Almost every yield figure in DeFi is a gross number, quoted at a moment in time, on a position that may not be liquid when you want to exit. The gap between that number and what you actually keep is not a rounding error — it is usually the entire difference between a strategy worth running and one that is not.
Where the yield actually comes from
There are only four sources, and they carry very different risk.
Trading fees paid by people using a market. This is the most durable source, because it scales with real activity rather than with a token's issuance schedule.
Lending interest, which is the utilisation curve doing its job. Durable, but it compresses the moment capital floods in.
Token emissions, which are a subsidy. High APYs in a new pool are almost always emissions, and emissions are paid in the thing you are not sure you want to hold. An 80% APY on a token falling 5% a week is a losing trade with extra steps.
MEV and liquidations, which are opportunistic and require infrastructure most users do not have.
Telling the difference is mostly a matter of asking what the number would be if the token were worth nothing. If the answer is "close to zero", the yield is a distribution of the token to you.
Emissions decay, and the decay is the point
Farm emissions are designed to run down. The schedule is published, the rate drops, and the reward per unit of liquidity falls as more capital arrives to chase it. Two forces compound: the emission rate declines on a fixed schedule, and the TVL in the pool rises because the APY is advertised. Your share of a shrinking reward pool, across a growing pool of capital, falls faster than either trend suggests.
This is why the honest version of a yield chart is not the current APY but the APY you would receive if you were the last person to enter. By the time a farm is well known enough to be on a dashboard, most of the emission is already spoken for.
Impermanent loss is not a fee and not a penalty. It is the difference between holding two assets and holding a constant-product position in them, measured in units of the better-performing asset.
The mechanism is simple: as one asset rises, the pool sells it for you. You end up with more of the loser and less of the winner, which is exactly the opposite of what a directional holder wants. The loss grows with the size of the price divergence, and it is asymmetric — a 2x move costs more than the equivalent move down, because the pool keeps rebalancing along the way.
The number that matters is the net: fees plus emissions minus impermanent loss minus gas. On a stable pair, impermanent loss is near zero and the trade is about fees. On a volatile pair in a quiet market, fees may not cover the divergence at all.
Concentrated liquidity changes the profile rather than removing it. You earn more fees in a narrow range, and you suffer impermanent loss much faster the moment price leaves that range — plus you are left holding a position that is entirely in the underperforming asset.
The aggregator layer adds its own risk
Vaults exist because farming is operationally tedious: find the pool, deposit, claim, swap, restake, repeat before emissions decay. A vault automates that loop and takes a performance fee.
Yearn-style vaults hold funds in a strategy contract that moves capital between protocols according to a stated rule. Autocompounders such as Beefy take the simpler route of claiming farm rewards and reinvesting on a schedule. The edge is execution frequency and gas efficiency, not secret alpha — when a vault advertises outperformance, the honest question is what it is doing that a user could not do themselves.
The risk you are actually taking sits one layer deeper than the vault's own audit. A vault is a smart contract wrapping another protocol's smart contract, so depositors inherit both. Aggregated strategies have failed on strategy bugs, on oracle mispricing inside the underlying protocol, and on governance decisions in a protocol the vault chose — none of which the vault's audit covers. And withdrawals can be gated by a timelock or by a strategy mid-cycle, which turns a paper APY into an illiquid position at precisely the moment you want out.
How to net it down
Estimate the emission value at a price you would be willing to hold at, not today's price. Subtract the performance fee, the gas to enter and exit, and the expected impermanent loss at the volatility you actually expect rather than the volatility that has been realised in a calm month. Then subtract the probability-weighted cost of a strategy failure.
For most vaults the honest conclusion is that they are a convenience product. They save you the operational work of compounding, and they add tail risk you cannot see from the APY. That is a reasonable trade if you size it accordingly, and a bad one if you treat the headline number as income.
References
Next, read Stablecoin Mechanics 101. More in DeFi Markets.