TL;DR: Compare currency paths only after splitting what reports blend: the funnel (attempts to settled payouts) and the economics (explicit fees vs the FX margin inside the executed rate). Then compare only equivalent corridor cohorts.
A payment funnel that ends in one currency and starts in another produces two numbers your dashboard happily merges: how many attempts became settled payouts, and how much value survived conversion. A blended conversion rate hides the second number, so cross border payment analytics built on a single rate will tell you a corridor is failing when it is merely expensive — or profitable when its conversion quietly leaks.
This is a worked analysis for growth, merchant operations, and revenue operations leads at fintech and payment businesses. It builds the cross border payment funnel step by step: definitions first, then a fee-versus-rate split, then cohort rules, then a worksheet with synthetic numbers you can reproduce. The discipline is the same one used to calculate automation ROI on defined units: fix the unit, fix the denominator, then count.
A currency path is the ordered pair funding currency → payout currency for one attempt: what the customer's account is charged in, and what lands in your settlement balance. USD→USD is a path with no conversion; EUR→USD routes every step of the funnel through an exchange rate. Each path gets its own funnel and its own cost accounting.
Define source and destination amounts
The comparison starts with two amounts per attempt, not one: the source amount in the funding currency and the destination amount booked in the settlement currency. Without both, plus the executed rate and its timestamp, you cannot say what any funnel step cost.
Your funnel has the same skeleton as any payment funnel — attempted, authorized, captured, settled — and the terminal-state discipline from measuring payment success beyond authorization applies unchanged: pick one end state (settled for finance-grade work), dedupe retries per purchase, and record when each state became final. Providers deliver outcomes asynchronously — Adyen's AUTHORISATION webhook carries a success flag and a refusalReason for failures, and asynchronous methods can sit in Pending for hours or days before a final Authorised, Refused, Cancelled, or Error code, per Adyen's webhook and result-code documentation.
What changes on a cross-border path is the field list. Every attempt record needs:
| Field | Purpose |
|---|---|
attempt_id, purchase_id |
Dedupe retries; one purchase can hold several attempts |
funding_currency |
What the customer pays in |
payout_currency |
What your settlement balance receives |
source_amount |
Amount authorized in the funding currency |
executed_rate, rate_timestamp |
The rate actually applied, and when |
fee_lines |
Explicit per-transaction fees, kept separate from the rate |
destination_amount |
Net amount booked in the settlement currency |
terminal_state, state_timestamp |
Settled / refused / cancelled, and when final |
corridor_id |
Stable key like EUR-USD-card for cohorting |
The denominator question answers itself once the fields exist: eligible funding attempts — checkout submissions that passed your own validation — per corridor_id, deduplicated to purchases for the customer-level rate. The front door matters too: track attempts per product view for each displayed currency, because currency unfamiliarity kills sessions before a payment attempt exists. A 2024 study of 3,010 consumers across six countries found 93% say local-currency pricing significantly impacts their purchase decision, and 35% would not return after hidden fees such as currency conversion charges, per Airwallex's cross-border checkout research. That drop-off happens upstream of every funnel metric you own.
Event capture on these paths is usually server-side — webhooks and settlement reports, not browser tags — so the plumbing described in server-side tracking for small business is the realistic transport, and the naming hygiene from GA4 key events and conversions keeps corridor_id consistent across tools.
Separate provider fees and rate effects
Two different costs ride on every converted transaction, and most reports show only one. An itemized cross border fee on the statement is the visible part; the FX margin lives inside the executed rate, and treating the blended rate as "the exchange rate" makes that second cost invisible.
The fee stack itself comes in layers. Adyen books per-transaction fees (interchange, scheme fees, markup, commission) at capture, processing fees as a daily total, and other charges through a monthly invoice, per its fee reconciliation documentation. The rate layer is murkier: when shopper and settlement currencies differ, Airwallex converts at the payment method provider's rate plus the platform's FX markup and "you see one blended FX rate in reports," per its FX reporting documentation. Dynamic currency conversion makes the same trade on the customer side — the checkout rate bundles conversion margin plus fees, and Stripe notes it typically costs more than issuer-side conversion. Mastercard's 2025 merchant guide shows the pair survives into scheme data: a DCC transaction records the original pre-conversion currency and amount in DE 54 subfields, per the DCC performance guide.
So compute the rate effect yourself. The working formula:
implied_fx_markup_bps = (reference_rate − executed_rate) / reference_rate × 10,000
fx_margin_value = source_amount × (reference_rate − executed_rate)
Pick one public reference and stick to it. The ECB publishes euro reference rates around 16:00 CET each working day, explicitly "for information purposes only" — which is exactly what a consistent benchmark needs to be, per the ECB's reference rates page. Corridor-level comparison works the same way in vendor tooling: capture quote ID, timestamp, expiry, executed rate, and net payout, then measure basis-point variance against a time-matched benchmark — the method described in this FX-markup field guide. Swift's analytics layer likewise reports which agent performed each conversion in the chain, per its Observer Analytics page.
Three boundaries keep this honest:
- Rebundling hides the split. Some providers can show identical economics as an adjusted rate or as a separate conversion fee. When the contract allows either, trust only executed rate vs reference rate plus itemized fee lines.
- Reconversions add noise. Refunds and disputes convert again at event-date rates, so a cross-border refund rarely returns the original settled amount; exclude reversed items from the funnel denominator and track them separately.
- The split is for diagnosis, not negotiation leverage. A measured 73 bps markup tells you what the path costs; whether your provider can do better is a procurement question, not an analytics one.
When the cost math feeds a build-vs-buy decision on routing or providers, the ROI calculator turns the per-transaction delta into a monthly figure.
Compare equivalent corridor cohorts
Once each path has a funnel and a cost split, the only legitimate comparison in cross border payments is cohort-versus-cohort: same currency pair, same method mix, same window, normalized to one reporting currency at the same-day reference rate.
The mechanics come straight from funnel analytics. Amplitude computes conversion as users completing an ordered event sequence inside a conversion window, applies segment filters at the first step, and wants n>30 with n·p≥5 and n·(1−p)≥5 per variant before calling a difference significant, per its funnel documentation. Treat that as a statistical floor, not an operating floor: for decisions, use roughly 200 eligible funding attempts per path per 4-week window as a working minimum, and rebuild cohorts when segment definitions change — funnel-derived cohorts go static, per Amplitude's funnel FAQ.
Four rules make cohorts comparable:
- Same currency pair path. EUR→USD is not comparable to USD→BRL; spread regimes and issuer behavior differ. Compare a path to its own baseline first, then to other paths only as ranking.
- Same method mix. Cards, wallets, and bank redirects refuse and settle differently; a corridor-level success rate split by rail and currency pair is the standard corridor analytics practice — a 94% aggregate can mask a 78% rate on one corridor.
- Same observation window. Settlement lags authorization, so compare windows where both paths had time to reach a terminal state.
- Normalize value at the reference rate. Convert both paths' amounts to your reporting currency at the same-day public reference. Converting at executed rates double-counts the FX margin you are trying to measure.
Reversed items — refunds, chargebacks — leave the denominator entirely and get their own cross-border report, since they convert twice. The settled side feeds downstream payment reconciliation across Stripe, QuickBooks, and your CRM; the funnel number is only as good as what reconciliation confirms landed.
The currency-path funnel worksheet
Copy this worksheet per corridor_id per reporting window. Inputs first, formulas second, one worked column per path.
Inputs per attempt: the ten fields from the first table. Roll them up per window.
Formulas:
funnel_conversion(path) = settled_payouts / eligible_funding_attempts
implied_markup_bps = (reference_rate − executed_rate) / reference_rate × 10,000
cost_share(path) = (explicit_fees + fx_margin_value) / reference_value
Worked example — synthetic inputs, reproducible arithmetic, no client data. One 4-week window, one product, two paths:
| Step | Path A: USD→USD | Path B: EUR→USD |
|---|---|---|
| Eligible funding attempts | 800 | 600 |
| Authorized | 720 (90.0%) | 510 (85.0%) |
| Captured | 704 | 498 |
| Settled | 700 | 490 |
| Funnel conversion | 87.5% | 81.7% |
Path B converts 5.8 points worse on count. Now the economics, per average transaction — Path A ticket $95.00, Path B ticket €92.00:
- Executed rate on Path B: 1.0840 USD/EUR → €92.00 becomes $99.73 gross.
- Same-day ECB reference: 1.0920 → reference value $100.46.
- Implied markup: (1.0920 − 1.0840) / 1.0920 = 0.73%, or 73 bps → $0.74 of rate cost per transaction.
- Explicit fees on both paths (synthetic schedule): 2.9% + $0.30 → Path A pays $3.06, Path B pays $3.19 on its converted $99.73.
- Net settled: Path A $91.95 of $95.00; Path B $96.54 of a $100.46 reference value.
- Cost share: Path A = 3.06 / 95.00 = 3.2%; Path B = (3.19 + 0.74) / 100.46 = 3.9%.
Read together: Path B loses 5.8 points of conversion and carries a 0.7-point heavier cost share — and the whole cost difference is the embedded FX margin, invisible in the fee lines. If you had compared only blended "revenue per attempt," part of that margin would have been misread as a conversion problem or a pricing problem.
Attribution then runs in three directions, which is the point of the split. The authorization gap (85.0% vs 90.0%) is issuer-side; the missing attempts before the funnel are customer-side, where currency display is a documented factor — Shopify reports 92.2% of customers prefer shopping in their own currency and about a third abandon USD-only checkouts, via FXC Intelligence's checkout benchmarking; and the 73 bps is pure economics, fixable in procurement rather than in the funnel. Three causes, three owners, one worksheet.
FAQ
What is cross border conversion?
Two different things share the phrase: a customer's conversion through your funnel on a multi-currency path, and the currency conversion executed inside the payment itself. Keep both meanings in your vocabulary and say which one you mean — the funnel conversion rate and the FX conversion economics are separate numbers by design.
Can I use my provider's blended FX rate for reporting?
For settlement totals, yes — it is what actually booked. For comparison, no: the blended rate embeds the markup, so two paths priced by different providers look like they faced different "markets" when they faced different margins. Normalize with a public reference rate instead.
Do currency paths matter for payouts as well as pay-ins?
More, if anything. Payout funnels add quote expiry and beneficiary-side deductions on top of the same fee-plus-rate structure, so the worksheet's executed-versus-reference check applies unchanged — just point it at disbursement records.
Should abandoned carts in foreign currency count as funnel failures?
Count them as pre-funnel drop-off, not attempt failures. Track product views → checkout start per displayed currency next to the funnel itself; the Airwallex and Shopify figures above are evidence that step has its own loss rate.
Does this require new tooling?
Usually not. Webhook payloads already carry outcome states, scheme data carries the pre-conversion pair, and providers expose fee reports. The work is joining them on a stable corridor_id — a data-definition job more than a purchase.
Answer clarity notes
- Dates: external sources were checked on 2026-11-17; the ECB rate examples use the publication context of that page, and vendor documentation is living — verify current formulas and fee schedules before rebuilding a report.
- Scope: written for growth, merchant ops, and RevOps leads at fintech and payment businesses measuring operational conversion. It covers analytics and attribution only — no FX forecasts, hedging, treasury, or financial advice.
- Evidence: the two-path example uses synthetic inputs with reproducible arithmetic — it is not a measured client result or a benchmark. Consumer statistics are third-party study findings cited with their original scope (Airwallex 2024, 3,010 consumers, six countries; Shopify figures via FXC Intelligence). The ~200-attempt cohort floor is an operator recommendation, not a statistical standard.
- Do not infer: basis-point gaps, cost shares, and conversion deltas are planning guidance, not guarantees; corridor mixes, providers, and rate regimes change the numbers.
Sources
- Adyen — Webhook structure and types
- Adyen — Payment result codes
- Adyen — Reconcile fees
- Airwallex — FX fees in payments
- Airwallex — Optimised cross-border checkout experience 2024
- Amplitude — How funnels compute conversions
- Amplitude — Funnel analysis FAQ
- ECB — Euro foreign exchange reference rates
- FXC Intelligence — Benchmarking cross-border checkout experiences
- Stripe — Dynamic currency conversion
- Mastercard — DCC merchant guide 2025
- Gruv — FX markup and exchange rate spread on payment platforms
- Fintel Analytics — Cross-border payments analytics
- Swift — Observer Analytics
If joining webhooks, settlement reports, and a reference rate into one clean funnel sounds like plumbing you would rather hand off, That'sGonnaHelp runs a bounded automation audit of exactly this kind of workflow.