follows from treating one object as a blanket yes.
An ID-1 card to ISO/IEC 7810 carries dozens of separately specifiable characteristics: substrate and core material, overall thickness around a 0.76 mm nominal, surface finish, print colour against an agreed target, front-to-back registration, edge and corner quality at a corner radius of about 3 mm, chip reference, antenna geometry, encoding, and read behaviour on a particular reader. One physical unit demonstrates a few of those well, a few badly, and several not at all.
A matte-laminated card proves what matte lamination looks and feels like on that core. It says nothing about whether MIFARE DESFire EV3 keys were loaded correctly, because key loading leaves no visible trace. The question to ask of a sample is not whether it is good, but which characteristics this unit can actually settle.
Write the scope in three parts before the sample ships, not after it is disputed: appearance, construction, functionality. A reference can carry one of the three, two of them or all three, and the record names which.
Appearance covers print colour against the agreed target, front-to-back registration, the surface finish by name, edge and corner quality, and any feature a buyer judges by eye such as a signature panel or a foil block. It is the scope most references are actually approved for, and the one least often written down.
Construction covers the substrate, the core material, the overlay, the layer count, the total thickness and the punching. It matters later because construction is what changes silently when a line is rebalanced or a material is substituted.
Functionality covers the chip reference, the encoding applied and the read behaviour on a named reader at a named distance. It needs a production-representative sample: a card built the way the order will be built, with the real inlay and the real chip. A display body with no electronics cannot carry this scope, however convincing it looks. Card sample types covers which sample each scope requires.
Two matched copies, one record. Both sides hold a unit from the same production moment at the same revision, and both copies point at a single approval document.
A single copy makes its holder the arbitrator. If only the supplier retains a reference, every comparison is the supplier's comparison. If only the buyer holds one, the plant is being asked to hit a target it cannot see. The retained counterpart is cheap, and the part most often skipped.
Storage is part of the control. Each copy goes in a sleeve, out of daylight, away from heat, with the record number written on the sleeve and not on the card. A reference with a felt-tip date across the reverse has stopped being a reference for finish.
The measurements live in the record, because the card drifts. Any physical reference changes while it sits in a drawer, and nothing about the approval changes with it.
PVC yellows under ultraviolet light and white cores go cream. Matte overlays burnish to satin wherever a thumb has held them. Inks shift, and a warm store adds bow and edge lift. After 18 to 24 months a reference card no longer represents what was approved, while looking as authoritative as the day it was signed.
So the record carries numbers rather than impressions: thickness in millimetres at named points, the colour as the agreed target with its tolerance, the finish as a named lamination rather than as the word matte, and the method behind each measurement. ISO/IEC 10373-1 gives those methods names, which turns an acceptance clause into something two parties can repeat. Photographs belong in the record as identifiers and nothing more: one fixes which card and revision this is, and preserves neither colour nor surface feel. Card quality plans covers how a requirement is tied to the test that proves it.
Lot acceptance runs on a sampling plan, not on the reference. These are two separate gates answering two different questions.
Comparing one new card against the approved reference is a match check on one unit, inside the scope the record states. Deciding whether a delivered lot of 20,000 cards can be accepted is a sampling decision: it needs a plan, an inspection level and an acceptance quality limit agreed before the lot exists. ISO 2859-1 supplies those, and it is what an inspection clause should point at.
The useful sentence in a purchase order names both gates: the reference and its revision for the match check, and the plan with its level and limit for the acceptance decision. Specifying a card order covers where each of those sits in the order documents.
A changed requirement means a new reference, not an annotated old one. When a requirement moves, issue a new matched pair at a new revision and close the previous one.
Annotating is how a reference ends up approving a combination nobody ever produced. A note saying the thickness is now 0.84 mm leaves a 0.76 mm card in two sleeves as the approved object, and six months later somebody compares against the card rather than the note. The card wins that argument, because it is physical.
The rule that holds up: a new reference gets a new revision number and a new matched pair, and the record states which parts of the previous approval remain valid and which are superseded. The old pair is marked superseded and kept rather than destroyed, because a dispute about a batch shipped last year has to be settled against the reference that applied then.
Changes that should trigger a new reference rather than a note: substrate or core material, thickness, chip or inlay including a revision inside the same chip family, lamination or finish, print process, and a move to a different line or plant. Reordering without batch drift covers the variables that move between batch one and batch two even when nothing was formally changed.
Illustrative scenario, not a Kaway case study. A buyer specifies 20,000 access cards and approves a colour reference early, because the brand team needs to sign off the print. The sample is made on a plain PVC body with no inlay, the fastest way to get an accurate colour on the real substrate.
The record is written to match that reality. Scope: appearance only, covering print colour against the agreed target, registration and a matte overlay. Excluded, in the same paragraph: construction, chip reference, encoding and read behaviour. Measured values: the colour target with its tolerance, and the overlay by name. Two copies, one in the buyer's quality file and one retained at the plant, both sleeved with the record number. A second sample follows for the functional scope, production-representative with the real inlay and chip, encoded as the order will be encoded and tested on the reader model that will be installed. Its scope excludes colour.
Six weeks later a delivery looks slightly warm against the colour reference, and three things are already settled. The comparison is legitimate, because colour is inside the scope. It is a match check on the cards compared, not a verdict on 20,000 units, so the agreed sampling plan decides acceptance. And the chip questions in the same email get answered from the second record, not from a colour sample never scoped to answer them.
Kaway manufactures in its own plant in Dongguan, China, so the reference sample, the line it came off and the retained counterpart sit in one place rather than in three. Cards have been manufactured at the Dongguan plant since 2007, in a 1,300 m² production plant with 87 people.
A sample in your material and chip goes out before any order, which is the object a reference record is written against. Encoding is done at the plant, so a production-representative sample for a functional scope is built with the real chip and the real encoding. Every chip is read-tested individually before shipping, 100 per cent and not sampling, which is a check on the delivered cards and separate from any reference comparison.
Minimum order is 500 units. Standard production is 13 calendar days, 8 calendar days on express and 12 calendar days for wooden cards. Quote requests are answered the same working day. The card supplier qualification record covers the wider approval a reference sample sits inside, and the PVC and recycled PVC range lists the constructions a reference is issued for.
Discuss the specification and acceptance evidence needed for your card project with Kaway Group.
Yes, and that is the normal case. State the approval scope explicitly, exclude encoding in the same record, and use a separate production-representative sample with the real chip for the functional test.
As long as its revision applies, with one caveat: the card degrades. Treat the measured values in the record as the durable approval and the card as a convenience, and expect a reference older than about two years to have drifted in colour and finish.
Stop and investigate the references before judging the lot. Check both revisions, both approval dates and how each copy was stored. A disagreement between two copies of one reference is evidence about the references, and using either to reject a lot before that is resolved just moves the error downstream.
No. It answers whether a card matches, within its stated scope. Accepting a delivered lot needs a sampling plan with an inspection level and an acceptance quality limit, agreed before the lot is produced.
About this guide
Published by Kaway Group. The date on this page shows when this guide was last updated. If anything here is out of date, or contradicts what your integrator tells you, tell us and we will correct it.
Ask for a free sample in your material and chip.