APR vs APY vs Real Yield: Reading a Rate Honestly
The number on the dashboard is annualised from a recent window, before fees, and often paid in a token whose price is falling. Here is how to get to the real one.
The number on the dashboard is annualised from a recent window, before fees, and often paid in a token whose price is falling. Here is how to get to the real one.
APR is the simple annual rate; APY assumes rewards are compounded and is therefore always the higher number for the same underlying yield. Neither tells you what you earn: displayed rates usually include token emissions valued at the moment of harvest, exclude fees and gas, and annualise a short recent window. Real yield is what remains after selling the reward token, paying fees, and stripping out incentives that end.
A vault advertising 42% APY and a lending market advertising 4.1% APR can be paying the same thing, and often the 4.1% is the better deal. The gap between an advertised rate and a realised return in DeFi is wider than in any other part of finance, and it is entirely explicable once you know what each number leaves out.
APR is the periodic rate multiplied out to a year with no compounding. Earn 1% a month and the APR is 12%.
APY assumes you reinvest each period. That same 1% a month compounds to 12.68%. The more frequent the assumed compounding, the larger the gap: at an underlying 100% APR compounded daily, the APY is over 170%. Nothing extra is being earned — the number simply assumes a behaviour.
So the first question about any advertised rate is which one it is, and if it is APY, whether the compounding it assumes actually happens. An auto-compounding vault does compound for you, so its APY is achievable net of its fee. A lending market that pays you continuously but requires you to claim and redeposit does not compound unless you do it, and doing it costs gas each time.
Most DeFi rates are the sum of two things: a base yield from actual revenue — interest paid by borrowers, trading fees paid by swappers — and an incentive component paid in the protocol's own token.
The base yield is real and persists. The incentive component is valued at the token's price at the moment of calculation, and you only receive that value if you can sell at that price. In a falling token, the realised yield is materially lower than the display, and in a thin token, selling your own rewards moves the price against you.
This is why the composition matters more than the total. DefiLlama splits base and reward APY on its yield pages, which is the single most useful disclosure in the category and the reason we score its methodology transparency highest among data tools.
Four costs. The protocol's performance and management fees, which come out of gross yield — 2% and 20% at Yearn, around 4.5% of harvests at Beefy. Gas, which on Ethereum mainnet can exceed a year of yield on a small position. Impermanent loss, if the position is a liquidity pair rather than a single asset. And slippage when you convert rewards to something you want to hold.
For a small deposit on an expensive chain, those four routinely turn a positive advertised rate into a negative realised one.
Most dashboards annualise a recent window — often seven days, sometimes twenty-four hours. A pool that had one enormous trading day shows an APY that assumes every day looks like that one. This is why new pools display absurd numbers on day two and settle by week three.
Look for a longer lookback, or check the fee revenue directly and divide it by liquidity yourself. If a rate is only quoted over 24 hours, treat it as marketing.
Work through it in this order. Start with base yield only. Subtract the platform fee. Subtract estimated gas over your intended holding period, divided by your position size. If the position is an LP pair, subtract a realistic impermanent-loss estimate for the volatility of that pair. Only then add the incentive component, discounted for how much you expect to lose selling the reward token — and treat any incentive with a published end date as temporary rather than as yield.
Compare what remains against the risk-free alternative in the same asset: a treasury-backed stablecoin or plain staking. If a DeFi position pays two percentage points over that and carries contract, oracle and liquidation risk on top, the question is whether two points is the right price for those risks — and that is a decision you can only make once the number in front of you is honest.
Revenue-funded yield is more durable than emission-funded yield, and you can check which you are looking at. If liquidity providers are paid from fees that users actually pay — trading fees, borrowing interest — the yield survives the end of an incentive programme. If they are paid in tokens the protocol prints, it does not. That distinction, not the size of the number, is what separates a position you can hold from one you have to time.

A framework for separating sustainable DeFi yield from token-emission dilution before you deploy a single dollar of capital.
The headline fee on an exchange's marketing page is the maker rate at a volume tier you will never reach. Here is what you actually pay.
A perpetual has no expiry, so something has to pull it back to spot. That something is a payment out of your account, several times a day.