You'll spend a fortune lowering acquisition cost and then lose the sale at the last step, the one analytics flags as a flat "abandoned at payment" with no cause attached. In a mini-program that blind spot is bigger, because the checkout runs through a host you don't own. The session, the payment sheet, the identity — all of it passes through someone else's runtime, and when it breaks, your dashboard just shrugs.

The leak nobody instruments

On the web you can watch the checkout funnel step by step. In a mini-program the payment sheet is handed to the host, and if it triggers an auth the user didn't expect, or a redirect that drops the session, or a currency that doesn't match what they saw, the user leaves — and your events show a clean "checkout started, no completion." No error, no stack trace, just a number that's quietly bleeding revenue.

The honest version: most mini-program checkout leaks aren't bugs. They're friction you didn't know you added because the steps live in a runtime you can't fully see.

Why mini-program CRO isn't web CRO

The biggest lever on the web is a faster, simpler form. In a mini-program the biggest lever is not having your own form at all. The host already has the user's identity and a payment rail that's one tap from settled. If you bolt your own web checkout on top — a webview, a redirect, a fresh login — you've rebuilt the exact friction the mini-program was supposed to remove.

A runtime that reuses the host's native identity and payment collapses the funnel. No re-login, no card re-entry, no bounce to a browser. One call, one confirmation, done.

What actually moves conversion

  • Reuse the host identity. If the user is already signed in to the host, your mini-program shouldn't ask them to make a new account. Every extra sign-up step is a withdrawal from the conversion account.
  • One native payment call. Don't open a webview to collect money. Use the host's payment sheet. It's already trusted and already has the card.
  • Show local price before the tap. A currency surprise at payment is one of the fastest ways to lose a sale you'd already won.
  • Prefetch the address. Asking for a shipping address after the user committed is a classic late leak. Pull it from what the host already holds.
StepYour own web checkoutRuntime-optimized
IdentityNew accountHost identity reused
PaymentWebview + card entryNative host sheet
AddressTyped after commitPrefetched
SurprisesCurrency at the endShown up front
Steps to paid5–71–2

The tradeoff you're making

Native payment means the host takes a cut of the transaction. You're trading margin for completion, and on a mini-program where the alternative is a leaked sale, that trade is usually positive. Pretending you can keep 100% and also convert is how teams ship a checkout that looks cheap and converts worse.

The real move

Stop measuring only acquisition. Pull the add-to-cart-to-paid rate on real devices, in the host's own payment flow, and watch where the session actually dies. The leak is there, it's expensive, and it's almost never the thing your dashboard is blaming. Fix the steps the runtime can collapse, and the conversion you've been buying with ad spend finally shows up in revenue.

Sources