80% of Your Wallet Users Are Gone in a Week. The Fix Isn’t UX

Industry-wide, fewer than 20% of Web3 wallet users return after 7 days. I used to think that was a competitor problem, then I saw it firsthand at a leading non-custodial wallet on TON. Non-custodial wallets are the only product category where users own their exit as completely as they own their assets. If someone decides to leave or loses their seed phrase, they’re gone, and the product has no lever to pull. The wallets that cracked retention in 2025 — Phantom growing to 15M MAUs , MetaMask holding 30M through a full market cycle — did one thing the others didn’t: they made themselves financially load-bearing. Most teams respond to the retention problem with better onboarding, cleaner UI, smoother flows. None of it moves the number, because the number isn’t an onboarding problem. They stopped being places to store assets and became places where users’ actual money was actively working. Here’s what that looks like in practice, from my own experience in the space and from what the public data shows about wallets that grew in 2025-2026. Your Best Users Don’t Come From Ads. They Come From Ecosystem Partnerships. The highest-quality users we acquired at the wallet I worked on didn’t come from performance marketing or App Store optimization. They came from cross-partnerships with ecosystem projects. The logic is simple : a user who downloads a wallet because they want to interact with a specific DeFi protocol, a staking product, or a token launch already has intent. They have a specific job to do, and the wallet is the tool they need to do it. That user funds their wallet on day one. The user who found the app through an ad, stares at an empty balance and churns. In the TON ecosystem, the best acquisition came through protocol-level integrations (DeFi projects, staking platforms, tokenized asset sales). The wallet was the required entry point; the protocol was the reason to enter. 2025 data from other wallets confirms the pattern. Phantom’s growth to 15 million MAUs by mid-2025 tracked the Solana DeFi ecosystem almost exactly — each major protocol launch or airdrop brought a wave of users who arrived pre-motivated to transact. Phantom didn’t grow despite being Solana-native; it grew because it was embedded deeply enough in the Solana ecosystem that every protocol tailwind translated directly into wallet installs. The Revenue Model Has to Match the Retention Model This is the structural insight that separates wallets with durable businesses from those that keep chasing the next bull market. At the wallet I worked on, revenue came from a percentage of the DeFi activity it facilitated (staking, borrowing, etc). Which meant the business model and the retention model pointed at exactly the same user behavior: an active user who transacts is both a retained user and a revenue-generating user. You’re not trading off growth for monetization, they’re the same thing. In 2025, every wallet that reported strong financials ran this same playbook. Phantom’s Hyperliquid integration is the clearest recent case. In July 2025, Phantom launched in-wallet perpetual futures trading powered by Hyperliquid’s API — users could trade 100+ crypto markets with up to 40x leverage without leaving the app. In the first 16 days, the integration processed $1.8 billion in volume and generated $930,000 in revenue from roughly 17,000 unique users . By early 2026, the cumulative volume through the integration had reached $37 billion , with $20 million in total revenue earned by Phantom — entirely from facilitating trades that users were already motivated to make. What makes this case instructive is the mechanic. Phantom didn’t add a subscription tier or a native token to monetize. They partnered with a DeFi protocol their users already wanted access to, embedded the experience in the wallet, and earned on the activity. The same users that drove revenue were the users most likely to return daily: active traders with open positions. MetaMask ran the same play in October 2025, launching in-wallet perpetual futures via Hyperliquid and introducing MetaMask Rewards — a program tying wallet activity directly to incentives. In December 2025, they added native Bitcoin support (following native Solana support in July), further expanding the set of assets users could manage without leaving the wallet. Each integration deepened the user’s dependency on MetaMask as financial infrastructure. Build your monetization model around the same actions that define retained users. If they’re different, you have a structural problem. Depth of Integration Is the Moat “Connect your wallet to this dApp” is table stakes, it gets you an install, not a retained user. What actually drives long-term retention is when the wallet becomes load-bearing for a user’s financial workflow, when switching wallets would mean unwinding real positions and losing real context. Phantom’s PSOL launch in May 2025 is a good example. Rather than just supporting SOL staking through an external protocol, Phantom launched its own liquid staking token natively integrated into the wallet interface. Users who hold PSOL have staking positions that live inside Phantom. They monitor yield, swap in and out, and track PnL directly in the app. Their financial activity is embedded in the product. The Hyperliquid perps integration works the same way. A user with open leveraged positions and stop-losses set inside Phantom opens the app multiple times a day not out of habit, but because they have active financial stakes there. The same pattern held firsthand. Users managing active staking positions or borrowing strategies retained at dramatically higher rates at 90 days than passive holders — the gap was large enough that aggregate retention metrics were almost entirely driven by which cohort dominated in a given period. The wallet that just stores assets competes on UI but the wallet that manages active financial positions competes on trust. Integration depth determines switching cost. Users with active positions embedded in your wallet have a reason to stay. The DeFi-Active User Is a Different Retention Curve Not all wallet users are equal and in a non-custodial context, the gap between user types is wider than in most product categories. The users who retain longest have active DeFi positions: staking yields, open perp trades, borrowing positions, yield farming strategies. These users return because the wallet is connected to real financial activity they care about. Remove that connection and there’s nothing pulling them back. Passive holders behave completely differently. They open the wallet when crypto is going up. They disappear when it isn’t. The aggregate retention stats for the non-custodial wallet market, sub-20% 7-day return rates, sub-10% 30-day dApp re-engagement, are largely driven by this passive cohort. Build your product to serve them and you’ll always be riding market cycles rather than building durable engagement. Every investment in deeper DeFi functionality serves your highest-retention cohort. Better transaction history, position tracking, gas management, protocol integrations — these features do almost nothing for the passive holder and everything for the active DeFi user. Phantom’s build-out in 2025 — perps, liquid staking, cross-chain swaps, SimpleHash integration for real-time token data, was a systematic investment in the infrastructure active users actually need. Trust Is Product, Not Marketing In a non-custodial wallet, trust is the product itself. Every interaction a wallet has with a user is either building or degrading that trust. A transaction that fails silently erodes it. An RPC that goes down when markets are moving erodes it. A push notification that isn’t about their funds erodes it. And crucially, a user who gets exploited while using your wallet doesn’t come back. There’s no customer support path that recovers lost funds in a self-custody environment. In 2025, the leading wallets treated security infrastructure as a retention investment. MetaMask’s Blockaid integration has prevented an estimated $5 billion in user losses since rollout, according to MetaMask’s own data. That figure is marketing, but the underlying mechanic is real: users who don’t get exploited stay. Separately, MetaMask launched an anti-phishing network with SEAL partners in 2025, responding to over $400 million in phishing-related crypto thefts in early 2025. The security investment is also the retention investment. Phantom acquired Blowfish specifically for transaction safety, and Trust Wallet’s in-app Security Scanner has blocked over $162 million in potentially harmful transactions, according to third-party trackers (the company’s own figures put the number higher, at $191 million ). The same principle applied directly: the wallets users trust long-term are the ones that have proven they won’t let them get burned. That trust is built through the product never failing them when it matters most. The Pattern Underneath All of This The wallet I worked on grew because the TON ecosystem grew, and because it became genuine infrastructure for ecosystem activity rather than just a token storage front-end. The same pattern holds across every wallet that built durable retention in 2025: deep integration with DeFi protocols that users actually want to use, a revenue model that earns on the activity it facilitates, and a security posture users can trust with real money. The industry retention numbers are brutal by default — sub-20% after 7 days across the market. The wallets that beat that number are doing it by making themselves genuinely necessary to the financial activity users care about. That’s the only retention strategy that holds through bear markets. The open question is what happens next: as DeFi activity consolidates into a smaller number of dominant protocols, the partnerships that drive this model will become harder to access for newer entrants. The moat widens for wallets that are already embedded. For everyone else, the window to become load-bearing infrastructure is closing.