TL;DR
TVL (Total Value Locked) is a balance-sheet snapshot that doesn't tell you whether the capital is doing anything. It distorts in three predictable ways: composability double-counts the same dollar across multiple protocols, USD denomination converts price moves into apparent growth, and mercenary capital makes incentive-bought TVL look organic. The fastest sanity check is the Fee/TVL ratio: productive capital generates fees in proportion to size; parked capital does not. Below 1% annualized, the TVL is mostly idle.
TVR (Total Value Redeemable), TVS (Total Value Secured), and REV (Real Economic Value) aren't replacements for TVL. They're complementary lenses that answer different questions. A serious read uses several together.
For every research page that cites TVL, we now pair it with the trailing-30-day annualized Fee/TVL ratio (DefiLlama). Below 1% gets a yellow flag. Above 10% gets a green flag. Why: a $25B protocol generating $50M of fees a year (0.2%) and a $1B protocol generating $200M of fees a year (20%) are not in the same business, but headline TVL would suggest the first is 25x more important.
The three ways TVL distorts
1. Composability double-counts
DeFi composability is real and useful. It also breaks the math. Deposit 1 ETH into Aave, get aETH back. Use aETH as collateral in a CDP. The CDP-issued stablecoin gets supplied to a money market. The same ETH appears in the TVL of three protocols. Restaking, LRT wrappers, and looping strategies push the count higher.
DefiLlama tracks an "Adjusted TVL" view that nets some of this out at the aggregator level, but per-protocol pages don't always reflect it. When two trackers disagree by more than 20% on the same protocol, double-counting is usually the reason.
2. USD denomination converts price into "growth"
TVL is reported in dollars. If ETH is up 30% week-over-week, every Ethereum-denominated protocol's TVL is mechanically up ~30%, even if not a single new deposit arrived. The reverse holds in drawdowns: a "TVL collapse" headline often hides that token amounts are flat and prices fell.
The fix is the inflows view. DefiLlama exposes a "Show Inflows" toggle that strips out price effects and shows only net deposits/withdrawals. If inflows are flat or negative while TVL is up, you're looking at a price story, not an adoption story.
3. Mercenary capital fakes adoption
Protocols routinely pay liquidity into existence. Run a $50M-per-month emissions program, attract capital that's there for the rewards, post a "$1.2B TVL milestone" headline, then watch it leave the day incentives cliff. The 2025-2026 cycle had several reset moments: Berachain shed roughly 90% of TVL when the genesis emissions schedule wound down, and Unichain saw a similar collapse.
The fix is to look at the organic baseline, not the peak. Pull TVL three months before the incentive program launched and three months after it ended (where data exists). The delta between launch-peak and post-cliff baseline is the mercenary-capital fraction.
The 7-point authenticity checklist
When TI evaluates a TVL claim before publishing, this is the sequence. None of these requires more than a DefiLlama page and 10 minutes.
On DefiLlama, toggle "Show Inflows" on the protocol's TVL chart. If TVL is up but net inflows are flat or negative over the same window, the move is price-driven, not adoption-driven. State the framing accordingly.
Healthy TVL is mostly independent assets (ETH, BTC, USDC, USDT). If a protocol's TVL is dominated by its own governance token, deposits are a circular subsidy: the protocol prints the token, pays it as rewards, the rewards get re-deposited, the deposit raises TVL, the higher TVL supports a higher token price. A 20%+ native-token share of TVL is the standard red flag.
If there's a live emissions program, mentally subtract it. Compare three windows: the three months before incentives started, the active-incentive peak, and the three months after incentives ended (if available). The first and third are the organic baseline. The second is the marketing story.
Productive capital generates fees in proportion to size. If TVL doubles and fees-per-dollar-TVL falls, the new capital is parked. The full live snapshot for TI's coverage is below; as a rough heuristic, anything under 1% annualized Fee/TVL is parked, 1-5% is normal, 5-15% is active, above 15% is exceptionally productive.
If DefiLlama and Token Terminal (or L2Beat for L2s) disagree by more than 20% on the same protocol, something is being double-counted or scoped differently. The Sky/Maker page surfaced this directly: SFF's broad-scope figure of $12.46B Sky Lending TVL vs DefiLlama's narrower-scope ~$5-7B for the same protocol, end of Q1 2026. Neither was wrong. They were measuring different things. State which scope you're using.
Use Etherscan, Arkham, or Nansen to check which addresses make up the TVL. If three to five wallets account for more than 50%, the TVL is fragile: any one of them rotating out is a single-day capital flight risk. Diffuse holder bases (hundreds-to-thousands of depositors) are far more resilient. This is one of the cheapest robustness checks available.
Pull the TVL chart through the Luna/UST collapse (May 2022), the FTX implosion (Nov 2022), the March 2023 USDC depeg, and recent macro-driven drawdowns. Protocols that retained the bulk of their TVL through these had real utility for real users. Protocols whose TVL evaporated within days were rented liquidity that left with the music.
Live Fee/TVL snapshot — TI's coverage
Below: trailing-30-day annualized Fee/TVL and Revenue/TVL ratios for the protocols TI covers, computed from DefiLlama. Fee/TVL measures all fees the protocol generates per dollar of TVL. Revenue/TVL is the subset that the protocol actually retains (excludes LP/staker pass-through). The gap between the two tells you how much value flows to the protocol vs. its liquidity providers.
| Protocol | TVL | Fees ann. | Revenue ann. | Fee/TVL | Revenue/TVL |
|---|---|---|---|---|---|
| Aerodrome | $368M | $88.7M | $88.7M | 24.07% | 24.07% |
| Jupiter | $684M | $141.7M | $52.6M | 20.70% | 7.68% |
| Hyperliquid | $4.75B | $726.2M | $646.1M | 15.28% | 13.59% |
| Uniswap | $3.45B | $522.6M | $44.0M | 15.13% | 1.27% |
| Sky Lending | $6.16B | $408.0M | $179.9M | 6.62% | 2.92% |
| Ethena | $4.43B | $216.6M | $2.9M | 4.89% | 0.07% |
| Aave V3 | $25.64B | $762.4M | $99.0M | 2.97% | 0.39% |
| Lido | $20.88B | $573.8M | $57.4M | 2.75% | 0.27% |
| Morpho Blue | $10.91B | $176.9M | $0 | 1.62% | 0.00% |
| Pendle | $1.61B | $9.3M | $9.2M | 0.58% | 0.57% |
Source: DefiLlama protocol endpoints + dailyFees / dailyRevenue summaries, 30-day windows annualized (×12.17). Snapshot date: 2026-04-30. TVL excludes borrowed positions, staking-program double-count, and pool2 incentive pools where present.
What the snapshot tells you
The Fee/TVL ranking surfaces a few non-obvious distinctions:
- Hyperliquid (15% Fee/TVL, 14% Revenue/TVL) is the standout. The narrow gap means almost all generated fees stay with the protocol or its token, which is the point of the buyback-burn architecture. This is "active TVL" by every measure.
- Aerodrome and Jupiter (24% and 21%) show the highest Fee/TVL because they're DEX/perp activity layers, not capital warehouses. Their TVL is small relative to the volume passing through.
- Aave (3.0% Fee/TVL, 0.4% Revenue/TVL) is the textbook lending profile: large fee base, but most fees pass through to depositors as yield rather than to AAVE holders. This is structural, not a problem, but it explains why Aave's market cap multiple looks different than a DEX's despite similar headline TVL.
- Morpho Blue (1.6% Fee/TVL, 0% Revenue/TVL) has zero protocol take-rate by design at the base layer. Curators (Steakhouse, Gauntlet, etc.) capture the fees, not the protocol. A reader who only looks at headline TVL misses this entirely.
- Pendle (0.6% Fee/TVL) stands out as the lowest in this set. Either the fee model genuinely doesn't generate proportional revenue at current PT/YT volume, or DefiLlama isn't capturing all of Pendle's fee streams. Worth a closer read for any Pendle-related thesis.
What TVL doesn't measure: complementary lenses
TVR, TVS, and REV are sometimes pitched as TVL replacements. They aren't. They're different lenses on different questions, and a serious read uses several together.
TVR — Total Value Redeemable
From Cambridge CCAF research. Strips out derivative wrappers (aTokens, LP tokens, restaked positions) and counts only the base assets a holder could actually redeem at par if they exited. This is the closest answer to "how much real capital is here, after netting out the composability stack." Gap between TVL and TVR is approximately the composability double-count.
TVS — Total Value Secured
From L2Beat. Specific to L2s. Counts every asset bridged to an L2 system, regardless of whether the asset is being actively used in a dApp. The right metric for evaluating L2 security and capital footprint, but it can be gamed: any contract holding funds can in theory be tagged as an "L2 bridge."
REV — Real Economic Value
Pushed by Blockworks. Combines protocol fee revenue, MEV captured, buyback flows, and staker-distributed value into a single number. This is the income-statement view, where TVL is the balance-sheet view. REV answers "what does this system produce" rather than "what is locked inside it." The Fee/TVL ratio used throughout this page is essentially a normalized REV-per-dollar-of-TVL.
VTVL — Verifiable TVL
Proposed in 2025 BIS working-paper research. Computed using only raw on-chain data from Ethereum archive nodes, deliberately ignoring off-chain APIs and aggregator adjustments. The argument is that aggregators inevitably make discretionary calls about what to count and what to net out, and a fully on-chain methodology removes that judgment layer. We have not yet seen a widely-adopted implementation; tracking.
For a balance-sheet question (how much capital is at risk?), use TVR. For an L2-security question (how much value is the bridge protecting?), use TVS. For a productivity question (is this capital working?), use Fee/TVL or full REV. For a "how big is this thing" question, raw TVL is fine, with the caveats above. Don't substitute one for another. Asking "what's the real TVL?" usually means you're using the wrong lens for the question you actually have.
The TI default going forward
Wherever a TI research page cites TVL, we now pair it with the trailing-30-day annualized Fee/TVL ratio. Three reasons:
- It's the cheapest sanity check. Two API calls per protocol. No methodology debates.
- It surfaces the parking problem immediately. A reader sees Aave at $25B TVL and might assume scale equals economic relevance. A reader sees Aave at $25B TVL with 3.0% Fee/TVL and Hyperliquid at $4.75B TVL with 15.3% Fee/TVL and immediately understands they're different categories of business.
- It matches how serious institutional allocators already evaluate the space. Headline TVL gets you into the conversation. Productivity gets you to a position size.
The numbers in the table above are also written into TI's canonical-figures registry. The monthly refresh worker (canonical-figures-refresh, runs the 2nd of each month at 14:00 UTC) cross-checks each figure against DefiLlama and emails a drift report when anything moves outside its tolerance. Manual verification before any update keeps the audit trail in git rather than auto-mutated by a cron.
What this doesn't tell you
Fee/TVL is not a complete metric either. It misses:
- Token-level value capture. A protocol with 15% Fee/TVL where 100% of fees go to LPs and 0% to token holders gives the token zero direct cashflow. This is where the value-accrual mechanism class matters: same Fee/TVL number, very different implications for the governance token.
- Sustainability of the fee base. Hyperliquid's 15.3% Fee/TVL is partly a function of high perp leverage. Sustained perp volume requires sustained directional speculation. Drawdown windows compress the ratio.
- Quality of the TVL sources. A protocol could have great Fee/TVL today by serving 5 whale-LPs who all leave next month. Wallet concentration (point 6 on the checklist) is the companion check.
- Comp-set effects. Different protocol categories generate fees on different bases. DEXes generate fees on volume; lending protocols generate fees on borrow demand; LSTs generate revenue from staking yield. Fee/TVL across categories is a directional signal, not a same-comp comparable.
The point isn't that Fee/TVL is the answer. It's that any single TVL number, by itself, is a poor proxy for what an allocator actually wants to know. The right read combines a balance-sheet metric (TVL or TVR), a productivity metric (Fee/TVL or REV), a concentration check, and a sustainability stress-test. Each of these is cheap. Together they're rigorous.