A payment card network is the infrastructure that connects the bank that issued your customer’s card with the bank that holds your merchant account. When a cardholder pays, the network carries the authorization request from the acquiring bank to the issuing bank and then returns the response. Without that relay, the transaction doesn’t move.
Visa and Mastercard dominate this infrastructure, together accounting for more than 80% of U.S. credit card transaction volume, with a combined average interchange rate that reached 2.36% in 2025 (Nilson Report via Clearlypayments.com). The other two global networks, American Express and Discover, run a structurally different model, and several regional networks carry significant volume in specific markets. Each has its own fee schedule, risk models, and rules that affect what happens when a transaction hits the network.
For developers building or maintaining payment integrations, understanding where the card network fits in the stack matters for two practical reasons: it determines which fees apply to which transactions, and it determines which routing options are available when a card can be processed across more than one network.
What is a payment card network?
A payment card network is a scheme that sets the technical standards and rulebooks governing how card-based transactions are authorized, cleared, and settled. The network owns the rails, defines the message format (ISO 8583 for legacy systems, JSON-based APIs for newer integrations), sets interchange rate schedules, and adjudicates disputes between issuers and acquirers.
The four global networks are Visa, Mastercard, American Express, and Discover. Visa processed 257.5 billion transactions in fiscal year 2025 and operates infrastructure capable of processing over 65,000 transactions per second (Visa Fact Sheet). Visa and Mastercard together, along with China UnionPay, process 97% of all credit card transactions worldwide (CoinLaw, 2026).
Networks are categorized as open or closed. Visa and Mastercard are open networks: they license their rails to thousands of issuing banks and acquiring banks, who issue cards and accept merchant accounts under the network’s brand. The network itself doesn’t issue cards or hold merchant accounts. American Express and Discover operate closed networks: they act as both the network and the issuer, issuing cards directly to cardholders and managing merchant relationships themselves. This structural difference affects both the fee economics and the number of parties a transaction touches.
Regional networks, including China UnionPay, India’s RuPay, and France’s Cartes Bancaires, operate within specific geographies and in some cases are co-badged alongside global networks on the same card.
The four-party model: how a card network fits into a transaction
An open-network transaction involves four distinct parties:
- The cardholder, the person making the purchase using a card issued by their bank.
- The issuing bank, the bank that issued the cardholder’s Visa or Mastercard. It approves or declines the authorization request based on available funds, fraud signals, and risk models.
- The acquiring bank, the bank that holds the merchant’s account and processes card payments on the merchant’s behalf.
- The card network, the infrastructure connecting the acquirer and issuer, carrying authorization requests, clearing messages, and settlement instructions.
When a cardholder pays, the payment flows outward: the acquiring bank sends an authorization request through the card network to the issuing bank. The issuer returns an approval or decline, which travels back through the same network to the acquirer and then to the merchant’s terminal or checkout.
Clearing and settlement happen separately, typically within one to two business days. The network oversees the transfer of funds from the issuing bank to the acquiring bank, minus interchange fees that flow to the issuer.
American Express and Discover collapse this into a three-party model: the network itself is the issuer, so there’s no separate issuing bank in the authorization loop. This gives them more control over underwriting and pricing but limits card acceptance in markets where their network coverage is thinner than Visa or Mastercard.
Card network vs. payment processor vs. acquirer: what’s the difference
The three terms are often confused because some companies play more than one role.
Role What it does Card network Owns the rails and the rulebook. Doesn’t touch the merchant’s money directly and doesn’t hold accounts on either side. Payment processor Sends the transaction to the appropriate network on the merchant’s behalf. Provides the API or [gateway](/glossary/#gateway-token) a merchant’s code talks to, manages PCI scope, and handles communication with the acquiring bank. Acquiring bank Holds the merchant’s account, assumes [chargeback](/glossary/#chargeback) and credit risk, and settles funds from the card networks into the merchant’s account.
Processors can be standalone (Stripe, Adyen, Braintree) or tied to a specific acquirer; in some cases the acquirer and processor are the same company, in others they are separate.
A clean summary: the acquirer is responsible for the money; the processor is responsible for the message. The card network is responsible for the rules and the relay between them.
When you integrate with a PSP like Stripe or Adyen, you’re working with a processor (and often a bundled acquirer). The card network is upstream of that integration and mostly invisible unless you’re looking at interchange fees or implementing network-specific features like network tokenization.
Major card networks compared: Visa, Mastercard, Amex, Discover, and regional networks
Network Model 2025 active cards Notes Visa Open 4.81 billion Largest global network by transaction volume; processed 257.5B transactions in FY2025 Mastercard Open 3.39 billion Second largest; tends to have slightly higher interchange on premium cards American Express Closed ~130 million Higher merchant fees, affluent cardholder base, lower acceptance in some markets Discover Closed ~57 million US-focused; also operates the PULSE debit network China UnionPay Open/hybrid ~10 billion Dominant in China, significant for cross-border merchants serving Chinese travelers RuPay Open ~700 million India’s domestic network, government-promoted via the National Payments Corporation of India Cartes Bancaires (CB) Open ~80 million French domestic network; co-badged with Visa or Mastercard on most French cards
Acceptance matters as much as volume when deciding which networks to support: the fee and coverage gaps in the table above cut into the reach that raw card counts suggest. For developers building multi-market checkout flows, the practical question is which networks your processor supports routing to, not just which cards a user might carry.
How card networks make money: network fees vs. interchange
The fee a merchant sees on a card transaction typically has three components:
- Interchange is the largest component, set by the card network’s published schedule and paid to the issuing bank as compensation for credit risk and reward programs. For U.S. consumer credit cards, interchange averages 1.5%-2.5% plus a per-transaction cent amount. Visa and Mastercard don’t receive this fee; they set the schedule.
- Assessment fees are paid to the card network itself for use of its rails. These average around 0.14% of transaction value (Swipesum, 2025), a fraction of what goes to the issuer. This is the fee the network actually collects.
- Processor markup is what the payment processor charges on top of interchange and assessment, in exchange for handling the technical integration and risk.
Confusing interchange with network fees is a common error when analyzing processing costs. A merchant isn’t paying Visa that average interchange rate per transaction; they’re paying the issuing bank approximately that amount, with Visa taking a small assessment fee off the top.
Assessment fees are roughly: Visa 0.13%, Mastercard 0.1375%, Amex (variable, typically higher). These numbers shift when Visa or Mastercard update their schedules, typically twice a year.
Total U.S. merchant card processing fees reached $198.25 billion in 2025, an annual record, according to the Merchant Payments Coalition citing Nilson Report data.
Network tokenization and why it matters for card-on-file transactions
Network tokenization replaces a stored card number (PAN) with a token issued and managed by the card network itself. When a cardholder’s card is reissued or its expiration date changes, the network updates the token automatically, without requiring the cardholder to re-enter their card or the merchant to trigger an account updater job.
The authorization-rate benefit is quantifiable.
Key stat: Visa’s network tokenization lifts card-not-present authorization rates by 4.6%; Mastercard’s by 2.1% (Solidgate, 2024-2025).
The mechanism is that the card network pre-validates each token before the authorization request reaches the issuing bank, providing a stronger signal than a raw PAN submission. Mastercard reports 5-8% fewer false declines from token-based transactions (NMI).
Visa provisioned 11.5 billion network tokens by the end of fiscal year 2024, and tokenized transaction volume grew 44% year-over-year in 2024, corresponding to a 6% improvement in approvals and a 30% reduction in fraud (Visa corporate data via AInvest, 2025). Tokenized transactions are projected to double from 283 billion in 2025 to 574 billion by 2029.
For developers managing card-on-file subscriptions or recurring charges, network tokenization is the more reliable alternative to raw PAN storage or third-party account-updater services. Visa also offers a 0.10% interchange cost reduction on consumer CNP credit card transactions that use a network token.
The tradeoff: network tokenization requires working with a processor or vault provider that supports it, since the token provisioning process goes through the card network’s token service provider. Not all processors expose this as a standard integration option.
Routing across card networks: co-badged cards and network selection
How network choice varies by card type:
Card type Network locked? Who chooses Credit (single-branded) Yes No choice, network fixed by the card’s primary brand Debit (US, Regulation II) No Merchant or acquirer selects among at least two unaffiliated networks Co-badged (EU, Art. 8) No Cardholder chooses; automatic override mechanisms are prohibited
Most credit card transactions are locked to a single network by the card’s primary brand. A Visa credit card routes through Visa. A Mastercard routes through Mastercard. There’s no choice to make.
Debit cards work differently. U.S. debit cards must carry at least two unaffiliated networks under the Durbin Amendment (Regulation II), enabling merchants and acquirers to choose which network processes the transaction. The Federal Reserve publishes average debit interchange by network specifically because of this competitive routing requirement.
In the EU, Regulation (EU) 2015/751, Article 8, extends similar logic to co-badged cards more broadly: when a card carries both a global network brand and a domestic or regional scheme (a Visa card also carrying Cartes Bancaires, for example), merchants who accept both networks must let the cardholder choose which one processes the transaction. Card schemes, processors, and acquirers are prohibited from inserting automatic mechanisms that override that choice, though merchants may configure a default priority that the cardholder can override.
The practical implication for checkout implementations: in the EU and other markets with co-badging rules, a compliant checkout flow needs to surface network choice to the cardholder when relevant, and the routing logic can’t silently force one network over another.
Authorization rates vary by network routing for the same underlying transaction. Networks apply different fraud-scoring models, and issuers respond differently depending on which network routed the request. For co-badged or multi-network scenarios, routing the same transaction through a different eligible network can produce a different authorization outcome.
How Orchestra routes transactions across card networks
Orchestra’s routing logic operates one layer above the card network, at the processor and PSP level. When a transaction comes in, Orchestra evaluates which connected processors are available, their current authorization rates, their fee schedules, and any rules you’ve configured, then routes to the optimal option.
For network-specific routing decisions, the relevant integration points are:
- Network tokenization support: Orchestra integrates with processors that support Visa and Mastercard network tokens for card-on-file transactions, making the authorization-rate uplift available without a separate vault integration.
- Co-badged card routing: when a transaction involves a card that can route across multiple networks and your processor supports network selection, Orchestra’s routing rules can encode the network preference alongside processor preference.
- Failover across processors: if your primary processor fails or declines a transaction, Orchestra automatically retries through a secondary processor, which may access the card network through a different acquiring relationship and produce a different authorization outcome.
Orchestra connects to 90+ payment providers through a single JavaScript library integration. The routing engine described in our payment routing optimization guide handles the processor selection layer; card network mechanics are enforced upstream by whichever network the processor routes to.
For teams building multi-PSP strategies, the card network layer is mostly static, but the processor layer that sits above it is where routing decisions, failover, and cost optimization happen. A multi-processor strategy improves authorization rates by changing which processor’s relationship with that network handles the transaction, not by changing the network itself.
See also: a guide to payment routing for routing rule configuration, and payment authorization vs. settlement for where the card network fits in the full transaction lifecycle.
Frequently asked questions
What is the difference between a payment card network and a payment processor?
The card network (Visa, Mastercard, Amex, Discover) sets the rules and carries authorization messages between the issuing bank and the acquiring bank. The processor is the technical layer that sends transactions to the network on the merchant’s behalf. A merchant’s integration code talks to the processor; the processor talks to the network. Some companies, like Adyen or Stripe, bundle both processor and acquirer functions.
How many major payment card networks are there?
Four global networks dominate: Visa, Mastercard, American Express, and Discover. Visa and Mastercard operate open-loop models with separate issuing and acquiring banks. Amex and Discover operate closed-loop models where they act as both network and issuer. Regional networks including China UnionPay, RuPay, and Cartes Bancaires carry significant volume in specific markets.
What is the four-party model in card payments?
The four-party model describes open-network transactions: the cardholder, the issuing bank (who issued the card), the acquiring bank (who holds the merchant’s account), and the card network routing messages between issuer and acquirer. Amex and Discover collapse this into a three-party model by acting as both issuer and network.
How do card networks make money if they don’t collect interchange?
Card networks charge assessment fees, a small percentage of transaction volume (around 0.14% on average) paid by issuers and acquirers for use of the network’s rails. Interchange, the larger fee merchants notice, is set by the network’s published schedule but paid to the issuing bank, not the network itself. Visa and Mastercard set interchange rates; they don’t receive them.
What is network tokenization and why does it matter for recurring payments?
Network tokenization replaces a stored card number with a token managed by Visa or Mastercard. When the underlying card is reissued or its expiration date changes, the network updates the token automatically. For recurring charges or subscriptions, this reduces failed-payment retries from stale card data. Network tokens also lift authorization rates: 4.6% for Visa CNP transactions, 2.1% for Mastercard (Solidgate, 2024-2025).
Can a single transaction be routed across more than one card network?
For co-badged debit cards in the US (under Regulation II) and co-badged cards in the EU (under Regulation (EU) 2015/751), yes. The merchant or acquirer can select which eligible network processes the transaction. For credit transactions on single-branded cards, the network is fixed by the card’s primary brand and can’t be changed at routing time.
Does card network choice affect authorization rates?
Yes. Networks apply different risk models, and issuers respond differently depending on which network routed the request. For co-badged or multi-network scenarios, routing through a different eligible network can produce a different authorization outcome on the same card. This is why routing logic that accounts for network eligibility, not just processor eligibility, matters for approval-rate optimization.