A PM's Deep Dive:
Cross-Border Payment Infrastructure
From Echo Japan's async payment challenge to understanding the global rails that move money: FX mechanics, local payment methods, and what it means for product decisions.
Type
Personal Research
Role
Product Manager
Scope
Payments · Infrastructure · Cross-Border
00 — Intro
What started this
Working on Japan market entry at Echo, I had to design around Konbini and bank transfer flows - async payment methods that broke every assumption our sync order system was built on. We solved the product problem. But the experience left me curious about what was actually happening underneath.
Why does cross-border payment confirmation take days? Why does the same card work in Tokyo and Paris but the fees look completely different? Why did we need a local payment method at all?
This is my attempt to answer those questions, not as a textbook, but as a PM trying to understand the infrastructure well enough to make better product decisions.
01 — The Players
Who's in the room
Before getting into cross-border specifics, it helps to know who's involved in any card payment, domestic or international.
Cardholder: Pays with a card
Hover or tap any node to see its role
Example
When Sage pays $6 at Starbucks with her Chase Visa: the POS sends a tokenized payment request via Stripe → Visa routes it to Chase → Chase approves → confirmation travels back in seconds. No money has moved yet.
💡 PM Insight
PSP choice isn't just a technical decision. It determines which markets you can enter, how FX risk is handled, and whether local payment methods are even available. Stripe and Adyen serve fundamentally different customers, not just different sizes.
02 — Local Payment Methods
The payment method is the market
Living across Japan, Korea, and the US, and working on a Japan market entry, made this visceral. The same product needs completely different payment infrastructure in each country. And the infrastructure shapes the product design, not the other way around.
Konbini + PayPay
Async cash · QR walletKonbini: pay at 7-Eleven/FamilyMart with a payment slip; confirmation in hours to days. PayPay (60M+ users) is Japan's dominant QR wallet.
PM: Two separate integrations. Konbini requires full async state machine design. PayPay is instant but needs QR-based checkout flow, not a card form.
Kakao Pay · Naver Pay · 통신사 결제
Super-app wallet · Carrier billingKakao Pay: auth via KakaoTalk push (95%+ app penetration). Naver Pay: dominant for e-commerce. 통신사 소액결제: charge to monthly phone bill via SKT/KT/LGU+.
PM: Carrier billing settles monthly, a completely different reconciliation cycle. KakaoTalk Business verification takes weeks. Skipping Kakao Pay means missing the auth method embedded in the app nearly every Korean opens daily.
WeChat Pay · Alipay
Super-app ecosystemQR-code-first (not NFC). WeChat Pay embedded in 1.3B-user super-app. Alipay backed by Ant Group / Alibaba ecosystem.
PM: You're not integrating a payment method, you're integrating into an ecosystem. Separate merchant portals, mini-program frameworks, platform-managed dispute resolution. Cards are largely irrelevant for domestic consumers.
UPI
Instant bank transferVPA (Virtual Payment Address) instead of card number. Interoperable across all banks. Zero MDR mandated by RBI.
PM: If your revenue model depends on interchange, UPI breaks it: zero merchant fee is law. Checkout UX is QR or VPA entry, not a card form. A completely different product surface.
Pix
Instant bank transferCentral Bank mandated. Operates 24/7/365 including holidays. Settlement in seconds. Free for consumers.
PM: Pix QR codes have configurable expiry. Lower fraud risk than cards (bank auth required). Now dominant: card-first checkout underperforms significantly.
iDEAL
Direct bank transfer~70% of Dutch online payments. Bank redirect-based: user authenticates in their own bank's app.
PM: Card-only checkout will underperform significantly. The redirect UX requires careful handling, as users leave your checkout to their bank app and return.
SEPA Direct Debit
Bank transferCultural aversion to credit. SEPA mandate-based: customer authorizes recurring or one-time debit from bank account.
PM: Requires a signed SEPA mandate upfront, adding a step before first payment. T+1 to T+2 settlement. Chargeback window is 8 weeks for unauthorized transactions.
How payment type shapes product design
Local payment methods aren't interchangeable. Each type demands fundamentally different product architecture decisions, before engineering starts.
Async Confirmation
Konbini (Japan) · SEPA (Germany/EU)
Hours → Days- —Order state machine: Pending → Confirmed / Expired / Partial
- —Inventory hold decision: hold = stockout risk; no-hold = fulfillment risk
- —Customer comms at every state transition
- —Idempotent webhook handling (providers retry, duplicates happen)
Example: Echo Japan
No inventory hold, backed by LTV data showing repeat-purchase collision rate was within acceptable range. All 5 payment states mapped to finance reconciliation events before engineering started.
Ecosystem Wallet
Kakao Pay · Naver Pay (KR) · WeChat Pay · Alipay (CN)
Instant (in-app)- —SDK integration, not just API: checkout is inside their ecosystem
- —Auth via the platform's own flow (KakaoTalk push, WeChat payment password)
- —Platform-managed disputes, not standard chargebacks
- —Korea carrier billing: charges monthly phone bill; T+30 settlement, separate integration per telecom (SKT, KT, LGU+)
Example: Korea launch
KakaoTalk Business verification takes weeks, it's not a same-day API key. If you're planning a Korean launch, start the merchant account process months before target date. Miss it and you're excluding the auth method in the app 95% of Koreans open every day.
Instant Bank Transfer
UPI (India) · Pix (Brazil)
Seconds- —Different checkout surface: QR code or VPA/alias, not a card form
- —Different failure modes: bank server unavailable ≠ card declined
- —UPI: zero MDR mandated, interchange-dependent revenue models don't work
- —Pix QR codes expire, must handle regeneration in checkout
Example: India entry
For a fintech product with a revenue model built on interchange rebates, UPI is a structural problem. The pricing model has to be redesigned before launch, not patched after. This is a pre-roadmap decision, not a post-launch optimization.
PSP choice determines what's even possible
Stripe
- Strong card infrastructure
- Growing local method support
- Best for: startups, SMBs, developers
Limitation: local method coverage thinner outside core markets; Konbini, carrier billing, WeChat Pay often require separate integrations
Adyen
- 150+ local payment methods
- Direct Acquirer licenses in major markets
- Best for: enterprise, global-first products
Why Uber and Netflix use Adyen: one platform, every market: FX + local methods + acquiring in one stack
💡 PM Insight
PSP selection upstream determines whether local payment support is even possible. The Echo Japan decision to support Konbini and bank transfer wasn't a feature addition, it was a market entry requirement. Make the PSP decision based on where you're going, not just where you are.
03 — Cross-Border Layer
What changes when money crosses borders
Add a cross-border transaction and three things happen that don't exist in domestic payments: currency conversion, additional fees, and settlement timing risk. Each one has product implications.
FX Conversion
When Sage pays €100 at a Paris café with her Korean Shinhan Visa card, Visa converts EUR → KRW at that day's spot rate, then adds a spread.
| Domestic | Cross-Border | |
|---|---|---|
| FX conversion | None | Visa spot rate + 0.5–1% spread |
| Cross-border fee | None | Issuer adds 1–3% |
| Settlement currency | Same | May differ |
Why Visa's rate beats a bank branch: Visa processes hundreds of millions of transactions daily; banks offer wholesale rates that individuals can never access. Even with the spread, it's better than most retail FX options.
Spot vs. Guaranteed Rate: merchant's choice
Spot Rate
- Rate floats with market
- Lower fee
Best when: low volume or stable currency
Guaranteed Rate
- PSP locks rate via forward contracts
- Predictability, pays a premium
- €100 → $108 fixed vs. $105–$112 at spot
Best when: high volume, volatile pairs
Volume-based negotiation
Enterprise merchants (Uber, Netflix) negotiate FX margins directly with PSPs. Small merchants take standard pricing. PSP selection matters more as you scale globally.
Cross-Border Fee
This is the “foreign transaction fee” on your card statement. Charged by the Issuer, not Visa, for three reasons:
Fraud risk: Cross-border transactions are harder to verify. If fraud occurs, the Issuer absorbs it.
FX hedging cost: The Issuer converts currency and carries exchange rate exposure until settlement.
Infrastructure: Routing through global networks costs more than domestic rails.
DCC: The Trap
⚠ Watch Out: DCC
DCC (Dynamic Currency Conversion) happens when a terminal abroad asks: “Pay in KRW?” It sounds convenient. It isn't.
Normal flow: €100 → Visa spot rate → Issuer converts → billed in KRW
DCC flow: Terminal pre-converts at their rate → adds 2–3% fee → billed in KRW
The merchant and their Acquirer split the DCC fee. That's why they offer it.
Always choose local currency.
Settlement Currency Mismatch
The gap between payment capture and actual settlement (T+1 to T+2) creates FX exposure.
Example
€100 captured. EUR/USD = 1.10 → $110 expected
EUR/USD = 1.05 → $105 received
$5 lost to rate movement
Three ways to handle it
Absorb it: Rate floats. Fine at low volume, painful at scale.
PSP hedges it: PSP locks a forward rate at capture. Upside capped, downside protected.
Adyen's approach: 150+ currency pools. Issuer sends €, Adyen pays $ from its own pool. FX risk eliminated, not just managed.
💡 PM Insight
FX strategy directly affects unit economics. A product that works domestically can become margin-negative internationally if FX handling isn't designed into the payment architecture. This is a PM decision, not just a finance one.
04 — PM Lens
What this means for product decisions
Payment infrastructure decisions aren't made by engineers. They're made by PMs who understand what the infrastructure can and can't do, and what it costs to find out the hard way.
PSP selection has long-term lock-in risk
Switching PSPs mid-scale is expensive. FX contracts, local payment integrations, and reconciliation pipelines all need to be rebuilt. Choose based on where you're going, not just where you are.
FX is a unit economics variable
Most P&L models for international expansion underestimate FX drag. Spot rate variance, cross-border fees, and hedging premiums can swing margins by 2–4% per transaction. Model it before launch, not after.
Local payment methods determine market entry feasibility
In some markets, card-only checkout means the product literally doesn't work for most users. This is a go/no-go input, not a post-launch optimization.
Async payment flows require product architecture decisions upfront
Konbini and bank transfer taught me this directly. Inventory hold logic, order state communication, idempotency handling, and reconciliation mapping all need to be designed before engineering starts, not discovered during QA.
Compliance and licensing are moats
Adyen's strength isn't just technology, it's the years spent acquiring banking licenses and local payment network memberships in 150+ markets. PSP selection is partly a bet on whose compliance infrastructure you want to sit on.
“The checkout button is the surface. Everything covered here is what makes it work, or not, when your customer is 6,000 miles away.”