Most users abandon at the wallet step. The fix is segmentation: WalletConnect-first for crypto-native arrivals, embedded-wallet-first for everyone else, with an upgrade path to self-custody that does not strand early balances.
What moves the needle
- SIWE sessions that survive refreshes without re-signing every visit.
- Simulation previews showing gas, receive amounts and failure reasons before signing.
- Signing copy that says what is approved in plain words, beside the technical detail.
- Device-lab QA across extensions, mobile deep links, in-app browsers and hardware wallets.
Measure connect-started to first-confirmed-transaction weekly. That single funnel is the wallet report card.
Related: dApp architecture · Multi-chain strategy · Wallet services