Appendices
Appendix C — Comparison Matrices
Side-by-side comparisons from across the book. Rows are terse by design — the chapter pointer has the nuance.
Network vs Gateway vs PSP vs Orchestrator
Who you contract with, who touches the money, who owns the risk. See Chapter 5.
| Dimension | Card network | Gateway | PSP | Orchestrator |
|---|---|---|---|---|
| What it is | The rails + rulebook | Technical message courier | Full-stack acceptance provider | Router above multiple PSPs |
| Contracts with merchant | No | Sometimes | Yes | Yes (software contract) |
| Holds merchant funds | No | No | Yes (merchant of record models vary) | No |
| Owns fraud/credit risk | Sets rules, arbitrates | No | Yes, underwrites you | No |
| Examples | Visa, Mastercard | Authorize.net-style | Stripe, Adyen, Rapyd | Spreedly-style, in-house |
Single-Message vs Dual-Message
One message or two — the split that separates PIN debit from everything else. See Chapter 10.
| Dimension | Single-message | Dual-message |
|---|---|---|
| Auth and clearing | One combined message | Separate auth, then clearing |
| Typical use | PIN debit, ATM | Credit, signature debit, e-commerce |
| Capture later? | No — immediate | Yes — tip adjust, delayed capture |
| Reversals | Harder, money moved | Void before capture is cheap |
Network Tokens vs PSP Tokens vs Vault Tokens
Three tokens that all "replace the PAN" and behave nothing alike. See Chapter 12.
| Dimension | Network token (DPAN) | PSP token | Vault token |
|---|---|---|---|
| Issued by | Card network | Your PSP | Your vault (own or third-party) |
| Survives card reissue | Yes — auto-updated | Only via account updater | Only via account updater |
| Portable across PSPs | Yes | No — lock-in | Yes — you control export |
| PCI scope for you | Low | Low | Vault provider dependent |
Billing Model vs Payment Method Suitability
Match how you charge to how the money moves. See Chapter 17.
| Billing model | Cards | Direct debit | Bank transfer / VA | Wallets |
|---|---|---|---|---|
| Fixed subscription | Strong (with dunning) | Strong, cheap, revocable | Weak — manual | Market dependent |
| Usage-based / metered | Strong — MIT on file | Good for large B2B sums | Good with VAs for B2B | Weak |
| One-off invoice (B2B) | Costly at size | Possible | Strong — VA reconciliation | Weak |
| Marketplace payouts | N/A (push needed) | N/A | Strong | Regional |
Push vs Pull Rails
Who initiates decides who bears the failure modes. See Chapters 18 and 20.
| Dimension | Push (credit transfer, RTP) | Pull (cards, direct debit) |
|---|---|---|
| Initiated by | Payer | Payee (with authorization/mandate) |
| Insufficient funds risk | None — fails upfront | Returns/declines after the fact |
| Reversibility | Low — often final | High — chargebacks, DD refunds |
| Merchant fear | Customer forgets to pay | Money clawed back later |
| Reconciliation | Needs references/VAs | Built into scheme reporting |
Cards vs RTP vs Stablecoins vs Crypto
Loading interactive…
Interactive — “Comparison explorer” is available in the web edition of this book.
The cross-rail scorecard. See Chapter 29.
| Dimension | Cards | RTP / instant transfers | Stablecoins | Native crypto |
|---|---|---|---|---|
| Speed to final | Days (T+1/T+2) | Seconds–minutes | Seconds–minutes | Minutes–hours |
| Reversible | Yes — disputes | Mostly no | No (issuer freeze aside) | No |
| Consumer protection | Strong, scheme rules | Thin | Contractual only | None |
| Reach | Global brand acceptance | Domestic schemes | Wallet + ramp dependent | Niche |
| Cost profile | Percentage + interchange | Flat, low | Near-zero rail, ramps cost | Volatile fees |
Closed-Loop vs Open-Loop Agentic Architecture
The cleavage that actually predicts agentic-payment traction. See Chapter 33.
| Dimension | Closed loop (e.g. Alipay) | Open loop (Western stack) |
|---|---|---|
| Who owns model, catalog, rail | One entity, one trust domain | Different parties per layer |
| Credential handoff | None needed | Tokenized across trust boundaries |
| Risk visibility | Full transaction graph | Fragmented per participant |
| Liability | One responsible party | Contested across the chain |
| Integration cost | Zero incremental | Adapter glue at every seam |