Why DeFiLlama's TVL is the most quoted, and most misused, number in DeFi
If you have spent even a week reading crypto research, you have seen the pattern. A thread on X announces a protocol "hits $4B in TVL". A fund manager posts a chart of DeFiLlama's ranking. A governance proposal argues that its treasury is "only" 2% of Lido's TVL. TVL, or total value locked, has become the default way to size a DeFi protocol, and DeFiLlama is the default place to read it.
That default is useful, and dangerous. DeFiLlama is the most comprehensive open dashboard in crypto. It tracks more than 2,000 protocols across more than 200 chains, ranks them by TVL, breaks them down by category and chain, and exposes the raw numbers behind the headlines. Most alternative dashboards either license DeFiLlama's data, scrape it, or quote it. For an open, free, community-maintained resource, it is genuinely impressive.
The problem is that "TVL" sounds like a single, objective measurement, the way "market cap" sounds for a stock. It is not. The same dollar can be counted several times depending on how the protocol works. A lending market's supplied assets and its borrowed assets are both reported. A bridged version of USDC is often counted separately from its canonical version. And almost none of the numbers are audited. DeFiLlama is a starting point for investigation, not the final word.
What the top of a protocol page is actually showing you
Open any protocol page on DeFiLlama, say Aave or Pendle, and you will see a row of large numbers at the top. They look similar, but they mean different things. Understanding the difference is the first step to reading the page correctly.
TVL is the headline number. It sums the dollar value of assets sitting in the protocol's smart contracts, using a price feed that DeFiLlama maintains. For a simple DEX like Uniswap, this is roughly the liquidity in its pools. For a lending market like Aave, it is the total supplied assets before subtracting borrows.
TVL double-counted is the raw figure before DeFiLlama tries to remove overlap. If Protocol A deposits $1B into Protocol B, DeFiLlama may show that same $1B in both A's and B's totals. The double-counted line keeps that overlap in place, which inflates the headline number for the wider DeFi ecosystem.
TVL excluding double-count is the cleaner figure. DeFiLlama walks the dependency graph between protocols, finds assets that are simply "wrapped" or "restaked" versions of each other, and removes the duplicated dollars. This number is usually the one serious researchers cite, because it approximates the real capital sitting in the system.
You will often see the gap between the two figures grow as the restaking and yield-aggregator meta heats up. In 2024, EigenLayer's growth pushed billions of dollars of ETH into restaking positions, and many of those positions were then deposited into Pendle, Spectra, or other yield markets. Each layer of wrapping adds another line on the dependency graph, and DeFiLlama's excluding-double-count figure is what stops that from being counted three or four times.
Lending pages: why "supplied" and "borrowed" matter more than headline TVL
Lending protocols are the easiest place to misread a DeFiLlama page. The top of Aave's page shows a TVL number that looks like a vault size. It is not. It is the sum of everything depositors have put in, including the part that borrowers have taken out. The protocol does not have that money as free collateral; borrowers do.
That is why the supplied and borrowed lines below the headline exist. Supplied minus borrowed is roughly the "available liquidity" sitting in the pool, the part that depositors could withdraw without triggering liquidations. If supplied is $10B and borrowed is $7B, headline TVL is $10B but the depositors' cushion against a bank-run scenario is closer to $3B.
For an extreme example, look at a market like wBTC on Aave during a stress event. If supplied stays flat and borrowed rises, the headline TVL stays the same while the real risk profile worsens. The utilisation line, supplied divided by supplied, climbs toward 100%, which is when borrow rates spike and liquidations become more likely.
DeFiLlama also breaks this down per asset. Aave's page lists every market with its own supplied, borrowed, and utilisation figures. If you are about to deposit into a specific market, that per-asset view matters more than the protocol-wide TVL. A protocol with $20B in headline TVL can still have one obscure pool that is almost entirely borrowed out and paying 30% to attract new depositors because risk is being repriced.
Staking, LSTs, and LRTs: where the definitions get tricky
Staking protocols are another place where DeFiLlama's categories require careful reading. Liquid staking tokens (LSTs) like Lido's stETH are counted in TVL, because they represent ETH that is locked in the beacon chain. Restaking tokens (LRTs), the yield-bearing wrappers issued by EigenLayer operators, are also counted, but their underlying assets often overlap with the LSTs they restaked.
This is where the double-count adjustment does the most work. If you stake ETH with Lido, you get stETH. If you then deposit stETH into Pendle, DeFiLlama sees stETH in Lido, a stETH derivative in Pendle, and possibly another wrapper in an LRT protocol on top. Without the excluding-double-count adjustment, the same ETH gets counted multiple times.
Staking is also where chains disagree most loudly about how to count. Solana's TVL includes SOL delegated to validators, plus liquid staking tokens like mSOL and jitoSOL. Ethereum's TVL excludes the ETH staked directly with validators, because DeFiLlama counts only assets inside smart contracts. That is why Solana's headline TVL looks high relative to its DeFi activity: a lot of the number is the network's staking base, not user-deposited capital.
Look at the "Staking" category on the homepage and you will see Lido, Rocket Pool, and similar protocols at the top. Those numbers include the underlying staked asset plus the protocol's own tokenomics. When evaluating a liquid staking protocol, it is worth opening the breakdown tab and checking how much of TVL is in the LST versus how much is in the protocol's governance token.
Bridged vs canonical: how chain breakdowns can mislead you
DeFiLlama's chain breakdown is one of its most useful features, and one of its biggest traps. Every protocol page lists TVL by chain, so you can see how much of Aave sits on Ethereum, Arbitrum, Base, Polygon, and so on. But that breakdown often includes bridged versions of the same asset.
Canonical TVL is the dollar value held in the asset's "home" form on its native chain. Bridged TVL is the same asset wrapped for use on another chain, for example USDC.e on Avalanche versus native USDC on Avalanche, or WETH on Optimism versus native ETH on Ethereum. DeFiLlama lists both. They are not always additive.
This matters when people use the chain breakdown to claim that Base has overtaken Arbitrum, or that Hyperliquid's TVL is bigger than its competitors'. The number can rise simply because more bridging volume passed through, not because new capital entered the ecosystem. A protocol can show $500M on Base and $300M on Arbitrum, but if the $500M is bridged USDC and the $300M is bridged USDC.e, the actual unique capital might be far smaller than the headline suggests.
It is also where "chain-hopping" shows up. A protocol that launches on Ethereum, then on Arbitrum, then on Base, then on a new L2 every quarter, can post rising TVL across each chain without ever attracting genuinely new users. The chain breakdown reveals this: you see TVL split roughly evenly across many chains, with no single chain dominant. That pattern is usually a marketing tour, not a real product-market fit.
The biggest limitation: DeFiLlama is self-reported
Every number on DeFiLlama comes from somewhere, and that somewhere is usually the project itself. Protocols submit their contracts to be listed, and DeFiLlama writes adapters that read on-chain balances or pull from the protocol's subgraph. For major protocols, those adapters are well-maintained and audited by community contributors. For smaller protocols, the adapter might be a pull request the team submitted themselves.
This means a malicious or careless protocol can, in theory, inflate its TVL. The most common pattern is "TVL wash trading": the team deposits its own treasury, plus borrowed funds from a lending market it controls, into its own pools. The on-chain balance is real, so DeFiLlama records it, but the actual user capital is zero.
DeFiLlama has added some safeguards. The team reviews new listings, marks protocols as "hacks" or "rugs" when they fail, and excludes known wash trading. But the surface area is enormous, and the project does not have an auditor reviewing every adapter. Treat the number as accurate for blue-chip protocols and as a starting hypothesis for everything else.
This is also why a single dashboard is never enough. Token Terminal prices protocols on fees and revenue, which are harder to fake. EigenFi focuses on restaking flows. Etherscan and the chain explorers let you verify the actual contract balances. Dune dashboards, when the SQL is shared, let you reproduce the numbers yourself. None of these replace DeFiLlama, but together they bound the range of plausible answers.
How to read a protocol page in practice
Put it all together and the workflow looks like this. First, open the protocol's page and check the headline TVL, then immediately switch to "excluding double-count" for a sanity check. Note the gap; if the gap is more than 30%, the protocol is heavily dependent on other protocols' assets, and the headline number is probably misleading.
Second, look at the chain breakdown. Is one chain dominant, or is TVL spread thin across many chains? Is the asset list mostly bridged versions of the same thing? A protocol with concentrated TVL on one chain and a balanced mix of canonical assets is usually more "real" than a protocol with TVL on eight chains and almost entirely bridged stablecoins.
Third, if it is a lending protocol, open the supplied/borrowed split. Calculate the utilisation per market and watch the trends over the past 30 days. Rising utilisation with stable TVL means risk is building. For a staking protocol, open the asset breakdown and see how much is the LST versus the protocol's own governance token.
Fourth, cross-check with one alternative source. Token Terminal's revenue and fees chart tells you whether the protocol is actually being used, not just whether capital is parked there. For lending markets, check the protocol's own risk dashboard, which usually lists liquidation thresholds and oracle sources. For restaking, EigenFi's operator breakdown shows where the assets really sit.
Fifth, check the history tab. TVL spikes that don't match a product launch, a yield program, or a market-wide move are a red flag. Wash trading looks like a single block of capital appearing and disappearing. Real organic growth looks like a slow climb.
Follow DeFi protocols with context, not just numbers
DeFiLlama gives you the raw material, but the context behind each line item changes weekly. New restaking launches, new bridge integrations, and new yield programs shift which protocols are double-counted and which are not. Tracking all of that by hand is a losing game. Zippfeed surfaces the protocol and chain headlines that move TVL, with sentiment scoring (bullish, neutral, or bearish) and an importance rating, so you can spot the changes before they show up in next week's chart.
