The phrase is a marketing word.
Impermanent loss is one of the most confusing terms in DeFi, and most of the confusion is the name itself. It's not impermanent. It's not really loss. It's just the price exposure you take on when you provide liquidity instead of holding the two tokens.
The original idea was that if you deposit into a pool and the price returns to where you started, the loss disappears. So it's impermanent compared to the moment of deposit. That framing is technically true and practically useless. Real markets don't return to starting prices on a schedule.
What's actually happening.
When you provide liquidity to a SOL/USDC pool, you're agreeing to a rule. As SOL goes up, the pool sells your SOL into USDC. As SOL goes down, the pool buys SOL with your USDC. The pool is a slow market maker, automatically rebalancing.
The result is that you end up with less SOL when SOL has gone up, and more SOL when SOL has gone down. Compare that to just holding fifty percent SOL and fifty percent USDC and not touching it. The held portfolio always wins on a directional move, because it doesn't sell into strength or buy into weakness.
That gap, between what you'd have if you held versus what you have inside the pool, is impermanent loss. It is real money. It does not magically disappear unless the price exactly retraces.
You are long the move down, short the move up.
This is the simplest way to think about it. As an LP in a SOL/USDC pool, you take more downside than holding fifty-fifty would, and you take less upside than holding fifty-fifty would. The pool is selling you a kind of mean-reversion bet whether you wanted one or not.
If you believe the price will chop sideways, that bet is good. The pool earns fees and your impermanent loss stays small. If you believe the price will run hard one direction, that bet is bad. The fees will not catch up to the directional drag.
When fees outweigh the drag.
On a calm pair with a lot of swap volume, fees can absolutely beat impermanent loss. SOL/USDC during a quiet week is the classic example. The pool grinds through volume, you collect fees on a wide enough range to stay active, and the net is positive.
On a volatile pair during a trend, fees rarely catch up. Concentrated liquidity makes the fee math better, but it also makes the impermanent loss worse on a per-dollar basis, because you're more exposed to local price moves. Tighter ranges earn more fees, but they also feel the move harder when it comes.
There is no general rule. There is only this question, asked over and over: is the fee rate at this concentration level high enough to compensate for the price drag at this volatility level. That's literally what an agent computes.
Why Lyra tracks inventory PnL, not just fees.
A common trick in LP marketing is to publish only the fee yield. Fees on a tight pair can look spectacular. Twenty, fifty, one hundred annualized percent is not unusual on a hot pool.
But fee yield without inventory PnL is a half-truth. If you collected fees worth ten percent of your position but the inventory drift cost you fifteen percent, you lost five percent. That's not yield. That's a loss with a side dish of fees.
Lyra's vault PnL is total: fees collected, plus harvests, plus the actual settlement value of the LP inventory at the time of close. We don't separate them in your top-line number. The split is in the activity feed if you want to see it.
Honest framing.
LPs can lose money. Even with a good agent, even with a positive fee yield, a hard one-sided move on the underlying pair can leave a position worth less than the same dollars held outside the pool. This is not a bug. It is the asset class.
The job of an agent is to pick pools and ranges where, on average, fees plus reshape discipline beat the inventory drag. The job of you, the depositor, is to know that you're taking price exposure to the pairs your agents are LPing on. If you wouldn't be comfortable holding a basket of those tokens in the first place, an LP vault on those pairs is not for you.