Back

Product Authentication

Connected Product Architecture: From Physical Item to Digital Service Layer

Thomas


20 Oct 2026

abstract layers movement

A spirits brand launches a tap-to-verify bottle. The scan screen is polished. Six months later, a distributor ships pallets to the wrong market. The brand cannot prove it, because the identifier points to a vendor’s domain, the events sit in a proprietary format, and nothing connects to the ERP. Connected product architecture has six layers: carrier, read and verify, identity resolution, event record, services and integrations. Most programmes design only the scan experience.

Why do connected product programmes stall after launch?

The scan screen is the visible part. The failures sit underneath it. Typical drivers:

  • A static NFC tag or plain QR code is used for authentication, so a clone passes as genuine.
  • Verification runs on the phone, so a modified app can return “authentic” for anything.
  • Identifiers point to a vendor’s domain, which makes changing vendor mean re-tagging products already in market.
  • Events are stored in a proprietary format that no distributor, customs authority or internal system can read.
  • No feed from ERP or WMS exists, so a scan cannot be compared with the market the item was meant for.
  • Regulation arrives late in the design: the central EU Digital Product Passport registry has been live since 20 July 2026 (European Commission).

The scale of the underlying problem explains the pressure. Counterfeit trade reached USD 467 billion in 2021, up to 2.3% of global trade, and USD 117 billion of EU imports, or 4.7% (OECD/EUIPO, Mapping Global Trade in Fakes 2025). Clothing, footwear and leather goods account for 62% of seized counterfeits (OECD/EUIPO, 2025).

What is connected product architecture?

Connected product architecture is the technical structure that links a physical item to digital services through its own identifier. It defines what is on the product, who verifies it, where its identifier resolves, how its history is stored and which systems use that history.

The distinction that matters: A scan experience is one service. The architecture is everything that lets that service be trusted, reused and replaced. Each layer constrains the one above it.

A carrier that can be copied caps the trust any service can offer. An identifier the resolver cannot parse caps who can read it. An event model nobody else uses caps integration.

How does a connected product work, step by step?

  1. Encoding. At the production line, each item receives a unique identity: a GS1 Digital Link URL built from the GTIN plus a serial number, written to a carrier such as an NFC chip.
  2. Commissioning. The line station records the item in the event record with batch, line and time. Production data enters from the MES.
  3. Aggregation and shipment. Items are linked to cases and pallets. ERP or WMS supplies the intended market for each shipment.
  4. Logistics reads. UHF portals and handheld readers write received and shipped events as stock moves.
  5. The tap. A consumer, retailer or inspector reads the item. The phone opens a web address that contains a one-time code produced by the chip.
  6. Server-side verification. The platform recomputes the code, checks the counter against replays and adds context: tap location, shipped market, closure status.
  7. Resolution. The resolver decides what this audience sees: an authentication result, a consumer experience or passport data.
  8. Recording. The tap itself becomes an event. The history grows, and every service reads from it.

What are the six layers of a connected product?

Layer Question it answers Typical technologies
Services What do people get? Authentication, traceability, ownership, engagement, DPP
Integrations Which business systems feed and use the data? ERP, MES, WMS, CRM, brand protection
Event record What happened to this item? GS1 EPCIS 2.0 with extensions
Identity resolution Who is asking, and what should they see? GS1 Digital Link on a brand-owned domain
Read and verify Is this read genuine? Phones, handhelds, fixed portals, line stations
Data carrier What is physically on the item? NFC, secure QR, UHF RFID, protected barcode

The layers should be separable. A brand should be able to change carrier on new products without changing its record format, add a service without re-tagging, or move hosting without changing identifiers already in market.

A practical test: for each layer, write down who owns it, which vendor provides it, and whether the brand could replace that vendor without touching the layers above and below. Any “no” marks where lock-in sits.

How does NFC product authentication work at the carrier layer?

The reference chip class for phone-tap authentication is NXP’s NTAG 424 DNA (ISO/IEC 14443-A, NFC Forum Type 4). On each tap it increments a counter and computes a message authentication code with an AES-128 key that never leaves the chip. This is SUN, Secure Unique NFC messaging.

The code travels in the URL the phone opens. The platform recomputes it and checks the counter. The server sends nothing to the chip, so it is not a classic challenge-response. The effect is the same: a fresh, verifiable proof on every tap that a copy cannot reproduce.

An analogy helps. A basic NFC tag is a door with the key hanging next to it. A secure NFC tag is a door that generates a new one-time code each time someone approaches, and only the genuine door can produce the right one.

The TagTamper variant adds a loop across a closure and reports whether it has been opened. That turns “authentic” into “authentic and sealed”, the evidence needed to catch refilled bottles.

Phone compatibility is rarely a blocker. iPhone XS, XR and later read tags in the background with no app. Android reads natively with the screen on.

Physical fit matters as much as security. Standard antennas detune on metal, so watches and metal closures need on-metal tags with a ferrite layer. Small or curved formats need thin, flexible inlays. Liquids reduce read range.

NFC vs QR: security comparison

Criterion Cryptographic NFC (NTAG 424 DNA class) GS1 Digital Link QR
Proof per read Fresh signed message on every tap Static code unless a copy-detection feature is added
Cloning A copy cannot produce a valid code Photographable and reproducible
Tamper evidence Available with a tamper-loop variant Requires a separate seal design
Consumer access Tap, no app on iPhone XS/XR and later Any camera
Read range Close range Line of sight, any distance a camera can resolve
Marginal print cost Chip and inlay per unit Close to zero
Role in DPP Trust layer for authentication Visible carrier for compliance access

The practical framing for passports is NFC for trust, visible carrier for compliance. The ESPR (Regulation (EU) 2024/1781) is carrier-neutral. In practice, GS1 Digital Link QR is emerging as the default access carrier. NFC and QR are not regulatory equals. They do different jobs, and one item can carry both if they resolve to the same identity.

When is QR the right choice?

QR wins in several situations, and an honest architecture says so.

  • Very high volumes at low unit value. The printed QR adds roughly zero marginal print cost. The real cost is serialisation and platform (industry sources, 2026).
  • No secondary market. If nobody has an incentive to copy the item, a fresh signature per tap buys little.
  • Cost-constrained lines. Where adding an inlay to the line is not viable, print is the practical route.
  • Regulatory access. A visible code lets anyone reach passport data with any camera.

QR can also be hardened. ACM-secured barcodes and Diamond Dust add copy resistance to print, which suits packaging-level authentication. The verification method depends on the feature used. UHF RFID covers warehouse reads at distance but is not readable by consumer phones.

A mature architecture uses several carriers and binds them all to one item identity in the record. A UHF portal read and a consumer tap then land on the same history.

How much does an NFC tag cost?

Published converter prices for NTAG 424 DNA-class labels cluster around $0.45 to $0.65 per unit at five-figure volumes. On-metal variants add 7 to 30%. Quotes at 50,000 units and above are on request only (public catalogues, 2026).

Standard non-secure NFC tags of the NTAG213 class list around €0.19 to €0.23 at 1,000 to 10,000 units (public catalogues, 2026).

The gap between those two figures buys cryptographic proof. The tag is also only part of the cost. Line encoding, serialisation, the platform and integrations decide the total.

How does server-side verification work, and why does it matter?

Verification must happen on the server. Anything decided on the phone can be faked by an app. The read layer collects the carrier output and context. The platform decides.

Context turns a scan into a finding:

  • Authentic is the minimum.
  • Authentic, sealed, and in a market it was not shipped to is a diversion case.
  • Authentic but opened, presented as new in a bar is a refill case.

Non-phone readers belong to this layer too. Fixed UHF portals, distributor handhelds and line encoding stations produce events that matter more for traceability than any consumer scan. Two decisions are often left to default:

  • Offline behaviour. A line station that loses connectivity should buffer timestamped events and sync later, not drop reads.
  • Reader credentials. A distributor’s handheld writing a “received” event needs authentication. Otherwise anyone can write into the record.

What does an identity resolver do?

A resolver turns an identifier into the right response for the right audience. With GS1 Digital Link, one URL can send a consumer to an experience, a retailer to an authentication result and an authority to passport data.

Think of it as a switchboard. Every product carries one address. The resolver checks who is asking and why, then routes the request. One tag serves three audiences without printing three codes.

Own the domain. If identifiers on millions of products point to a vendor’s domain, changing vendor means re-tagging or permanent redirects. Language and market logic also belongs here. The same item scanned in Tokyo and Paris can serve different content and legal notices without separate tags per market.

Which event standard should connected products use?

GS1 EPCIS 2.0 is the most widely adopted event model. It carries production, packing, shipping, receiving and transformation events in a common shape: what, when, where and why.

Product identity needs two more things: verification results and ownership events such as registration and transfer. These can be modelled as extensions, so the core stays interoperable.

Two rules keep the record usable as evidence. Events are added, never edited. Events follow a common standard, so a distributor or customs authority can read them without custom translation.

This is where value accumulates. A brand can answer questions it now answers with estimates: where allocated stock actually sold, which shipment leaked, which units of a batch are still in circulation.

Connected product architecture by sector

The six layers are the same everywhere. The carrier, the key integrations and the first service change.

Sector Carrier layer Key integrations Lead services
Spirits TagTamper NFC in capsule; UHF on cases Line commissioning, ERP intended market Refill evidence, diversion, engagement
Luxury leather goods Sewn-in NFC Boutique sales, CRM, client services Authentication, ownership, DPP
Watches On-metal NFC or certificate card After-sales, warranty systems Service history, resale transfer
Beauty and fragrance NFC or secured codes on pack High-SKU commissioning, distributor data Diversion detection, reorder journeys

Textiles and apparel are the first product group expected under delegated acts for the DPP, indicatively 2027. Application starts at least 18 months after each act, so enforcement is likely around 2028 to 2029. Batteries follow their own passport from 18 February 2027.

What should a connected product RFP ask?

Functional demos rarely cover what decides production success. Specify these:

  • Carrier proof. Name the chip or security feature per carrier, and show a cloned tag being rejected.
  • Verification. Server-side, with counter and replay handling described.
  • Identifiers. GS1 Digital Link on a brand-controlled domain.
  • Events. A standard model (EPCIS 2.0), exportable in full.
  • Line fit. Support for your line speeds and co-packers.
  • Intended market. Taken from ERP or WMS at shipment.
  • Access. Role-based views for consumer, partner, authority and brand from one identifier.
  • Operations. Stated latency, availability, data residency and longevity commitments, including export format and key escrow.

A down verification service makes genuine products look suspicious. Availability is a brand issue, not only an IT one.

How do you migrate from a campaign architecture?

Brands with earlier campaign tags in market do not need to start again.

  1. Inventory carriers, domains, records and products.
  2. Classify by security. Cryptographic carriers stay. Static tags and plain QR codes can stay for engagement but should not be used for authentication.
  3. Take control of resolution. Redirect existing identifiers through a resolver the brand controls.
  4. Unify the record. Import historical items and events into the standard model.
  5. Apply the full architecture to new production. Secure carrier and full event chain from day one.

40+ brands · 12 years · €1.5B protected

Connected products reward the teams that design the layers nobody sees: the carrier, the resolver, the record and the integrations. Selinko, founded in 2013 in Belgium and part of Toppan since 2026, reviews these layers with brands before they commit to a vendor.

FAQs

What is connected product architecture?

Connected product architecture is the layered structure that links a physical item to digital services. Its six layers are the data carrier, read and verify, identity resolution, event record, services and enterprise integrations. Decisions at the lower layers set the limits of what the services above can do.

Should authentication be verified on the phone or on the server?

Authentication should be verified on the server. A phone app can be modified to return “authentic” for any input. The server recomputes the chip’s code, checks the counter against replays and adds context such as market and closure status before deciding what to show.

What is an identity resolver?

An identity resolver is the layer that maps a product identifier to the right response for each audience. With GS1 Digital Link, one URL can route a consumer, a retailer or an authority to different content. The brand should own the resolver domain to avoid vendor lock-in.

Which event standard should connected products use?

GS1 EPCIS 2.0 is the most widely adopted event standard. It records what, when, where and why in a common format. Verification results and ownership events can be added as extensions, keeping the core readable by distributors, customs authorities and internal systems.

Which NFC chip is used for product authentication?

NTAG 424 DNA is the reference chip class for phone-tap authentication. It follows ISO/IEC 14443-A and NFC Forum Type 4. It uses AES-128 and SUN messaging to produce a fresh authentication code on every tap. ISO 15693 chips suit supply chain scanning, not phone authentication.

Can one product carry more than one data carrier?

Yes. One item can carry NFC for consumer authentication, UHF for logistics and a GS1 Digital Link QR for passport access. All carriers must resolve to the same item identity in one record, so every read lands on a single history.

Do connected products need an app?

No. iPhone XS, XR and later read NFC tags in the background without an app. Android reads tags natively with the screen on. A QR code opens in any camera app. The verification itself runs on the platform, not on the phone.

Review your architecture

Blog

Discover more articles

All our articles