A coffee shop in Tokyo accepts a Visa card issued by a small credit union in Ohio.
The barista has never heard of the credit union. The credit union has never heard of the coffee shop. There is no direct contract between them, no shared customer database, and no direct communication channel.
Yet, in under two seconds, the transaction is authorized.
How?
The answer lies in what happens behind the scenes. When you tap or swipe your card, you aren’t simply initiating a payment. You are executing a query across one of the most sophisticated distributed networks ever built—a global four-party system that allows millions of independent nodes to achieve state consensus without direct peer-to-peer connections.
1. The Four-Party Topology
Paying with cash is a simple, local state mutation: You and the Merchant. The exchange is immediate and final.
A card payment operates on a fundamentally different topology. Underneath the UI, it involves four distinct participants.
1.1. The Cardholder: The entity initiating the transaction. 1.2. The Merchant: The business selling the goods or services. 1.3. The Issuer: The financial institution that issued the card and holds the underlying ledger. It extends credit (or holds the deposit account for debit cards). Examples include Chase, Capital One, or your local credit union. 1.4. The Acquirer: The acquiring bank (often working with a payment processor like Square or Stripe) that provides the merchant’s ingress to the network and ultimately funds their account.
One of the biggest misconceptions about card payments is that Visa and Mastercard generally do not issue consumer credit cards or extend credit themselves. Instead, they operate the payment network.
They provide the routing protocols, security standards, and global messaging infrastructure that allow independent issuers and acquirers to transact securely. Think of them as a stateless, global message bus that routes authorization requests between isolated nodes.
2. The Card Payment Lifecycle
A card payment operates on a strict two-phase commit protocol: synchronous authorization, followed by asynchronous clearing and settlement.
2.1. Phase 1: Authorization — The Distributed Consensus
When you tap your card, the merchant’s edge device sends an authorization request through the Visa or Mastercard network to your issuing bank.
The request essentially asks: “Someone presenting this credential wants to execute a $5 state change in Tokyo. Is the credential valid? Does the ledger have sufficient available balance? Does the fraud scoring system flag this?”
Your issuer evaluates the request:
- Available credit or account balance
- Credential status
- Real-time fraud detection algorithms
- Risk limits
If the constraints are satisfied, the issuer replies: “Approved.”
At this point, the issuer’s ledger updates the available_credit through a temporary hold or “memo post.” The merchant hands you the coffee.
Importantly, no money has moved. Authorization is not a funds transfer; it is a guaranteed promise to settle at a later time.
2.2. Phase 2: Clearing and Settlement — The Async Reconciliation
After the transaction is authorized, the merchant still needs to receive the funds. This happens through two separate, asynchronous processes.
2.2.1. Clearing — Reconciling State Later—often in batches—the merchant submits its authorized transactions for clearing. The acquiring side pushes these transaction details through the card network to the relevant issuers.
During clearing:
- Transaction data is exchanged between the merchant, acquirer, network, and issuer.
- The network calculates the net financial obligations between participating institutions.
- Network routing and interchange fees are applied.
Clearing answers the question: “What state changes occurred today, and what is the net delta between institutions?”
2.2.2. Settlement — Executing the Net Transfer After clearing, settlement takes place. But just like domestic batch rails (ACH) or SWIFT, individual $5 coffees are not wired back and forth. The system relies on Netting.
At the end of the day, the network calculates the net position between all acquiring banks and all issuing banks. If Acquirers are owed $100 million total, and Issuers owe $80 million total, they only settle the $20 million net difference.
This net settlement is executed through designated settlement banks or central bank RTGS systems. Once settlement is completed, the Acquirer pushes the funds to the merchant’s actual bank account. Because of this batching process, merchants typically receive funds on a T+1 or T+2 basis.
3. The Cost Architecture: Merchant Discount Rate (MDR)
Every card transaction involves a cost structure deducted from the merchant’s transaction, expressed as the Merchant Discount Rate (MDR). This is the funding mechanism that keeps the distributed network operational.
The MDR is typically made up of:
- Interchange Fee: Paid to the issuing bank to subsidize the cost of the credit risk and ledger infrastructure.
- Network Assessment Fee: Paid to Visa or Mastercard for operating the global routing infrastructure.
- Processor or Acquirer Fee: Retained by the acquiring bank or payment processor for maintaining the merchant’s ingress gateway and risk management.
4. The Subsidy Model: Why Rewards Exist
What happens to the interchange revenue received by the issuer?
A significant portion of this revenue is used to fund the issuer’s ecosystem:
- Cash-back rewards, airline miles, and hotel points
- Real-time fraud detection infrastructure and customer service
- Credential production and technology overhead
- Credit losses (when users default) and regulatory compliance
- Profit
From an architectural perspective, rewards programs are a deliberate subsidy. They are funded by the MDR to drive transaction volume through the issuer’s node on the network.
5. The Trade-Offs: Why Merchants Accept the Fees
Giving up 2–3% of a sale creates a direct margin hit. So why do merchants accept this cost?
- Guaranteed State Consensus: An approved authorization provides mathematical assurance from the Issuer that the funds are reserved, drastically reducing payment uncertainty compared to physical checks or cash handling.
- Offloaded Fraud Risk: Card networks and issuers invest massive compute into fraud detection, encryption, and tokenization. Depending on the network rules and authentication methods (like EMV), liability for fraudulent transactions often shifts away from the merchant.
- Increased Throughput: Decades of data show that consumers spend significantly more when using digital credit instruments versus cash. The reduction in user friction often drives higher average transaction values, offsetting the MDR cost.
6. The Illusion Revealed
The brilliance of modern card networks isn’t that they move money instantly. It’s that they abstract a massively complex, multi-day asynchronous batch reconciliation process into a two-second synchronous user experience.
Behind that tap are independent financial institutions, real-time fraud compute, global message routing, cryptographic validation, and bulk netting operations.
A merchant in Tokyo and a small bank in Ohio—two institutions with no direct relationship—are able to transact because they both agree to participate in a shared, rule-based routing topology. That two-second tap isn’t a payment; it’s the visible tip of a vast, asynchronous financial infrastructure.