The Power of Automated Decision-Making in DeFi Risk Management Processes
Intro
Last week’s events, with SVB insolvency and the depegging of USDC which affected lending markets across the board, have shown once again the need for more automated, and transparent, economic risk management processes in DeFi. The rapid changes in the crypto markets make it hard for governance-savvy processes to manage the risk exposure of users in real-time and in an effective manner.
Transiting from opaque, manual risk management into an automated process that is based on quantifiable and verifiable real-time data streams, can offer a more robust framework for DeFi lending markets to manage their platforms’ risk.
In this post, we will focus on the automation aspect of Risk Oracle, and the way it helps mitigate human intervention in the risk management processes of DeFi lending platforms.
Current state
Economic risk management in DeFi lending markets is consisted mainly of setting the right risk parameters such as collateral factors (aka LTV — Loan to Value) and liquidation thresholds of assets. Other risk management activities include setting borrowing and lending caps, effective interest curves for utilization rates, and ultimately the listing, and de-listing, of assets according to their risk profiles.
Today these processes are done manually, according to recommendations coming either from the team behind the platform or by a 3rd party hired by them, requiring the approval of the DAO through a governance vote of token holders. As crypto markets are very dynamic, the changes required to maintain the risk levels in place are quite often. During 2022, both Aave and Compound, the leading lending DeFi platforms, have made dozens of governance votes to set or adjust these risk parameters according to the changes in the market, which are mainly driven by changes in liquidity or volatility of assets, or by the depegging of a stablecoin.
Though DeFi can harness the transparency nature of blockchains to offer much tighter risk management than SVB or Credit Suisse can, last week’s events have shown also how hard a task is for DAO governance processes to keep track of the rapid changes in crypto market conditions. As USDC plunged below 88 cents and re-gained its peg within roughly 48 hours, risk parameters had to be adjusted and re-adjusted frequently according to market reactions. But as each of these changes goes through a forum post, community discussions, on-chain vote, and time-locks before the vote is executed (assuming the vote reached quorum) — platforms couldn’t keep up with the pace of the market, exposing its users to potentially excessive risks than intended by its gatekeepers.
Automation of risk management
What if the decision-making process of adjusting these risk parameters could have been automated to keep track of market dynamics to give safer grounds to lenders and borrowers while minimizing the human intervention required in the process?
There are two separate issues with managing risk in DeFi:
The framework that is used to assess the risk and recommend the right parameters.
The processes which are required to execute these changes.
We’ll focus on the first part in another post, though it’s worth mentioning that most risk frameworks used today in DeFi are not open-source which makes it hard to verify the calculations that were made for the given recommendations.
As for the process — a way to make more hands-off adjustments of risk parameters, e.g. setting the collateral factor and liquidation threshold of an asset, is to base it on on-chain feeds of risk data such as the liquidity and volatility of assets.
By providing real-time, verifiable, streams of an asset’s liquidity available on-chain, a smart contract can automatically adjust the CF and LT in a gradual way in response to the current market condition. It would still need a template to “crunch” the data into specific risk parameters, but once this is available, a DAO can agree in advance on the template and its methodology and set the right parameters according to the liquidity and volatility feeds of an asset — alleviating most of the governance processes required to adjust its risk parameters.
Risk Oracle is setting the ground to do exactly this by providing on-chain feeds of liquidity and volatility of assets, that DAOs and risk teams can use to embed into their platform’s smart contracts to make their economic risk management process more automated and less governance-heavy.
Keeping the process automated, besides being more cost-effective, dictates a pre-defined and transparent set of rules. The automatic execution of this rulebook guarantees the decision-making process to be executed as intended, regardless of the “politics” around specific situations the platform may encounter due to changes in market conditions.
USDC Depegging Case Study
The abrupt depegging of USDC after the insolvency of SVB forced the Aave DAO, like all other lending platforms, to take some fast decisions and adjust some of its platform’s risk parameters to keep users on the safe side and minimize the accumulation of bad debt.
Aave has a few parties managing its risk and the initial recommendations came out fast, suggesting to pause the platform due to the new market conditions. Others have suggested different proposals, and it took a few hours of online discussions to decide on one course of action. You can track the timeline of events here.
Had Aave used a risk oracle, the recommendations coming out of the initial simulation would have been self-executed immediately, both on pausing the markets, and re-activating them once market conditions recovered and the risk framework used to assess the risk would approve it.
MakerDAO governance process was also lagging behind the pace of events during the USDC depegging event, with some risk-preventing votes being executed after USDC had already re-gained its peg due to timelocks and governance process time.
Challenges
Just like professional market makers might perform better than Uniswap, the Automated Market Maker option enabled a deterministic set of rules (or algorithms) to be run by a self-executing smart contract — minimizing, and eventually eliminating, human intervention in the process, for better or worse.
Risk Oracle takes the same perspective toward DeFi lending protocols, where risk management is a core aspect of the success of the platform — and the safety of its users. Though there might be cases where human intervention for decision-making will still be required, the Pareto principle is probably true here as well — automation can eliminate 80% of the governance processes that are required for risk management in DAOs, and over time it can grow gradually to cover nearly 100%. In the meanwhile, the automated rulebook can include the conditions where the DAO should get involved and make decisions manually.
Conclusion
Automating the risk management of DeFi lending platforms using a risk oracle can add transparency to the decision-making process while verifying it will be self-executed in real time as intended. By minimizing human intervention that might withhold “pressing the button” due to political outcomes of the pre-determined required action, DAOs can make sure their platform and their users are being guarded without favoring any specific stakeholder.
If we envision an on-chain, self-executing future, we should thrive to build the solutions that can get us there. A Risk Oracle for DeFi lending protocols is taking us a step closer to that goal.
Last updated