TL;DR: Fiat on ramp conversion is only honest when the denominator is every funding intent and the numerator is a usable product balance, not an accepted payment. Track one funding path, split rejection from timeout and abandonment, and count confirmed end states.
Your dashboard says the on-ramp converts at 60%. Finance says funded accounts grew by a third of that. Both are reading "conversion" and dividing by different numbers — and neither knows which users ended with spendable balance.
This guide is for growth, customer operations, and commercial leads at Web3 products and infrastructure businesses that embed a fiat-to-crypto on-ramp. It gives you a status dictionary, a denominator rule per funnel stage, and a worked synthetic calculation you can reproduce against your own events. It is conversion analytics only — nothing here is advice on buying crypto.
Why the reported number looks better than it is
Every system in the chain reports "conversion" against its own favorite base. The widget reports quote-to-payment, the provider reports payment-to-delivery, and your ad platform reports click-to-signup. None of them answer the question a growth team actually owns: of the users who showed funding intent, how many ended with usable balance?
The gap is structural, not cosmetic. The 2023 Global Crypto Onramp Report found that more than 50% of fiat-to-crypto transactions failed even after users completed KYC, with abandonment during the purchase flow reaching as high as 90% in the flows studied. Success rates also varied sharply by user location, fiat currency, cryptocurrency, payment method, amount, and provider — local payment methods and bank transfers generally outperformed cards, while Europe posted the highest average success rates in the study.
Providers see the same leak from their side. Mercuryo reported its conversion rate rising from 12% in H1 2025 to 21% in H1 2026 after streamlining its purchase flow, and noted that abandonment concentrates before payment submission. Treat these as vendor-reported direction, not your benchmark.
The measurement discipline is identical to calculating business process automation ROI: define the unit, define the end state, then count.
Track one selected funding path end to end
Pick one rail — for example, card purchase through one provider in one market — and map every event from funding intent to usable balance before you average anything. A blended funnel across cards, bank transfers, and wallets hides exactly where users die.
A minimal event map for a card-rail funnel:
| Stage | Event to log | Authoritative source | Typical loss |
|---|---|---|---|
| Funding intent | Quote accepted / order created | Your app and provider order record | Entry point, not a loss |
| Identity passed | KYC approved | KYC vendor or provider | Document and liveness friction |
| Payment submitted | User confirmed payment | Provider status update | Pre-submission abandonment |
| Payment authorized | Issuer or provider approved | Provider webhook | Card declines, fraud blocks |
| In delivery | Processing or pending delivery | Provider webhook | Liquidity and compliance holds |
| Usable balance | Balance spendable in product | Your ledger or chain confirmation | Delivery lag, reversals |
Two rules keep this honest:
- One path per view. A crypto on ramp that accepts cards, Apple Pay, and SEPA transfers is three funnels with different failure shapes, not one. Report per rail, per market, per provider — the 2023 report found success rates diverging enough between fiat on ramp providers that any blended average is fiction.
- Log the intent, not the screen. Widget opens and quote views are browsing. The denominator starts where the user commits — quote accepted or order created — because earlier steps belong to product marketing, not funnel integrity.
KYC deserves its own stage, not a footnote. Plaid notes that identity checks are often the highest-friction step in a crypto on-ramp, and some apps run KYC twice — once at signup and again inside the on-ramp, which compounds drop-off; one vendor analysis estimates stablecoin on-ramps lose 30-50% of users at the KYC step. If KYC is your biggest leak, remediation is a separate job — there is a playbook for reducing KYC abandonment with status-aware messages.
Separate rejection, timeout and abandonment
A failed intent has three different causes with three different owners. If they share one "failed" bucket, nobody owns anything.
The rule is simple: classify by who ended the attempt.
| Provider status (Transak example) | Bucket | Who ended it | First fix to try |
|---|---|---|---|
FAILED — card decline, KYC rejection |
Rejection | Provider, issuer, or KYC vendor | Payment-method coverage, provider mix |
EXPIRED — payment not made in timeframe |
Timeout | The clock | Deadlines, local rails, reminders |
CANCELLED |
Abandonment | The user, mid-flow | UX, trust proof at the decision point |
AWAITING_PAYMENT_FROM_USER (stuck) |
Abandonment in progress | The user, so far | Watch until the window closes |
Transak publishes this order status reference with REST, webhook, and WebSocket tracking. Your crypto on ramp provider will use different labels, but the split holds: a FAILED is a rejection verdict, an EXPIRED is a timeout, and a CANCELLED — or a user who never reaches submission — is abandonment.
Three traps to avoid:
- Abandonment before submission is invisible to the provider. The 2023 report's failure counts cover submitted transactions. Users who close the widget before paying exist only in your own events — instrument them client-side.
- Webhook duplicates fake volume. MoonPay's widget webhooks can arrive duplicated and out of order, and Ramp Network retries failed deliveries four times at 3-minute delays. Dedupe on order ID and apply statuses by timestamp, not arrival order.
- Timeouts are not rejections. A bank transfer that expires in 48 hours is a design problem — deadline too short, rail too slow — not a card problem. Bucket it separately or it pollutes your decline metrics.
The same status discipline applies downstream: support teams need their own wording for each provider state, which is why a stablecoin payment status map for support teams exists as a separate article.
Measure the confirmed end state
The honest numerator is not "payment accepted." It is usable product balance — the moment the user can actually do the thing the funding was for. Stripe's fiat onramp guide lands on the same definition: delivery completes when the user sees their balance update and the business sees a complete conversion.
Three decisions make this number trustworthy:
- Wait for the terminal state. Provider delivery is asynchronous — an order sits in
PROCESSINGorPENDING_DELIVERYbeforeCOMPLETED. Counting the payment response as conversion overcounts every order still in flight. - Pick a conversion window per rail. Funnel tools already work this way: Amplitude counts a conversion only when all steps finish in order inside a rolling window that defaults to one day. Cards settle in minutes to days; bank transfers take longer. Publish the window next to the rate, and never mix windows in one trend line.
- Report mature cohorts only. A funding intent from yesterday has not finished expiring, retrying, and delivering. Measuring it today understates conversion — the same reason deduplicated payment rates dip before retries finish, covered in payment success rate beyond authorization.
Once the balance is usable, a different funnel starts — activation. Measuring what users do after their first funded session is the job of Web3 activation metrics beyond the wallet connection, not of this rate.
A worked example with synthetic numbers
Synthetic example — invented inputs, reproducible arithmetic, no client data. Assume one week of card-rail funding intents:
| Funnel state | Count |
|---|---|
| Funding intents (quote accepted) | 1,000 |
| Payment submitted | 640 |
Rejected (FAILED: declines, KYC) |
90 |
Timed out (EXPIRED) |
120 |
Abandoned (CANCELLED or closed pre-submission of the submitted set) |
30 |
Provider COMPLETED |
400 |
| Usable balance inside the 7-day window | 380 |
The same week produces five different "conversion rates," all technically true:
- Provider-facing delivery rate — completed ÷ submitted: 400 / 640 = 62.5%
- Intent-to-completion — completed ÷ intents: 400 / 1,000 = 40.0%
- Honest funded conversion — usable balance ÷ intents: 380 / 1,000 = 38.0%
- Pre-submission abandonment — never-submitted ÷ intents: 360 / 1,000 = 36.0% of demand lost before the provider could even fail it
- Rejection share of submitted — 90 / 640 = 14.1%, the number your fiat on ramp payment gateway dashboard probably shows
If leadership quotes 62.5%, the report is silently dividing by payment submitters and discarding the 36% who left before paying — the largest single leak in this funnel. None of the five numbers is wrong; quoting one of them as "the" conversion rate is.
What honest denominators buy: one agency case study describes a crypto exchange where roughly 70% of verified users dropped between KYC and first deposit — invisible because deposits lived only in the backend and were never reconciled against acquisition reporting. After instrumenting the full path, deposit conversion moved from 31% to 58%, cost per funded trader fell 44% in a quarter, and a geo holdout exposed a 2.6x overstatement in platform-reported CAC. If your spend decisions run on platform-reported conversions, that last number is the one to sanity-check with a ROAS leak calculator before the next budget review.
The on-ramp funnel worksheet
Copy this spec into your tracking plan. It satisfies three requirements: track the selected funding path, separate rejection from timeout and abandonment, and measure the confirmed end state.
| Field | Rule |
|---|---|
| Funding path | One rail × one provider × one market per funnel view |
| Denominator | Funding intents: quote accepted or order created, deduplicated by user and order |
| Rejection | Provider or issuer verdict ending the attempt (FAILED, KYC denied) |
| Timeout | Clock ended the attempt (EXPIRED, stale awaiting-payment) |
| Abandonment | User ended the attempt (CANCELLED, or no submission inside the window) |
| Numerator | Usable product balance confirmed in your ledger, not provider COMPLETED alone |
| Conversion window | Stated per rail; report mature cohorts only |
| Retry rule | One intent, one outcome; dedupe webhook events by order ID |
Checklist before you publish the number internally:
- Can every intent be traced to exactly one end state, or do retries mint extra attempts?
- Does the report name its denominator — intents, submissions, or orders?
- Are rejection, timeout, and abandonment three separate rows?
- Is the window stated, and is the newest cohort marked immature?
FAQ
What is fiat on ramp?
A fiat on ramp is the service or flow that converts government-issued money — dollars, euros — into cryptocurrency inside a product, usually via card, bank transfer, or a local payment method. This article measures that funnel; it does not recommend where or whether to buy crypto.
What is a good fiat on-ramp conversion rate?
There is no honest universal number: the 2023 report found success varies by market, currency, method, amount, and provider, and vendor-reported figures like Mercuryo's 12% to 21% reflect their own denominator choices. Treat published rates as orientation, build a baseline from your own intent-to-balance cohorts, and improve from there.
Which provider statuses count as a timeout versus a rejection?
In Transak's model, EXPIRED means the user did not pay within the timeframe — a timeout — while FAILED covers verdicts like card declines — a rejection. CANCELLED and never-submitted intents are abandonment. Whatever your provider's labels, classify by who ended the attempt.
Fiat on ramp vs off ramp — what is the difference?
An on ramp converts fiat into crypto so a user can fund a product; an off ramp converts crypto back into fiat for withdrawal. The measurement discipline is the same in both directions — intents in, confirmed end states out — but the failure buckets and rails differ.
Should the denominator count widget opens or funding intents?
Funding intents. Widget opens and quote views measure browsing, not commitment. Starting the denominator at quote-accepted or order-created keeps the rate tied to users who actually tried to fund.
How long should the conversion window be?
Long enough for the slowest rail you count: card orders resolve in minutes to days, bank transfers longer. State the window next to every rate — one day is a common analytics default, but it will understate bank-rail funnels.
Answer clarity notes
- Dates: provider documentation and the Mercuryo release were checked on 2026-11-19; the Global Crypto Onramp Report is a 2023 source and its figures describe the flows studied then, not current rates. Verify current vendor statuses and definitions before rebuilding a report.
- Scope: written for growth, customer operations, and commercial leads at Web3 product and infrastructure businesses measuring an embedded fiat-to-crypto on-ramp. It is analytics guidance only — not investment, compliance, or purchase advice.
- Evidence: provider statuses, webhook behavior, and funnel-window mechanics come from linked primary documentation. The 50%+ failure and up-to-90% abandonment figures, the Mercuryo conversion numbers, and the 30-50% KYC loss estimate are vendor- or report-published claims cited with their own scope. The Wameq case is an agency-published client study. The worked example is explicitly synthetic.
- Do not infer: all percentages here are planning guidance and worked illustrations, not guarantees; your rates depend on rails, markets, provider mix, and denominator choices.
Sources
- Global Crypto Onramp Report (2023)
- Mercuryo — crypto purchase conversion press release
- Transak — Track order status
- Ramp Network — Webhooks
- FluidRWA — Ramp onboarding evidence map
- Amplitude — Interpret your funnel analysis
- Stripe — Fiat onramp guide
- Plaid — How to build a crypto onramp
- Vove ID — Why on-ramps lose users at KYC
- Wameq — Crypto exchange funded-trader growth case study
Most on-ramp conversion problems are event-plumbing problems before they are UX problems. If you want a second pair of eyes on how funding events flow into your reports, That'sGonnaHelp runs a bounded automation audit of workflows like this.