How to Read a DeFi Data Caveat
A guide to locating limitations about coverage, estimates, timing and calculation method.
This article may contain affiliate links. Commercial relationships are disclosed in the affiliate policy.
A data caveat is the small piece of text that tells you how much a number is allowed to be wrong. DeFi interfaces show a great deal of precision: eight significant figures, a green arrow, a percentage that updates while you watch. None of that precision comes from the data. All of it comes from the method, and the method is what the caveat describes.
The reason to read caveats is not caution. It is that a caveat changes what a number means, and a number read without its meaning leads to a wrong decision while looking precise enough to be trusted.
The four things a caveat usually limits
Caveats are written in many styles, but almost every one of them limits one of four things: coverage, estimation, timing, or calculation method.
Coverage answers which data is included and which is excluded. A volume figure that counts only one venue describes that venue accurately and the market incorrectly. The caveat says so, usually in a clause that begins with excludes or only.
Estimation answers how a figure was produced when no exact answer exists. Average token price, impermanent loss, impermanent loss-adjusted yield, and most on-chain analytics dashboards estimate rather than measure. An estimate is still useful. It is just not a count.
Timing answers when the data was true, as opposed to when it was published. Block time, indexing delay and confirmation lag all separate the moment something happened from the moment a screen shows it. A TVL number that is accurate for yesterday morning is not wrong today; it is a different number.
Calculation method answers the formula. Whether an APR is simple or compounded, whether fees are gross or net of gas, whether returns are annualised from a single day of data, and whether a value is denominated in USD or in ETH.
Where caveats hide in plain sight
The interface rarely labels these as caveats. They appear in places that do not look like warnings.
A footnote under a yield figure. A help article reachable only from an information icon. A glossary definition. A line of fine print in the documentation rather than in the product. A sentence in a governance forum post.
The practical rule is to look in three places before you rely on any DeFi figure: the tooltip on the figure itself, the methodology or docs page behind the word about or learn more, and the dated announcement where the metric was last changed.
Reading a gas figure’s caveat
Fee data is the clearest example because the method is documented in detail. Ethereum gas is measured in units of gas multiplied by a cost per unit, and the total splits into two components: a base fee set by the protocol, and a priority fee, the tip, set by the user to make the transaction attractive to validators.
The caveat that matters is about how long a quote holds. The base fee is recalculated from previous blocks, and it moves up to 12.5 percent per block when blocks are over target. A fee quoted as a number is therefore a claim about a moment, not a property.
This is also why a max fee parameter exists and why you are refunded the difference. You set a ceiling, the network charges base fee plus tip, and the unused ceiling comes back. A displayed fee that looks like the amount you will pay is reading it wrong: it is a ceiling you set, and the amount you pay is determined when the transaction is included.
One more caveat on fees: the fee is paid whether the transaction succeeds or fails. A failed transaction that burned gas still cost money. Any interface that reports only successful swaps understates your real cost.
Reading a price or output figure’s caveat
An output figure is a quote, not a fact about a price. The interface calculates the result it can obtain if the transaction executes immediately under current pool conditions.
Slippage is the gap between the quote and the execution price, and it exists for structural reasons. On an automated market maker the trade itself moves the pool price, so the larger the trade relative to the pool, the larger the price impact. On any network the market can move between submission and confirmation. Searchers can also front-run or sandwich a large order, which widens the gap deliberately.
The caveat you will usually find states the default tolerance. The Uniswap interface has historically defaulted to 0.5 percent, with the setting adjustable. Read that number as a decision you have already made, because the interface made it for you.
A high tolerance means the trade is more likely to succeed and more likely to settle at a worse price. A low tolerance means the opposite. Both defaults are defensible. Neither is a fact about the market.
When the caveat is the whole answer
Some caveats do not adjust a figure; they void it.
A token whose contract contains a variable buy or sell fee means the output estimate is conditional on a value the contract can change. A token with a honeypot function means the displayed sell capability is not a prediction of anything, because the transfer logic is written by someone other than the market. An interface that flags a transaction as expected to fail is applying exactly this kind of caveat before you sign.
The habit worth building is small. When a figure appears, ask what it excludes, what it estimates, when it was true, and how it is computed. Four questions take seconds. Skipping them takes the rest of your position.
The short version
Precision on a screen comes from formatting, not from accuracy. The caveat is the only place the method is stated, and the method decides what the number means.
Read it before you read the figure. If the caveat says the number covers one venue, one moment, one formula, or one estimate, then the number is a fact about that and not about the market you are trading.