DeFi liquidations can look brutal from the outside: collateral disappears, bots make money, and no one asks for permission. But behind that seemingly ruthless process is one of the mechanisms that keeps decentralized lending markets solvent.
You go to sleep with an ETH-backed loan that looks healthy.
By morning, ETH has fallen sharply. Part of your collateral is gone. Your debt is smaller too, although you never repaid it yourself. Somewhere on Ethereum, a wallet you have never interacted with made money from the transaction.
No bank called you. No risk officer approved the sale. No customer-service representative pressed a liquidation button.
The protocol simply did what it was designed to do.
That process is a DeFi liquidation.
At its simplest, liquidation occurs when the value supporting a collateralized loan falls below the safety level required by a lending protocol. Once that threshold is crossed, another participant can repay some or all of the borrower’s debt and receive collateral in return, typically with an economic incentive.
The mechanics sound straightforward. The reality is more interesting.
A liquidation involves collateral rules, price oracles, smart contracts, automated bots, network competition and carefully calibrated incentives. Understanding how those pieces connect explains not only why borrowers sometimes lose collateral, but also how lending protocols survive violent markets without a traditional bank sitting in the middle.
A DeFi Loan Does Not Wait for a Monthly Payment
Traditional loans usually revolve around time.
A mortgage has monthly payments. A credit card has a billing cycle. A bank may consider whether a borrower has missed payments before taking action.
Overcollateralized DeFi lending works differently.
Consider a simplified example. A user deposits US$10,000 worth of ETH into a lending protocol and borrows US$6,000 of a stablecoin against it.
The protocol does not primarily care whether the loan has existed for six minutes or six months. What matters is whether the collateral continues to provide enough protection against the outstanding debt.
That protection is defined through risk parameters.
One is loan-to-value, or LTV, which determines how much can be borrowed against an asset at the beginning. Another is the liquidation threshold, which determines when a position becomes eligible for liquidation.
Those two numbers should not be confused.
A protocol may allow a borrower to open a loan at one LTV while keeping a higher liquidation threshold as a buffer. That gap is essentially breathing room between “how much you are allowed to borrow” and “how far the position can deteriorate before the protocol intervenes.”
On Aave, these parameters are set at the asset or market level and can be adjusted through the protocol’s risk-management and governance processes.
This is why an overcollateralized loan can be perfectly healthy one minute and liquidatable later without the borrower doing anything.
The debt may not have changed much.
The collateral did.
The Number That Matters Is the Health Factor
Aave makes the relationship easier to see through a metric called the health factor.
In simplified form:
Health Factor = Collateral Value × Liquidation Threshold / Borrowed Value
When the health factor is comfortably above 1, the position has room before liquidation.
When it falls below 1, the position becomes eligible to be liquidated.
Aave’s official documentation on health factors and liquidations gives a useful example: US$10,000 of collateral with an 80% liquidation threshold supporting US$6,000 of debt results in a health factor of roughly 1.33.
Now follow what happens if the collateral falls.
Assume:
- ETH collateral value: US$10,000
- Debt: US$6,000
- Liquidation threshold: 80%
At the beginning:
US$10,000 × 80% / US$6,000 = 1.33
There is a meaningful buffer.
If ETH falls and the collateral is now worth US$8,000:
US$8,000 × 80% / US$6,000 = 1.07
Still healthy, but much closer.
If the collateral falls to US$7,500:
US$7,500 × 80% / US$6,000 = 1.00
The buffer has effectively disappeared.
If it drops to US$7,200:
US$7,200 × 80% / US$6,000 = 0.96
Now the position can be liquidated.
Nothing about the blockchain “decided” the borrower was irresponsible.
The math simply crossed a boundary that had already been encoded into the lending system.
There is also no single health factor that is universally comfortable. A position backed by highly volatile collateral can move toward liquidation much faster than one involving closely correlated assets. Aave itself cautions that an appropriate safety margin depends on the volatility and correlation of the assets involved.
That makes health factors less like a credit score and more like a continuously moving risk gauge.
The Oracle Does Not Liquidate You
At this point another question appears.
How does a smart contract know that ETH has fallen?
Blockchains are deterministic systems. Ethereum does not naturally know the dollar price of ETH, BTC or any other external asset.
That information has to arrive from somewhere.
This is where price oracles come in.
Oracle networks such as Chainlink provide data that smart contracts can use when valuing collateral and debt. Chainlink’s Data Feeds documentation, for example, describes how market information can be aggregated from multiple data providers and independent oracle nodes before being made available on-chain.
But saying that “the oracle liquidated the borrower” would be misleading.
The oracle’s job is primarily to provide information.
The lending protocol determines what that information means.
A simplified sequence looks like this:
Market price changes → oracle reports a new reference price → protocol recalculates collateral value → health factor falls below the threshold → position becomes liquidatable.
The distinction matters.
Imagine a thermometer in a hospital. It can report that a patient’s temperature has reached a certain level, but the thermometer itself does not administer treatment.
Similarly, an oracle can provide the price that causes a lending position to cross its risk boundary, but liquidation rules live in the protocol’s smart contracts.
This separation also explains why oracle design is so important to DeFi lending.
If a protocol acts on unreliable or manipulable pricing, a bad price does not remain merely informational. It can affect whether loans appear healthy or insolvent.
That is why oracle architecture is part of lending-market risk, not an accessory attached after the fact.
So Who Actually Liquidates the Loan?
Here is where DeFi becomes very different from a bank.
Once a position becomes eligible for liquidation, there usually is not an employee at the protocol waiting to process it.
Liquidations are permissionless.
On Aave, any network participant can initiate a valid liquidation. In practice, however, the job is highly competitive and often performed by sophisticated automated systems rather than by someone manually clicking a button.
These participants are commonly called liquidators.
A liquidator monitors lending positions and looks for accounts whose health factors have moved below the required threshold.
When one appears, the liquidator can submit a transaction that repays part of the borrower’s debt.
In return, the liquidator receives some of the borrower’s collateral, including an incentive commonly described as a liquidation bonus.
That bonus exists for a reason.
Liquidating a position costs money and carries execution risk. The liquidator may need capital, blockchain infrastructure, gas fees and systems capable of reacting faster than competitors.
Without an economic reward, there would be little reason for independent participants to perform the service.
The clever part of the design is that protocols do not need to hire these liquidators.
They create an incentive, publish the rules through smart contracts and let independent actors compete to capture it.
The borrower would understandably see the liquidation bonus as a penalty.
The protocol sees it as payment for an essential risk-management function.
The liquidator sees it as an opportunity.
All three interpretations can be true at once.
Inside One Liquidation, Step by Step
Return to the borrower whose ETH-backed position has fallen to a health factor of 0.96.
What actually happens next?
1. The market moves
ETH falls, reducing the market value of the borrower’s collateral.
The debt, denominated in a stablecoin, remains roughly unchanged in dollar terms.
2. New price information reaches the protocol
The protocol receives updated pricing through its oracle infrastructure.
With the lower ETH value, the borrower’s position is recalculated.
3. The health factor crosses below 1
The smart contract can now recognize the account as eligible for liquidation under its predefined rules.
This eligibility does not itself move the collateral.
It creates an opportunity for a liquidator to act.
4. A liquidation bot detects the position
Bots continuously monitor lending markets for liquidatable accounts.
The bot evaluates whether executing the liquidation makes economic sense after accounting for the collateral it can receive, the debt it must repay, transaction costs and competition from other liquidators.
5. The liquidator submits a transaction
The transaction calls the relevant liquidation function in the lending protocol.
The liquidator specifies, among other things, which borrower’s position it wants to liquidate and which debt and collateral assets are involved.
6. The protocol checks the rules
The smart contract does not trust the liquidator’s claim that the borrower should be liquidated.
It verifies the position.
If the account is ineligible, the transaction cannot seize collateral just because someone requested it.
If the conditions are satisfied, execution can continue.
7. Debt is repaid and collateral changes hands
The liquidator repays an allowed portion of the borrower’s debt.
Corresponding collateral is transferred from the borrower’s position according to the protocol’s rules, with the liquidation incentive incorporated.
The borrower emerges with less debt but also less collateral.
8. The position returns toward safety
In a partial liquidation, removing debt can push the remaining position back toward a healthier state.
The system has effectively reduced leverage before the collateral becomes insufficient to cover what lenders are owed.
That last point is the core purpose of the entire mechanism.
Liquidation is not primarily designed to punish the borrower.
It is designed to stop an undercollateralized loan from becoming the lending pool’s problem.
Why Doesn’t the Protocol Sell Everything?
Not every liquidation necessarily means losing an entire position.
Aave V3, for example, uses rules governing how much debt can be repaid during liquidation.
Under its documented mechanics, up to 50% of a position’s total debt can generally be liquidated when the health factor is above 0.95 but below 1, provided both collateral and debt meet the relevant minimum size thresholds. Under more severe conditions, including a health factor of 0.95 or below, as much as 100% may become liquidatable.
The distinction reflects a trade-off.
Liquidating too little could leave the protocol exposed to further losses.
Liquidating too much can unnecessarily punish the borrower.
That trade-off is important enough that newer lending designs are experimenting with more granular approaches.
Aave V4, for example, redesigns the mechanism around a target health factor rather than relying solely on a fixed close factor. In Aave’s explanation of the V4 liquidation engine, the amount a liquidator can repay is generally calculated to restore a borrower toward a predefined target health factor.
The design also introduces a variable liquidation bonus that increases as a position becomes riskier.
That change exposes something easy to miss:
Liquidation design is not finished technology.
Protocols are still trying to answer a difficult question: How do you protect lenders quickly without taking more collateral from borrowers than necessary?
Liquidation Bots Are Competing With Each Other
It is tempting to imagine liquidation as a simple automated function: a position crosses a line and a bot executes it.
But valuable liquidations attract competition.
Several bots may identify the same unhealthy position at nearly the same time.
Only successful execution earns the opportunity.
That changes the economics considerably.
A liquidator needs to ask:
- Is the liquidation bonus large enough?
- How much gas will the transaction consume?
- How much capital is required?
- Can the collateral be sold or hedged?
- Will another bot arrive there first?
- Could the transaction fail after fees or infrastructure costs have already been committed?
Aave describes its liquidation market as highly competitive and notes that successful operators often rely on custom bots or scripts capable of monitoring positions and responding quickly to price changes.
This is why calling liquidators “people who are discounted collateral” misses half the picture.
They are participants in a competitive execution market.
The discount is the revenue side.
Infrastructure, capital and failed competition are part of the cost side.
Ravel On-Chain Case Study: Where the Money Went in a Real Aave Liquidation
The economics become clearer when examining an actual liquidation rather than a hypothetical position.
Ravel examined an Ethereum transaction executed on November 21, 2025, involving Aave V3 and a borrower whose WETH collateral was liquidated against a USDT debt.
The transaction’s Etherscan record and decoded LiquidationCall event provide the underlying evidence.
The liquidation event recorded two important amounts:
- Debt covered: 106,425.797094 USDT
- Collateral liquidated: 41.7008437097 WETH
The transaction also shows what happened during the execution.
A liquidator-associated contract swapped approximately 40.2626969056 WETH through Uniswap V3, receiving 107,188.449024 USDT.
The swap proceeds exceeded the debt amount recorded in the liquidation event by approximately 762.65 USDT.
However, that difference cannot simply be treated as the operator’s net profit. Funding arrangements, trading costs and other transaction flows must be considered before determining the final economic return.
The more revealing detail concerns the WETH that remained after the swap.
Subtracting the amount swapped from the collateral recorded in the liquidation event gives:
41.7008437097 − 40.2626969056 = 1.4381468041 WETH
The transaction trace shows an equivalent amount being unwrapped and distributed as ETH.
Of that amount:
- 1.0142415918 ETH went to beaverbuild, the block builder identified in the transaction.
- 0.4239052123 ETH went to the address that submitted the transaction.
In other words, roughly 70.5% of the observed ETH distribution went to the builder, while approximately 29.5% went to the transaction-submitting address.
That calculation is based on the two recorded ETH transfers, not on an estimate of total liquidation profit.
The trace does not establish every participant’s final profit after funding expenses and other costs. Nor should this single transaction be treated as representative of every Aave liquidation.
Still, it reveals something that a simplified description of liquidation bonuses can miss.
The collateral received from a distressed borrower does not necessarily translate into a reward retained entirely by the liquidator.
Execution competition can create another layer of payments, including incentives for transaction inclusion.
In this case, the on-chain record shows a large payment to the block builder as part of executing a liquidation.
That makes the economics more complicated than a borrower losing collateral and a bot collecting a discount.
The liquidation incentive must support an entire execution process involving capital, market liquidity, infrastructure and competition for block inclusion.
This is one example, not evidence of a universal profit-sharing pattern.
But it provides a concrete illustration of how a permissionless liquidation can create financial flows extending beyond the borrower and the lending protocol.
When Liquidation Works, Almost Nothing Dramatic Happens
The best liquidation systems are strangely uneventful.
A collateral price falls.
Risky positions get reduced.
Liquidators earn an incentive.
Lenders remain protected against some of the resulting credit risk.
The protocol keeps operating.
That is success.
The problem appears when markets move faster than the mechanism designed to protect them.
Imagine collateral collapsing so quickly that its value approaches the outstanding debt before liquidators can react.
Or network congestion makes execution expensive.
Or liquidity for the seized collateral disappears.
Or pricing behaves unexpectedly during market stress.
Eventually, a position can move from “unsafe but liquidatable” to distressed debt territory, where the collateral is no longer sufficient to cover the obligation.
This is what lending protocols are trying to prevent.
Liquidation thresholds therefore cannot be considered in isolation. They interact with asset volatility, liquidity depth, oracle design, borrower concentration and the willingness of liquidators to act.
A token that looks perfectly acceptable as collateral during calm markets may behave very differently when everyone tries to sell it simultaneously.
That is also why protocols frequently apply different risk parameters to different assets rather than treating one dollar of every token as economically equivalent collateral.
The Most Dangerous Liquidation May Be the One Everyone Shares
Individual borrowers often think about liquidation at the wallet level.
Protocols have to think about it at the system level.
Suppose hundreds of borrowers use the same volatile asset as collateral.
The price begins falling.
Their health factors move toward 1 together.
Liquidations start.
Liquidators receive collateral and may sell it to recover the capital used to repay debt.
That selling can create additional market pressure.
More positions approach their thresholds.
More collateral becomes available for liquidation.
This is the basic anatomy of a liquidation cascade.
It does not mean every market crash automatically becomes a death spiral. Deep liquidity, conservative risk parameters, diversified collateral and efficient liquidation infrastructure can absorb significant stress.
But it shows why borrowing against volatile crypto assets creates a feedback loop that does not exist in a simple spot portfolio.
If you own ETH and it falls 30%, you still own the ETH.
If you use ETH to support leveraged borrowing, the same 30% decline may trigger an irreversible smart-contract process that changes your position before the market has a chance to recover.
Leverage turns volatility into a deadline.
Liquidation Is a Feature, Not a Bug
For borrowers, liquidation is usually experienced as failure.
For lending protocols, the ability to liquidate is part of what makes permissionless borrowing possible in the first place.
Aave does not know whether a wallet belongs to a billionaire, a university student, a hedge fund or another smart contract.
There is no salary slip.
No credit bureau.
No phone call to an employer.
The system therefore cannot rely on traditional borrower identity and credit enforcement.
Instead, it relies heavily on collateral.
If anyone can borrow without asking permission, the protocol needs a mechanism that can also protect lenders without waiting for permission when collateral becomes insufficient.
Liquidation is that mechanism.
This is the uncomfortable bargain at the heart of overcollateralized DeFi lending:
The protocol does not need to know who you are, it already knows what it can do with your collateral.
That sounds ruthless.
It is also what allows lending pools containing capital from strangers around the world to operate without those strangers having to trust each individual borrower.
What Borrowers Can Actually Control
Liquidation risk can never be eliminated from a leveraged collateralized position, but it can be managed.
The first defense is obvious: do not borrow to the maximum available limit simply because the interface allows it.
Available borrowing capacity and prudent borrowing capacity are not the same thing.
A larger buffer between the current health factor and the liquidation threshold provides the collateral with the collateral more room to absorb volatility.
Borrowers can also improve a deteriorating position by adding collateral or repaying debt. Aave explicitly identifies both as ways to increase the health factor.
Asset selection matters too.
Borrowing a stable asset against highly volatile collateral exposes the position directly to a decline in the collateral’s dollar value. Other structures may involve assets whose prices are more closely correlated, producing different risk dynamics.
And monitoring matters.
A position with a health factor barely above 1 is not “safe” merely because it has not yet crossed the line.
It is a position waiting for the market’s next move.
What Really Happens When Your DeFi Loan Gets Liquidated?
Strip away the terminology, and the sequence is surprisingly clean.
A borrower posts collateral.
The protocol assigns risk parameters.
An oracle helps value the assets.
The market moves.
The position crosses its allowed boundary.
An independent liquidator repays debt.
The protocol releases collateral according to predefined rules.
No debt collector arrives.
No manager makes an exception.
No one needs to know the borrower’s name.
The smart contract only needs to know if the numbers still work.
That is what makes DeFi liquidations both unsettling and important. They expose the part of decentralized finance that interfaces rarely emphasize: underneath the smooth dashboards and one-click borrowing flows is an automated risk engine built to protect the pool before it protects the borrower.
When markets are calm, that machinery stays mostly invisible.
When prices collapse, it becomes the system.
