Avocado. Self-custody crypto that feels like a banking app.
A consumer crypto wallet built around a single belief: most people don’t want a Web3 dashboard, they want their bank app — but with the keys in their own hands. Wallet UX, recovery flows, transaction signing, and the boring trust moves that make the whole thing feel safe enough to actually use.
Context
Most self-custody wallets fail one of two ways. They either look like Etherscan with a paint job, or they hide so much of the chain that users can’t tell what they just signed. Avocado was a consumer wallet that wanted to land in the middle: a banking-app surface for people, with full chain transparency one tap away for when they need it.
The brief was specific. Make the first transaction feel as forgettable as a Venmo. Make the recovery flow worthy of the word “custody.”
Approach
Research moves
We ran twelve unmoderated tests against three competitors and our own prototype, all on the same task: send 0.01 ETH to a friend. The pattern was loud. Users didn’t struggle with the send button — they froze on the signing prompt. Modal language like “sign this transaction” reads as legal, not financial. People treat it like an EULA and click through.
Design moves
We borrowed the language of consumer banking. Confirm payment. You’re sending to Sarah, on Ethereum. This will cost about 30 cents in network fees. Then, underneath, a small expand: see the raw transaction. The chain is still there. We just stopped making it the front door.
Decisions
Recovery was the hardest call. The honest pattern is twelve seed words on a piece of paper. The friendly pattern is iCloud backup with a master password. We did both — but defaulted to the encrypted-cloud pattern with a clear, opinionated path to the paper-seed option for users who knew they wanted it. Security advisor signed off; it cost us two extra weeks of pen-test time; it was worth it.
Outcome
The wallet shipped to TestFlight with a thousand invited testers. The first-transaction completion rate moved from 47% on the pre-redesign prototype to 81% on the shipped beta. Users reported the signing prompt as “the screen they didn’t notice” — which, for a signing prompt, is the goal.
The chain is not the product. The chain is the plumbing. People use plumbing. They don’t look at it.
Reflection
What I’d do differently next time: start with the signing prompt, not the home screen. The home screen looks like the product. The signing prompt is the product. We spent three weeks on the home before we touched the signing flow; that order was wrong.
What I’m proud of: the recovery flow. It’s the part of any self-custody product where bad design becomes someone losing their savings. We took it seriously enough that I’d still ship it today, two years later, without changing the structure.