TVL, Liquidity and Incentive Distortion

Total value locked is the number that arrives first. It appears in the opening paragraph of funding requests, in ecosystem updates, in comparisons between chains, and in arguments that a programme worked or did not. It is quoted more often than it is defined, and the gap between the quoting and the defining is where reviewers get into trouble. This lesson is not an argument that the figure is worthless. It is a set of tools for reading it: what it measures, what it cannot measure, and the specific ways an honest number can still mislead a careful person.

• Define total value locked precisely enough to say what a specific figure includes and excludes.
• Separate a stock from a flow, and refuse comparisons that mix the two.
• Spot denomination effects, double counting and recursion inside a headline number.
• Ask four questions that make any locked-value figure readable by someone who was not there.

The first discipline is to hold the difference between a stock and a flow. A locked-value figure tells you how much value is sitting inside a set of contracts at the instant of measurement. It tells you nothing on its own about how hard that value is working. A pool holding a large balance that nobody trades against produces almost no fees and provides almost no service, while a smaller pool with constant turnover may be doing far more useful work. The stock says how much is parked. Flows - volume, fees, unique users, transactions, borrowing utilisation - say what is happening.

This matters because proposals routinely present a stock as though it were an outcome. "The programme increased TVL by a large margin" is a statement about a balance. It is not a statement that anyone traded, borrowed, built, or benefited. The corresponding review question is not hostile and should be asked in a neutral register: alongside the locked figure, what happened to volume, to fees, to the number of distinct participants, and to whatever activity the proposal actually cares about? If the stock rose and every flow was flat, that is a real and interesting finding, and it belongs in the record without editorial.

The mirror-image mistake is to dismiss the stock because it is not a flow. Depth of liquidity is a genuine service: it is what allows someone to transact at a predictable price, and it cannot exist without capital sitting there. The point is not that stocks are inferior to flows. It is that they answer different questions, and a proposal should be assessed against the question it claims to be answering.

Every locked-value figure is denominated in something. On Cardano the natural units are ada and a fiat reporting unit, and the two tell noticeably different stories whenever the ada price moves. If a pool holds exactly the same quantity of ada and the same quantity of a token from one week to the next, and the ada price rises, the fiat-denominated figure rises. Nothing changed about participation, depth in ada terms, or the service the pool provides to an ada-denominated user.

The practical rule: whenever a growth claim is made in a fiat unit, ask for the same series in native units, and vice versa. Where the two disagree in direction, that disagreement is the finding. Where a figure includes assets whose prices are themselves thin or reflexive - for example a governance token that is largely held inside the protocol it governs - note that the valuation depends on a price that a modest sale could move.

When you deposit an asset into a protocol you often receive something back: a receipt token, a liquidity-provider position, a wrapped representation. That receipt is a claim on the deposit, not new value. If the receipt can itself be deposited somewhere that counts it as locked value, the same underlying asset now appears in two totals. Aggregate the two protocols and the ecosystem figure counts one unit of value twice. Methodologies differ in how aggressively they net this out, which is one reason two credible sources can disagree about the same protocol.

Recursion is the same mechanism with a loop. Deposit an asset, borrow against it, buy more of the asset, deposit again. Each round adds to the headline figure while the underlying capital that entered the system from outside has not changed. Leverage is a legitimate financial activity and this course does not treat it as suspect. But a reviewer must be able to say whether a reported increase reflects new capital arriving or the same capital being counted through more layers, because the two behave very differently under stress: layered positions unwind together.

Two features of Cardano shape how these figures are constructed. First, assets are native to the ledger: a token is tracked by the ledger itself under a minting policy rather than by a separate contract standard. That makes it comparatively easy to enumerate what an address or script actually holds, and it makes the inclusion question sharper - a methodology has to decide which minting policies count as assets with a meaningful price and which do not. A token that anyone can mint freely can be deposited into a pool and, if priced naively, can inflate a total dramatically.

Second, the extended UTXO model means pool state lives in specific outputs consumed and recreated by transactions, rather than in a mutable account balance. In practice this affects how protocols batch user actions and how contention is handled at busy moments. For the reviewer this matters mostly at the edges: figures may be derived from indexed script addresses, and understanding whether a methodology follows all of a protocol's script addresses - including newer versions after an upgrade - is part of asking what the figure includes. A protocol that migrates to a new script version can appear to lose most of its locked value overnight in a source that has not updated its address list.

It would be easy to leave this lesson believing that locked value is a vanity metric. That conclusion is as unhelpful as uncritical acceptance. Consider a lending protocol whose review question is whether a borrower of moderate size can obtain a loan without exhausting the available supply. Here the stock is exactly the right measure: the quantity of assets available to be borrowed is the service. A reviewer who dismisses the figure because "TVL does not measure activity" has answered a question nobody asked.

Similarly, a large and stable locked balance held across many distinct depositors, in assets with liquid external markets, over a long period and without an incentive programme running, is meaningful information about durability. The figure earns its weight from the qualifiers, not from its size. The reviewer's job is to attach the qualifiers, not to accept or reject the metric wholesale.

• Quoting a figure without the date, the source and the methodology version.
• Comparing a stock (locked value) with a flow (volume, fees) as though they were the same kind of quantity.
• Reading a fiat-denominated increase as participation growth without checking the native-unit series.
• Adding protocol figures together across an ecosystem without asking whether receipts are counted twice.
• Treating a decline as automatic evidence of failure, when a methodology change, a script migration or a price move may explain it.
• Treating the metric as a measure of public benefit. It measures deposits, not usefulness.

You can now take a locked-value figure and say what kind of quantity it is, in what unit, under what inclusion rule, as of when, and where it might count the same value more than once. You can state whether a claimed increase is consistent with new participation, with a price move, with layering, or with a change in how the figure is compiled. That is enough to stop a headline number from doing argumentative work it cannot support, and enough to use it properly where it genuinely is the right measure.

Decomposing the figure does not tell you whether the deposits are useful, whether the depositors are distinct people, whether the capital will stay, or whether the protocol is safe. It does not tell you the quality of execution a user would receive, which is the subject of the next lesson, nor whether any of it would have happened without funding, which is the subject of the two after that. And unless you have read the methodology yourself, the inclusion rules remain a claim by the aggregator rather than something you have verified.

Locked value is a stock in a denomination, as of a date, under a rule. Missing any of those four, it is a headline rather than evidence.

Receipts, wrapped claims and recursive deposits can make one unit of capital appear several times, especially in ecosystem-level aggregates.

A fiat-denominated increase may be a price move. Always ask for the native-unit series alongside it.

The metric is neither a vanity number nor a verdict. It is the right measure for some questions and the wrong one for most claims made with it.