A useful DeFi review begins with the position a user will actually hold. Supplying USDC to a lending market, providing concentrated liquidity, and locking collateral to mint a stablecoin expose different state transitions and different ways to lose money. The analyst's job is to reconstruct those mechanics closely enough to explain who owes what, which price or invariant drives action, and which transaction can turn a risk into a realized loss.
That work is closer to investigating a small financial system than completing a universal checklist. Contract addresses, reserve configuration, oracle behavior, liquidation routes, governance powers, token liquidity, and interface assumptions all matter, but their importance depends on the protocol. TVL and audit reports can orient the review; they cannot replace a protocol-specific explanation of withdrawals, bad debt, and control.
What you will learn
- Reconstruct the contracts and state changes behind one real user position
- Identify the protocol-specific price, liquidity, and solvency mechanisms
- Separate borrower or trader payments from incentives and token emissions
- Diagnose a concrete stress path and state which participants absorb the loss
Start with one position, not the homepage
Choose a representative action and follow it from wallet approval to exit. In a lending market, record the asset supplied, receipt token or accounting balance, collateral setting, debt token, interest index, health measure, oracle call, liquidation function, and withdrawal source. For an automated market maker, the equivalent path includes tick range or invariant, fee accrual, routing, slippage protection, and removal of liquidity.
Verify the deployment and active implementation through official registries and block explorers, then read the parameters that govern this particular market. A protocol name may span several chains and isolated pools with different assets, caps, administrators, and liquidity. Note which calls a user signs directly, which calculations the interface performs, and which safeguards live in contracts even if the main website is unavailable.
Find the mechanism that turns price into action
For lending, the decisive chain runs from oracle price to account health to liquidation eligibility; for an AMM, reserve balances or concentrated-liquidity ranges determine the execution curve; for a stablecoin, collateral valuation and redemption rules connect the peg to solvency. Write that chain explicitly. Then identify update cadence, fallbacks, stale-price handling, and any administrator able to replace a feed or change a threshold.
Test the assets inside that mechanism. Issuer freezes, bridge verification failures, rebasing behavior, thin secondary markets, and concentrated ownership can break assumptions even when core code behaves as designed. Estimate whether liquidators or redeemers could execute the stressed amount across available venues after fees and price impact. Dashboard liquidity under ordinary flow is not the same inventory available during a correlated exit.
Follow the cash flow and the shortfall
Separate fees paid by users from token incentives, treasury transfers, and self-referential activity. For lending, connect supplier yield to borrower interest and utilization. For trading, connect provider fees to volume while accounting for divergence and incentives. State who pays each return and whether the rate depends on a token's market price.
Next locate the shortfall account. A lending protocol may route part of borrower interest to reserves and rely on collateral liquidation before suppliers face bad debt; an AMM normally passes inventory divergence directly to liquidity providers; a stablecoin may auction collateral or socialize a deficit through governance. Scenario design should follow that exact waterfall instead of attaching the same list of shocks to every protocol.
Match reviews and admin powers to deployed code
Audits are scoped reviews of particular code versions and assumptions. Read the scope, date, severity definitions, unresolved findings, and whether deployed bytecode corresponds to the reviewed release. Bug bounties, formal verification, tests, monitoring, and simple architecture add different evidence. Multiple audit logos do not combine into a guarantee.
Map owners, multisignature signers, timelocks, upgrade administrators, pausers, guardians, and governance executors. Determine signer independence and operational security where disclosed. A powerful emergency key may limit exploit damage while creating compromise and censorship risk. The correct conclusion describes the trade-off instead of awarding decentralization based on one label.
Read governance when it changes the position
Read proposals that changed the assets, caps, oracle, liquidation settings, rate curve, or privileged roles relevant to the position. The discussion reveals how risk contributors use evidence and how quickly executable changes can pass. Incident records should be read the same way: what state first became abnormal, which role could intervene, where accounting recognized the loss, and who funded any reimbursement.
Finish by attempting the exit path with current contract methods and realistic size. Confirm unborrowed liquidity or redemption inventory, lockups, approval revocation, alternate interfaces, and the chain or bridge needed to receive the underlying asset. The final note should state a trigger, the protocol mechanism it activates, the party exposed, available mitigants, and what remains unknown in plain language.
Common misconceptions
“High total value locked proves that a DeFi protocol is safe and trusted.”
TVL is an asset-counting metric that can include incentives, leverage, repeated receipt tokens, concentrated deposits, or assets that are difficult to exit. It does not prove solvency or security.
“Several audits mean the protocol cannot be exploited.”
Audits examine defined code and assumptions at a point in time. Upgrades, integrations, economic attacks, key compromise, and findings outside scope can still cause loss.
“A mature protocol no longer needs dependency analysis.”
Age provides operating evidence but does not remove evolving governance, assets, bridges, oracles, interfaces, upgrades, and market structure. Dependencies can change after years of operation.
Risks and limitations
- The wrong unit of analysis can hide risk when aggregate protocol statistics obscure one chain, pool, collateral asset, or user position.
- A stale deployment registry, audit, dashboard, or parameter page can describe code and market conditions that no longer govern the position.
- Liquidation can fail economically even when contract logic is correct because keepers lack debt assets, blockspace, or a market for seized collateral.
- Governance can alter caps, price feeds, liquidation settings, rates, or privileged roles faster than a passive position holder notices.
- Exiting may depend on unborrowed reserves, active price ranges, bridge availability, or redemption inventory that disappears during common stress.
Key takeaways
- Begin with one real position and reconstruct the state changes required to open, maintain, liquidate, and exit it.
- Name the protocol-specific rule that converts prices, reserves, or inventory into a financial outcome.
- Trace yield to borrower interest, trader fees, or emissions instead of grouping unlike returns together.
- Test whether liquidators, redeemers, or liquidity providers can transact at the size assumed by the loss model.
- Compare audits and governance descriptions with the active deployment and current parameters.
- Conclude with a readable failure path and loss bearer, not a universal score.
Primary and further reading
Test your understanding
Score at least 2 out of 3 to complete this lesson. Explanations appear after you submit.