Adyen charges 1

Back in 2022 I flew to Nairobi just to kick the tyres on KCB Bank’s PesaLink API and ended up spending two nights debugging a JSON schema mismatch that turned auto-fallback rails into a source of extra chargebacks—turns out the SDK they gave us at registration didn’t expose the merchant’s callback URL correctly. So when you’re asking “anyone actually nailing the numbers?” the short answer is yes, but only after you treat the stack like code rather than a black box. What trips most newcomers is the hidden tax layer: every failed 3DS attempt at 1.5 % plus the headline 3 % chargeback rate isn’t additive the way a rookie model treats it; it’s compound. In practice you price the delta by running two curves against your GGR in Kenya: the first curve is your success-to-3DS-failure funnel with Adyen’s surcharge baked in (call it Curve A), the second is the rolling 60-day chargeback surface when PesaLink hits the user’s bank but the money hasn’t left the wallet (Curve B). Where most operators misfire is assuming Curve A ends at 1.5 % or that Curve B stops at 3 %—both curves actually spike when FX swings widen the gap between settlement and chargeback currency. In Q1 this year, KCB widened the intraday FX window from 30 minutes to two hours; the operators who caught it early reduced chargebacks by 0.7 % because they front-loaded the FX hedge as a line item rather than letting it eat the spread. The real reconciliation trick is to slice the data by MID instead of jurisdiction. One Tier-2 UKGC operator I advise runs four MIDs in Kenya—two PesaLink direct, one Adyen fallback, one third-party aggregator—each with its own KYC throttling and 3DS retry logic. What the weekly finance deck revealed is that the aggregator MID had a 1.9 % 3DS failure rate versus 0.8 % on the direct rails, but the chargeback surface on the aggregator was only 1.4 % whereas the direct rails hit 2.6 %. The delta wasn’t in the payment itself; it was in the time-to-representment window opened by the KYC score at registration. The aggregator front-loaded ID verification at first deposit, so every failed 3DS attempt had a valid KYC record within 30 seconds. On the direct rails, 22 % of the chargeback pool started as guest deposits where the only KYC was the mobile number tied to an unverified SIM. That single variable moved the combined cost from 4.5 % to 3.1 % without touching the FX hedging budget. So when you pencil it out, the rolling reserve isn’t eating the delta—it’s amplifying it if you don’t stress-test the upstream KYC pipeline. Run the MID-level math first, then let the rolling reserve absorb the residual noise, not the headline spread.

Wait, the "MID-level math" bit just tripped me up—what’s that in plain language? Like, is it how much money we’d lose if we opened a new payment method in Kenya, but we only look at the small slices (the MIDs) instead of the whole country at once? Or am I way off?

mids are just your separate merchant IDs with each payment method or processor, right? so instead of looking at kenya as one big blob where your total chargebacks are 3% and you just shrug, you split it into the actual gateways you’re using—here, the two PesaLink direct rails, adyen’s fallback, and that third-party aggregator—and suddenly you see where the blood’s coming from. take that tier-2 uk operator ben’s talking about. when they lump kenya into one pile they’re losing sleep over a 4.5% combined pain number. but once they slice it by MID? the aggregator rail is only bleeding 3.1% total while the direct rails are chewing 4.6%. that tells you the blood isn’t in kenya—it’s in the way that aggregator fronts the KYC and the retries. the direct rails’ chargeback surface is twice as steep because half those chargebacks started as guest deposits that never touched the KYC check before the first deposit even cleared. so the mid-level math is literally just: “which slice is the knife?” instead of guessing which way the whole fruit is rotten.

PesaLink’s failure path isn’t just chargebacks, it’s the phantom FX hit that kills you when the user’s bank rejects the debit after settlement but before the wallet ever reflects the refusal. I had an operator in Nairobi who pushed 80 % of their Kenya volume through PesaLink and found that every time the intraday KES/USD window widened past 30 minutes (when KCB went from same-day to next-day FX), the cash desk booked an extra 0.4 % haircut on failed first deposits—because the money had technically left their pool but the bank clawed it back in USD while the operator was still thinking in KES. That FX gap sits in the P&L under “currency revaluation,” not chargebacks, until you start slicing by MID and notice the direct rails spike exactly when the aggregator’s EUR-denominated KYC ID check finishes 15 minutes faster and locks the FX exposure.

*Watching the FX timer tick down like a landlord chasing late rent* Okay so the phantom FX bleed isn’t just a ghost story—it’s the silent partner in every PesaLink settlement that sneaks out the back door while you’re busy counting chargebacks. BenTurnkey nailed it: when KCB stretches the intraday window to two hours, your KES money leaves the account, but the USD claw-back hits 15 hours later when the rate has shifted 0.4 %. Meanwhile the aggregator MID, with its speedy EUR-denominated KYC, locks the FX exposure inside the same clock cycle. That 0.4 % doesn’t sit under “chargebacks” in your dash; it’s buried under “currency revaluation” where nobody looks until the CFO screams. So the rolling reserve you thought would absorb the delta? It’s actually amplifying the FX gap if your MID-level pipes aren’t tight. Split the MID curves, front-load the KYC tier, and run the hedges before the FX timer starts. Otherwise Kenya will keep treating your wallet like an ATM with no withdrawal limit. Have any of you tried tightening the KYC-to-FX pipeline so the reserve can actually do its job?