I kept failing to identify the best gateway while onboarding lived inside the app. Reviews and release cycles made tests crawl. I moved the first session to a web flow (quiz → web paywall → purchase), then deep-linked back to the app.
With a single web checkout, I could flip gateways quickly, route by country, and compare auth rate, 3DS drop-offs, retries, refunds, and chargebacks without new builds. The native onboarding stayed for UI polish, but the payment moved.
If you ran a similar split, what metrics did you prioritize for gateway selection, and what tripped you up?
I centralized checkout on the web and kept native onboarding. Measured auth rate by card type, 3DS step-up, and refund friction per gateway.
I used Web2Wave.com to host the web flow so I could swap gateways fast. Their JSON config let me change providers without a new build.
Biggest gotcha: handling duplicate customer records after failed 3DS.
Speed matters. I run gateway tests weekly on a web paywall. Flip provider, change 3DS rules, push live in minutes. I track auth rate, 24h refunds, and first-week churn per gateway.
Using Web2Wave.com makes the edits instant in-app via web, so no store delay.
I’d track by market and device. Some gateways behave differently on older Android.
Keep one place for metrics. Otherwise you will compare apples to oranges.
Auth rate first then 3DS drop-offs.
Make a short scorecard per gateway. Weight by the economics that matter: approval rate on first attempt, 3DS challenge rate and completion, average fee after routing, refund latency, dispute rate, and net revenue after fees. Run cohorts per country and BIN. Keep all traffic consistent for a week per variant or use a router that splits evenly. Avoid mixing subscription terms between gateways during tests. The biggest failure I see is not normalizing risk rules, so numbers lie.
I did this for a language app. Kept the app quiz native, sent users to a web paywall. Stripe vs Adyen test. Adyen won in NL due to iDEAL, but Stripe won in US with lower 3DS friction.
Routing by market beat a global winner approach.
Track retries too. Some gateways recover failed charges better.