Japan and South Korea look similar on the revenue slide—high card penetration, famous wallets, affluent buyers. Payments tell a different story. If you treat both markets as “just add cards and wallets,” you’ll end up with stalled conversions, slow refunds, and reconciliation chores nobody volunteers for. The teams that win take a quieter path: domestic bank rails first, virtual accounts as the backbone, konbini and easy-pay where they make sense, and a treasury layer that keeps FX away from customers and squarely inside finance policy.
This is the playbook we used to move an EU marketplace from “EU MID + cross-border everything” to clean JPY/KRW collections, predictable refunds, and month-end that closes itself.
What was really broken (and why spreadsheets kept multiplying)
The initial stack leaned on a single EU acquirer and a global wallet. Authorizations passed on small carts and then crumpled on mid-ticket orders, especially when issuers saw a foreign MID and unfamiliar descriptors. Bank transfers came in as free-text “furikomi” in Japan and as generic interbank credits in Korea; references were inconsistent, so finance matched by amount and timing—never fun. Refunds were worse: cross-border card credits took days; cash-based payments needed manual intervention; customers opened disputes because “tomorrow” meant nothing without a value date.
We didn’t need more methods. We needed rails that buyers already trust—and a ledger that could understand them.
The north star in one line
Let Japanese and Korean buyers pay like locals—virtual accounts for bank transfers, konbini in Japan, easy-pay in Korea—while refunds land on the same rail with a date support can say out loud. Treasury aggregates JPY/KRW and sweeps to EUR on rules; buyers never feel FX.
Virtual accounts: the unsung hero on both sides of the Sea of Japan
The single most important change was issuing payer-bound virtual accounts. In Japan, customers already expect furikomi (bank transfer) with a precise reference; in Korea, virtual accounts (가상계좌) are a standard e-commerce pattern. We gave every order—or every customer, depending on business rules—its own account number. When a credit arrived, a webhook closed the invoice without a human opening a CSV. The remittance comment could be blank; the account number was the truth.
Two things made this work day one. First, the beneficiary name on the payment screen matched what banking apps display, down to character set and spacing. Second, each payment object had a time-to-live. If a late transfer landed after TTL, the ledger parked it in exceptions with a scripted path (apply to open balance, return to source, or request confirmation). That single decision took auto-reconciliation above 95% and cut “missing payment” tickets to a trickle.
Konbini in Japan: powerful when you respect its shape
Konbini is not a generic “cash at store.” It’s a barcode or numeric slip with an expiry, a fee expectation, and a receipt that buyers treat as proof of purchase. We kept it for use cases where it truly lifts conversion: teen and young-adult segments, late-night carts, and buyers who distrust online card entry. We avoided pushing konbini for high-value orders or subscriptions. Refunds honored the rail: either reversal via the provider’s bank payout back to the payer’s account or, where supported, cash-out paths communicated at checkout. Most importantly, the expiry clock was visible and short; when it ran out, we issued a fresh slip rather than leave a zombie order.
Customers stopped asking “did you get my payment?” because the merchant portal mirrored what the buyer saw: created → slip issued → paid (store/time) → confirmed. Finance saw the same events and never had to reconcile “mystery cash.”
Korea’s “easy-pay” and when to use it
South Korea’s easy-pay ecosystem (card-backed and account-backed wallets) is ubiquitous on mobile. We enabled one card-backed option and one account-backed option per buyer profile, not a parade of logos. The rule of thumb was simple: if an easy-pay method refunded back to the same instrument quickly and preserved descriptors buyers recognized, it made the homepage; if refunds were detoured or one-way, it didn’t. For B2B carts and higher AOV consumer orders, we put virtual accounts first and easy-pay behind a “more ways to pay” link. Being selective improved approvals and tamed support workload.
Cards are not the enemy, but they’re not the workhorse
We kept domestic acquiring for cards in both countries, tokenized credentials for returning buyers, and risk-scored 3-D Secure where rules demanded it. The point was not to abandon cards; it was to route to them intentionally: first purchase, guest checkout, or when a buyer explicitly preferred them. A domestic descriptor that included our brand and a city slug—JP or KR—did more for chargeback prevention than a dozen risk rules. Refunds went back through the same acquirer; settlement files carried the same routing IDs our ledger used for bank rails, so reconciliation stayed one language across all methods.
Refunds with clocks, not poetry
We published promises and then hit them. Bank-rail collections (virtual accounts) refunded as domestic bank credits with next business day value dates during banking windows. Konbini refunds followed the provider’s bank payout path and landed on the same SLA; the help center showed examples of receipts and timelines so expectations were real. Card refunds rode the original authorization and used refund advice where available, so buyers saw “credit pending” quickly in online banking. Support agents never wrote “we’re working on it.” They wrote a date and a time in local time—and we shipped on that.
Under the hood, the choreography was always the same: lock entitlement or inventory → post a credit memo tied to the original tax context → fire the outbound payment on the same rail and currency → store the confirmation ID with a value date calculator that knew weekends and holidays in Tokyo and Seoul.
Tax receipts without late-night heroics
We didn’t reinvent local tax law in the app. We made the invoice consistent. Japanese buyers received receipts with the consumption-tax fielding their AP teams expect; Korean buyers saw VAT treatment they could book without questions. For B2B customers, business numbers were captured and validated; for B2C, we avoided pretending we had information we didn’t. Credit notes mirrored the original document—same tax country, same rate—so quarterly returns never required manual edits. It wasn’t about being clever; it was about never surprising the accountant on the other side.
Treasury and FX: sweep by rule, not by instinct
Collections lived in JPY and KRW. We held a working buffer—about a week of forecasted outflows for refunds and local opex—and swept the rest to EUR on time/threshold rules. Every conversion logged achieved rate vs a neutral benchmark. There was no per-transaction FX, no mid-flow currency flips, no DCC. Buyers paid and were refunded in their currency; finance optimized in the background. When seasonality stretched holding periods, we added light hedges; otherwise, policy beat improvisation.
Risk and the “boring controls” that save real money
We enforced a name/account match on the first payout or refund to a new destination and a cooling-off period on changes, verified out-of-band. Velocity caps for brand-new accounts throttled rapid pay-in/pay-out loops common in abuse patterns. Device/session binding prevented someone from rerouting funds after logging in from a new fingerprint. None of this appeared in marketing slides; all of it removed the most expensive afternoons.
What changed—measured like a CFO would, not a press release
Approval rates climbed when mid-ticket orders stopped looking foreign and when easy-pay was used where it belongs instead of everywhere. More importantly, DSO turned into a schedule. Konbini and virtual account credits posted the same day or the next morning; refunds hit on the promised date; the support queue stopped filling with “where is my money?” Auto-reconciliation crossed 95% once virtual accounts and structured references were standard; the remainder were true exceptions, not mysteries. On the P&L, effective cost per successful payment dropped by triple-digit basis points when volumes moved off cross-border cards to domestic rails. Treasury stopped arguing about spread because the dashboard made it a number.
How we shipped without boiling the ocean
We staged the rollout by habits, not by logos. First, we turned on virtual accounts in both countries, wired the webhooks, and banned free-text references from our process. Second, we added one konbini provider in Japan and one account-backed easy-pay in Korea, both chosen for refund symmetry. Third, we localized card acquiring and descriptors, then taught support to quote value dates with a calculator that knew local cut-offs. Only after this was boring did we add a second konbini chain or a second easy-pay brand. The point wasn’t coverage; the point was predictability.
Red flags worth a polite “no, thanks”
If a wallet can’t refund to source cleanly, it doesn’t belong on the first screen. If a provider can’t guarantee stable virtual account issuance or loses references at settlement, keep looking. If someone proposes per-transaction FX “because it’s simpler,” remember who pays for every extra reconciliation edge. And if anyone suggests skipping the beneficiary-name match “just once,” expect to learn this lesson the hard way.