The crypto industry has spent the better part of a decade obsessing over throughput. Every chain claims tens or hundreds of thousands of TPS in benchmarks. The Ethereum L2 stack has compressed transaction costs to fractions of a cent. Solana routinely processes blocks at TPS counts that exceed Visa's peak. The protocol-level scaling debate has, in any meaningful sense, been won. We have throughput. What we don't have is consumer experience, and that is the actual frontier where the industry's next billion users are going to be won or lost.
Most users still cannot reliably tell which chain a transaction settles on, why their balance shows different amounts in different wallets, why a "gas" fee suddenly spiked, or what happened to the funds that didn't arrive. They can't recover from a phishing approval. They can't tell which bridge is safe. Every "Web3 onboarding" deck shows wallet creation as a single happy-path step; the reality is a maze of seed phrases, network selection, and gas-token funding that nontechnical users cannot navigate without expert hand-holding. The industry has solved the throughput problem and not even started the experience problem.
The data on user retention reflects this honestly, even when industry presentations don't. Internal funnel data from major exchanges shows that fewer than 12% of self-custody wallets created in 2024 are still active 90 days later. Embedded-wallet products like Privy and Dynamic that hide key management from end users see retention rates triple that. Phantom and MetaMask both quietly shipped major UX overhauls in 2025 because their internal product teams identified user abandonment, not throughput, as the binding constraint on growth. The protocol-level engineers solved the easy problem; the consumer-facing teams now have to solve the hard one. The data also shows a clear bifurcation in user paths: technical users who arrive through hardware-wallet purchases retain at materially higher rates than mobile-app users who arrive through advertising. The latter cohort — which the industry needs to grow to reach mass adoption — is exactly the cohort that the current UX serves worst, and the retention curves reflect that mismatch with brutal clarity.
The category that has actually solved consumer UX in adjacent industries — fintech, ride-sharing, food delivery — did it by ruthlessly hiding the underlying machinery. Email users do not think about SMTP. Web users do not think about TLS. The crypto industry's insistence that users must learn to "self-custody" has been, in retrospect, a category error — confusing a technical option for a UX requirement. Account abstraction, embedded wallets, and intent-based interfaces are the first credible attempts to apply that lesson, and they are working where they have been deployed seriously.
The defenders of the current status quo make two main arguments. The first is that self-custody is the entire point of crypto, and any consumer product that hides it is selling out the underlying values. The second is that simplicity comes at the cost of genuine sovereignty, exposing users to platform risk in ways that pure self-custody avoids. Both are real concerns. But framing the choice as binary misreads the engineering opportunity. Modern account abstraction lets builders offer custodial-feeling experiences with sovereignty escape hatches; users can feel like they're using a normal app while retaining the on-chain receipts that make crypto crypto.
Until the industry takes the experience layer as seriously as it took the throughput layer, the next billion users are not coming, and the throughput numbers are mostly self-soothing. The signs of progress — Privy's growth, Coinbase Smart Wallet's adoption, Phantom's consumer overhauls — are real but isolated. What we need is a sustained reallocation of engineering attention toward consumer experience, comparable in scale to what the industry threw at L2 scaling between 2020 and 2023. The protocols are ready. The wallets need to catch up. The next five years of consumer crypto will be defined by whether they do.