Reference🛠️ Merchant setupAdvanced⏱ 16 min read

🪙 Managing tokenized cards

PAN, PSP token, network token: three representations of the card with very different properties. Card-on-file, the CIT/MIT framework, account updaters, and the token lifecycle.

Clear PAN, PSP token, network token: three different things

The PAN (Primary Account Number, 13 to 19 digits) is the number that identifies the card account at the issuer. A compromised PAN opens the door to fraud, and storing it puts the merchant in full PCI DSS scope. Two levels of tokenization can replace it in merchant systems. The PSP token (or acquirer token) is an internal reference generated by the provider. The network token (DPAN) is issued by the card network through its token service provider (TSP), with the issuer's approval. At Visa, the TSP is VTS (Visa Token Service); at Mastercard, it's MDES (Mastercard Digital Enablement Service).

OwnershipClear PANPSP tokenNetwork token (VTS/MDES)
Issued byCard issuerPSP/gatewayNetwork (TSP), approved by the issuer
Scope of useUniversalLimited to the PSP that issued the tokenAny acquirer, but bound to the merchant (TRID) / domain pair
Merchant PCI DSS burdenMaximum (SAQ D)Minimal (the PSP holds the PAN)Minimal
Update when the card is reissuedManual or account updaterAccount updater via the PSPAutomatic: the token survives PAN reissuance
Per-transaction cryptogramNoNoYes (TAVV/DSRP): dynamic credential
Impact on authorization rateReference≈ same as the PAN (it's a PAN on the network side)+2 to +3 pts (Visa 2024), e-commerce fraud −26%
Portability to another PSPYes (if PCI compliant)Negotiable; THE lock-in pointYes in theory (re-provisioning), with the networks' cooperation
Cost–Included or marginalNetwork fees per token/transaction (a few cents or bps)
Clear PAN vs. PSP token vs. network token
Customerenters the PAN onceMerchant / PSPnever stores the PANToken ServiceVisa VTS · Mastercard MDESIssuerapproves the tokenPANtoken requestTARDPAN (network token)bound to one card × merchant pairprovisioningSubsequent paymentstoken + dynamic cryptogramone-click / MITCard reissued or expired: the token staysvalid (updated by the scheme) →+2 to 3 pts of approval rate
🔑
The two tokens don't compete; they stack
In an online merchant's target architecture, the PSP token (or a token from an independent vault) serves as the internal pivot. It is the only reference stored in the CRM, the OMS, and the subscription database. The network token serves as the authorization credential sent to the network, with a fallback to the vaulted PAN when the network token is unavailable or performs worse on a given BIN.

Card-on-file: the stored card

Card-on-file (COF) means storing a payment credential for future use, such as one-click checkout, subscriptions, auto top-up, and merchant wallets. It is the economic foundation of recurring revenue models. In this segment, the quality of tokenization determines the share of charges that fail, and therefore the level of churn.

≈ €175B
French e-commerce in 2024, with a growing share from one-click purchases and subscriptions
FEVAD, 2025
25-30 %
of the card base reissued each year (expiry, loss, renewal)
3- to 4-year issuance cycles
20-40 %
of subscription churn is involuntary, caused by payment failure rather than a customer decision
subscription economy studies
  • Explicit consent at the time of storage (a dedicated checkbox, clear wording), required by network rules on stored credentials;
  • The enrollment transaction must be authenticated with SCA (even at €0, as an account verification), because it anchors the subsequent MIT chain;
  • Show the customer the stored card (brand + last 4 digits + expiry date) and allow one-click deletion;
  • When a card is close to expiry: a proactive banner + account updater, rather than waiting for the charge to fail;
  • Store only the token reference on the merchant side, never the PAN, and never the CVV (storing it is prohibited in all cases).
ℹ️
COF ≠ MIT
Card-on-file describes how the credential is stored; the CIT/MIT distinction describes who initiates the transaction. A one-click payment with a stored card is a CIT, since the customer is in session and triggers the transaction. A monthly subscription charge is an MIT. Confusing the two is the leading cause of incorrect network flagging.

The CIT/MIT framework and mandatory network flagging

Flagging a stored-credential transaction tells the network who triggered it: the cardholder or the merchant. The flag is mandatory on every such transaction. There are two flags. A CIT (Customer-Initiated Transaction) covers cases where the cardholder is present and acting. An MIT (Merchant-Initiated Transaction) covers cases where the merchant acts alone, under a mandate given during an initial, authenticated CIT. The requirement comes from Visa's Stored Credential Framework (2017), later adopted by Mastercard, and then from PSD2. MITs are out of scope for SCA, but only if the flagging chain is correct.

the issuer sends it backstored by the merchantInitial CIT3DS2 or zero-value + mandatenetwork transaction IDTransaction ID (Visa) · Trace ID (MC)Vaultto store and migrateMIT recurringfixed amount and dateMIT installment3 interest-free installmentsUCOFusage-based, variableResubmissionretry after 51referenced in EVERY MITMIT flagged correctlyoutside SCA scope, no CVV expectedMIT sent as a CITsoft decline 1A / 65 in a loopreference passed throughreference missing or wrongcardholder present (CIT)merchant only (MIT)flagging errorAn orphan MIT is not outside SCA scope: to the issuer, it is an unauthenticated transaction.
MIT typeDefinitionExampleAmount / schedule
RecurringCharges of a fixed amount at a fixed frequency€13.99/month streaming subscriptionFixed / fixed
InstallmentA single purchase paid in several installments3 interest-free installmentsFixed / fixed, set number
Unscheduled COF (UCOF)Usage-based charge, no scheduleAuto top-up of a mobility account when the balance drops below €5Variable / variable
IncrementalIncrease to an existing authorizationExtended hotel stay, minibarAdded to the initial authorization
ResubmissionResubmission after an insufficient-funds declineRetry of a subscription charge declined with 51Same amount as the original
Delayed chargesAdditional charge after the service is deliveredMissing fuel on a returned rental carAfter the main CIT
No-showCharge for a reservation that wasn't honoredHotel night not canceledUnder the terms accepted at the CIT
ReauthorizationNew authorization for a delayed orderSplit shipment, preorderReplaces an expired authorization
MIT types (network categories)

Network chaining links each MIT to the CIT that authorized it. The initial SCA-authenticated CIT returns a network transaction identifier, called the Transaction ID at Visa and the Trace ID at Mastercard. The merchant must store it and reference it in every subsequent MIT, along with the MIT type indicator. That reference proves to the issuer that a mandate exists, which is what justifies skipping SCA on the charge.

Flagging a recurring MIT (typical fields, annotated JSON)
{
  "amount": { "value": 1399, "currency": "EUR" },
  "paymentMethod": { "storedPaymentMethodId": "tok_8f3a2b" },  // vaulted token
  "shopperInteraction": "ContAuth",         // cardholder not present
  "recurringProcessingModel": "Subscription",  // MIT type: recurring
  "additionalData": {
    "networkTxReference": "015821943260929" // Transaction ID of the initial CIT
  }
}
// Golden rules:
// 1. The enrollment CIT was authenticated with 3DS (proof of mandate).
// 2. Every MIT references the networkTxReference of the initial CIT.
// 3. No CVV on an MIT (it cannot be stored): this is expected,
//    and the issuer should not penalize it if the flagging is correct.
⚠️
Mis-flagged MITs = a decline machine
A recurring charge sent as a standard e-commerce transaction (CIT) without 3DS triggers a soft decline (1A/65) at a European issuer. Since the cardholder isn't in session, there's no way to resolve that decline. The result is repeated failures, involuntary churn, and billed retries. Conversely, flagging a customer-present transaction as an MIT deprives the merchant of authentication and the liability shift. Auditing CIT/MIT flagging is therefore the first workstream in any program to optimize recurring payments.

Account updaters: RTAU, VAU, and ABU

An account updater is a network service that sends the merchant, through its PSP, the new details of a reissued card. When a card is reissued after expiry, loss, theft, or a product change, the stored PANs become obsolete. The networks run update services: VAU (Visa Account Updater) in batch and RTAU (Real-Time Account Updater) in real time at Visa, and ABU (Automatic Billing Updater) at Mastercard. The PSP queries the service overnight for the entire stored base, or on the fly when an authorization would otherwise fail. The vault is then updated.

  • New PAN: the card has been reissued; replace the stored credential;
  • New expiry date: PAN unchanged; extend;
  • Account closed: stop attempts and notify the customer (to avoid billed retries);
  • Contact cardholder: the issuer asks the merchant to have the cardholder update the card.
+3 to +5 pts
success rate on recurring MITs with an active account updater
PSP subscription benchmarks
J-15
the right time to run a proactive update before a subscription's billing date
market practice
ℹ️
Network tokens make account updaters (almost) unnecessary
A network token is maintained by the issuer across reissues. The DPAN stored by the merchant keeps working when the underlying PAN changes. Account updaters remain useful for the base of non-tokenized PANs and for networks or BINs not covered by VTS/MDES. Their role shrinks to that residual scope, whereas they used to be the main update mechanism.

The network token lifecycle

A network token is a stateful object, managed jointly by the TSP, the issuer, and the token requestor: the PSP acting on the merchant's behalf, identified by its TRID (Token Requestor ID). The token moves between states over the life of the underlying card, and the merchant has to process these lifecycle events to keep its vault up to date.

Cardholderenters or scans the PANWalletApple / Google · device signalsTSPVisa VTS · Mastercard MDESIssuerdecidesPAN + CVV2tokenization + device dataID&V requestlow riskmedium riskhigh riskGreen pathdirect approvalYellow pathadditional verificationRed pathdeclineapproval in the banking appSMS OTP alone = fraud vectorDPAN provisioneddevice-bound token with its own expirySecure Element: keys in the chipHCE: LUK keys refreshedPaymentDPAN + dynamic cryptogram + CDCVM = SCAfrictionless pathadditional verificationdecline / fraud vectorA fraudster who gets the PAN + SMS OTP adds the card to THEIR wallet and pays with THEIR biometrics.
Provisioning an e-commerce network token
Merchant / PSP
Tokenization request
The PSP (token requestor, TRID) sends the PAN to the network's TSP
TSP (VTS/MDES)
Checks and approval request
BIN eligibility check, forwarded to the issuer
Issuer
Approves (or declines) provisioning
Risk-based decision; possible step-up (ID&V)
TSP
Generates the DPAN
Token + its own expiry date, bound to the TRID and the usage domain
PSP
Stores the token and obtains cryptograms
For each transaction: requests a TAVV/cryptogram from the TSP
State / eventTriggerImpact on the merchant
ActiveProvisioning approvedToken usable for authorization with a cryptogram
SuspendedIssuer (suspected fraud, temporary block)Authorizations declined; do not delete, since the token can become active again
Reissued / updatedUnderlying card reissuedThe token stays valid, with no change on the merchant side (that's the whole point)
TerminatedAccount closed, permanent block, cardholder requestDelete the credential, notify the customer, stop MITs
Metadata updateNew card art, new token expiry dateUpdate the display (last4, card art) in the customer interface
Lifecycle states and events

These events arrive through the PSP's lifecycle webhooks/notifications. An integration that ignores them lets suspended or terminated tokens pile up in its vault, and the state change only surfaces when a charge fails. The merchant then ends up in exactly the failure scenario that tokenization was supposed to prevent.

PSP configuration and measured gains

Network tokens are turned on through configuration at the PSP. That covers enrolling the merchant as a token requestor (the PSP usually shares its own TRID across merchants) and choosing the scope by card brand, BIN range, and channel. It also sets the fallback policy to the PAN and the rules for displaying the payment method. The table below summarizes the usual options.

ParameterOptionsRecommendation
Tokenization scopeAll COF / recurring only / on the fly on CITsAll COF, with asynchronous provisioning as soon as the card is enrolled
PAN fallbackNone / automatic if token unavailable / BIN-drivenAutomatic + BIN-level steering: some issuers respond less well to tokens, so measure it
Credential choice per transactionAlways token / A/B test by issuerData-driven routing: token by default, PAN on BINs where the token underperforms
CryptogramTAVV per transaction (default)Make sure the cryptogram is always present on CITs; some MIT flows don't need it
Customer displaylast4 of the PAN (recommended) vs. last4 of the tokenAlways the last4 of the PAN: customers don't know their token
LifecycleWebhooks on/offEnable all events + a monthly vault reconciliation job
Network token settings at a PSP: options and recommendations
+2 to +3 pts
authorization rate on card-on-file migrated to network tokens
Visa, 2023–2024 communications
−26 %
e-commerce fraud rate observed on Visa tokenized transactions
Visa, 2024
> 10B
network tokens issued worldwide by Visa (milestone reached in 2024)
Visa
⚠️
Tokens are also a lock-in tool
PSP tokens are not portable by default. Switching providers without a migration plan means losing every stored card, and with them the subscribers. Exporting PANs to a third-party PCI processor (an encrypted vault-to-vault migration) and/or using network tokens that can be re-provisioned should be negotiated in the contract at signing. Some merchants treat this as a dealbreaker when selecting a PSP.