What TVL measures, and what it does not

Total value locked is the headline figure for a DeFi protocol or a chain, and it measures less than its name suggests. It is a dollar valuation of what sits in contracts, so prices move it. It can count the same dollars again at each layer they pass through. And it follows a published set of rules about what counts. Read with those three facts in mind it supports one comparison well, like with like.

The definition

Every total value locked figure on this site’s chain pages and protocol pages comes from DefiLlama, as a dated snapshot, and means what DefiLlama says it means. Its documentation, read on 3 October 2026 like every source cited here, gives the definition in two places. The methodology page counts the value of tokens locked in the contracts of a protocol. The data definitions page adds the chain case: the TVL of a chain is the sum of the TVL of the protocols on it. It likens the figure to assets under management and states the purpose, which is to proxy the risk depositors take, so only assets actually deposited by users and exposed to the protocol are counted.

Two mechanics sit under that definition. Each listed project has an adapter, code that returns its TVL, and the FAQ says the adapters are open source and open to review. Almost all tokens are priced through CoinGecko’s API, with on-chain methods where that cannot be done, so the dollar figure is a quantity of tokens multiplied by a price.

What the definition leaves out is stated with some care. The data definitions page and the page on what to include as TVL give these rules:

• Five types are tracked beside TVL, not in it. Staking is the protocol’s own governance token staked in its staking system. Pool2 is staked liquidity-provider tokens where one side of the pair is the protocol’s own governance token. Borrowed is value borrowed from a lending protocol, kept separate, DefiLlama says, so that cycled lending cannot inflate TVL. Vesting is tokens not yet circulating or issued, and offers are funds approved for spending but not deposited. DefiLlama shows them as optional toggles next to TVL rather than including them by default.
• A protocol’s own minted token is not counted back in. DefiLlama’s example is a stablecoin deposited back into the protocol that minted it, which is not counted because the collateral behind it already is.
• Some holdings are not counted. Bonds and fiat, such as the dollars in the bank account behind a stablecoin, are not on a blockchain. Funds in smart contract wallets are not counted, and neither is the native staking of a chain’s own token (ETH proof-of-stake and ADA are DefiLlama’s examples). DefiLlama’s FAQ says counting that would overshadow the TVL of DeFi protocols and make chain TVL just a proxy of token market cap, removing usage as the thing the metric tracks. A liquid-staking protocol is the exception: it has its own TVL, and only chain TVL leaves it out by default.
• Deposits that serve no function are being removed. DefiLlama says it is improving the metric by removing unproductive or artificial liquidity, meaning assets deposited only to earn rewards or lift the figure without taking real market risk.

The data pages here show what DefiLlama’s API returns. The endpoint descriptions read do not say which of the types above a figure includes, except that the chain history series is described as excluding liquid staking and double-counted TVL, so a figure is best read as a constructed number with decisions built into it, not a raw sum of balances.

Price moves it without anyone moving money

TVL moves for three reasons the headline figure cannot separate: people deposited or withdrew, prices moved, or what is counted changed (the last is taken up under the trend of one entity, below). DefiLlama’s data definitions page makes the point for the first two: a protocol’s TVL can fall while more assets are deposited, if the prices of what it holds fall, so the chart alone is not the best way to tell whether money is arriving or leaving. Its worked case is a protocol with all of its TVL in ETH: if the price drops 20% with no new deposits or withdrawals, TVL drops 20% and its USD inflows metric, which values each asset’s change in balance at that asset’s price, is zero.

An illustration with invented numbers, not a reading of any protocol. A protocol holds 1,000 units of one token and nothing else. At $10 a token its TVL is $10,000. The price falls to $8 and nobody deposits or withdraws: TVL is $8,000, a fall of 20%, and the 1,000 units have not moved. Suppose instead that 100 more units are deposited on the same day. TVL is 1,100 units at $8, or $8,800, a fall of 12%, although the quantity held rose by 10%. The dollar chart shows a fall in both cases, and only the quantity tells them apart.

The quantity series, tokens deposited rather than dollars, carries no price. The data pages here show dollars only, and DefiLlama’s API documents its per-protocol endpoint as returning breakdowns by token and chain. What exchange reserve data shows, and what it hides sets out three ways to separate price from flows for a reserves figure, and the same arithmetic applies here. Tokens pegged to the dollar are the least affected: a token built to hold a dollar keeps about that value in TVL for as long as its peg holds (see what holds each kind of peg), so a protocol whose deposits are mostly in such tokens has a dollar series that price barely moves.

Counted again at every layer

A deposit in DeFi can produce a receipt, a token that represents a claim on what was deposited, and the receipt can be deposited somewhere else. TVL is the value of the coins held in a protocol’s contracts, so each protocol that holds the receipt counts it.

An illustration with invented numbers. A holder deposits $100 of a token in a lending market and receives a receipt token worth $100. The lending market’s contracts hold $100, so its TVL is $100. The holder deposits the receipt in a yield vault, whose contracts now hold a receipt worth $100, so the vault’s TVL is $100 as well. One holder and one original $100 sit in two TVLs, and adding the two gives $200. Put a liquid-staking token underneath and there is a third. The holder stakes $100 of ether with a liquid-staking protocol and receives a token that ethereum.org describes as a receipt for the staked ether, one that can be put to use as collateral in DeFi applications. The token goes into the lending market and the lending receipt into the vault. The same $100 is in three TVLs, and $300 when they are added.

Restaking is a further use of the same position. Ethereum.org calls it a separate risk category built on top of liquid-staking tokens, and the EigenLayer contracts repository lists native ETH, liquid-staking tokens and ERC-20 tokens among the assets its protocol supports for restaking. DefiLlama files some protocols under a Restaking category, which the chain pages here use as a label, and none of the DefiLlama pages read states a rule for restaking as such. If a restaking protocol’s contracts hold the liquid-staking token, its TVL counts it, and the stack has another layer over the same value.

DefiLlama has rules for parts of this. Within one protocol, a receipt deposited in another part of the same protocol is counted once, the same logic as the minted-token example above. Across protocols the answer is a flag: when receipt or LP tokens from one protocol are deposited into another, the second protocol is flagged as double-counted so that the same underlying capital is not summed twice, and the website offers a doublecount toggle to count such tokens or not. None of the DefiLlama pages read says how a protocol comes to be flagged or how completely the flag is applied. A table of protocols ranked by TVL shows each protocol’s own figure and does not net one against another.

Liquid staking and lending, two of the layers above, are among the sources where crypto yield comes from takes in turn, each with its own payer and its own loss case.

The same rules explain why chain TVL and the sum of its protocols’ TVL will not match, by design. DefiLlama defines chain TVL as the sum of the TVL of the protocols on the chain, and also states exclusions that apply at chain level: native staking never counts, liquid-staking protocols are not counted towards chain TVL by default, bridge TVL is attributed to no chain, and capital in protocols flagged as double-counted is not summed twice. A sum over every protocol on a chain can therefore run above the chain’s figure. The chain pages here list the largest protocols on a chain, each at its own figure on that chain, and the note on each says the two totals will not match. That table is not a decomposition of the headline. It can include liquid-staking and bridge entries, the categories DefiLlama’s chain rules treat separately.

Bridged assets, and where they are

Assets move between chains through bridges, and ethereum.org lists three designs: lock assets on the source chain and mint on the destination chain, burn on the source and mint on the destination, or swap with another party. In the first, the original is locked on one chain while a representation is minted on the other. Where the value is then has two defensible answers, and DefiLlama’s methodology page says as much: there are arguments for counting bridge TVL on the origin chain and on the destination chain.

An illustration with invented numbers. 100 tokens are locked in a lock-and-mint bridge on chain A, and 100 representations are minted on chain B and deposited in a lending market there. A rule that counted both ends would put 100 on chain A, for the locked originals, and 100 on chain B, for the deposit: the same value on two chains’ tables. DefiLlama’s rule does not. It counts the TVL of bridge projects but contributes none of it to the TVL of any chain. The lending market’s contracts are on chain B, so the definition puts its 100 there. The bridge’s 100 counts as the bridge’s own TVL and on no chain, and chain A shows neither. At the protocol level the same 100 is still in two TVLs, the bridge’s and the lending market’s.

A bridge’s TVL is therefore a balance: the value held in the bridge project’s contracts. For a lock-and-mint design that is the collateral behind the representations. For the other designs ethereum.org lists it is whatever the contracts hold. It is not a count of the value that has crossed the bridge, which would be a flow.

Where assets sit is a separate measure in DefiLlama’s documentation. Bridged TVL, also called Chain Assets, is the value of the assets on a chain split by origin, whether native to the chain or bridged through its canonical bridge or through third-party bridges, tracking where possible the collateral backing bridged tokens on the source chain rather than trusting the minted representation. The data pages here do not show it. A chain’s TVL on these pages says where deposits were made into listed protocols, under the rules above, and is not an inventory of what sits on the chain.

What TVL is good for

The comparison TVL supports is like with like: figures built the same way, in the same unit, on the same date basis. It serves three uses.

Scale within one category. Lending markets against lending markets, or liquid-staking protocols against each other, hold the same kind of thing under the same rules, so the size of one against another is a comparison of scale. Across categories it compares unlike things: a bridge’s figure is a balance held in its contracts, a lending market’s is deposits less what has been borrowed out, and a liquid-staking protocol’s value can be counted again in the figures of the protocols that hold its token. Inside a category the asset mix still matters, because protocols holding different assets respond differently to the same price move.

The trend of one entity. A protocol’s series comes from its own adapter and a chain’s from the adapters of the protocols counted towards it, each under one definition, so either has fewer moving parts than a comparison across entities once price has been allowed for. That basis is not fixed: adapters are open code that can change, and the unproductive-assets section of DefiLlama’s what-to-include page lists removals protocol by protocol, with reasons and no dates, and does not say whether history is recalculated afterwards. A step in a series may therefore reflect a change in what is counted rather than a movement of deposits, and nothing on the page places the one against the other.

A change worth explaining. A sharp move in the figure is a question, not a finding. Price comes first, then flows, then a change in what is counted.

What it is not follows from the construction. It is not a count of users: the figure is a value deposited, and the same dollars can come from one depositor or from many. It is not revenue. DefiLlama tracks fees and revenue as separate metrics, and the protocol pages here show the fees it recorded beside TVL, under methodologies that vary between protocols. Fees and revenue are defined and measured differently from TVL, and the companion piece on protocol fees and revenue sets them out. And it is not a measure of safety or quality: it says how much value is exposed to a protocol, not how well the protocol is built, tested or run.

Reading a TVL chart

Four checks, in this order.

1. The unit and the date. The headline figures on the data pages are US dollars at the prices of the snapshot date, and each page says the date. The history chart is dollars by date, drawn as weekly points for the last two years and monthly before that, so a spike shorter than a week can fall between points.
2. What price did in the same window. Find what the protocol mainly holds from its own documentation or contracts, and what those assets’ prices did over the same dates, from any price source. If the dollar change is close to what those prices alone would produce, no deposits or withdrawals are needed to explain it.
3. The category. DefiLlama files each protocol under one category, and the protocol pages here show it beside the number of chains the protocol is deployed on. Compare a figure with figures from the same category, and ask which layer it sits at, as the sections on counting above set out.
4. The breakdown. A protocol page splits its TVL by chain, and a chain page lists the largest protocols on the chain. Use them to find where a move came from. A protocol page’s split divides that protocol’s own figure by chain (the largest twenty are listed), whereas a chain page’s list is on a different basis and does not add up to the chain’s figure. The figures from the latest snapshot are in the chains table and the protocols table, and each row opens a page laid out as described here.

Read in that order, a TVL figure answers a narrow question well: how much value sits in these contracts, at these prices, under DefiLlama’s rules, compared with figures built the same way.