Benddao Improvement: Segmenting of lending into different pools by risk


The recent NFT bear market has exposed major fault lines in the Bend protocol. Loans taken by holders during the previous market peak are now at risk of impairment. The fear of cascading liquidations have set off a chain of events that have submerged the protocol into a vicious downward spiral:

  1. The fear of cascading liquidations and loan impairments have caused lenders to flee the protocol en mass, creating a liquidity crunch that has limited withdrawals and additional borrowing. Protocol APRs have spiked to incentivize additional deposits but fears of a bank run have largely kept would-be depositors on the sidelines.
  2. As result of the spike in APRs, borrowers woke up over the weekend to untenable triple digit interest rates. This kicked off a subsequent round of repayments by credit worthy borrowers with overcollateralized loans, leaving the pool with a great proportion of loans where the borrowers either have no incentive (loans are already underwater) or means to repay.
  3. The combination of higher rates and riskier loans accelerates liquidations, putting further stress on the protocol and increasing the risk that remaining depositors see impairment from bad loans

While the core team and community at Bend have already put forward emergency proposals to stabilize the protocol, I believe structural issues must be addressed to ensure the protocol’s success in the long term.


The core issue with the existing Bend lending model is its inability to disaggregate and price different levels of risk. Under the current platform, a loan made at the maximum 40% LTV pays the same interest rate as one made at 10% LTV, even though the former is inherently more risky than the latter. This results in a classic moral hazard where degens are more inclined to borrow aggressively because part of their risk is being subsidized by more responsible borrowers in the pool. When a small number of these loans go bad, they taint the entire pool of loans and the resulting contagion causes an exodus of lenders putting Bend in its current predicament. To ensure that such events do not happen again in the future, it is vital that different levels of risk are appropriately isolated and priced to ensure more fairness and transparency.

Proposed Solution:

To address this issue, I propose that future lending / borrowing pools be separated into four pools based on their level of risk with different levels of interest rates (still subject to the same/similar formula that Bend uses to incentivize deposits):

  1. Pool A: For loans up to 10% LTV - this the lowest risk pool with a base APR of 7.5%
  2. Pool B: For loan amounts between 10 to 20% LTV - this is considered a moderate risk pool with a base borrow APR of 15%
  3. Pool C: For loan amounts between 20 to 30% LTV - this is considered a higher risk pool with a base borrow rate of 25%
  4. Pool D: For loan amounts between 30 to 40% LTV - this is the highest risk of pool for the most degen borrowers/lenders with a base borrow rate of 35%

For borrowers: Borrowing on Bend effectively becomes an act of borrowing from up to 4 pools depending on the amount being sought. A borrower looking for a 10% LTV loan will borrow exclusively from Pool A (while paying the lowest interest rate) while someone looking for the maximum borrowable amount will borrow from all 4 pools.

ie. Borrower A and B both have an BAYC that they want to take out a loan on. Assuming a floor price of 70 eth the maximum amount borrowable under Bend would be 28eth. Borrower A elects to borrow 14 eth - this results in him borrowing 7eth from Pool A at 7.5% and 7 eth from pool B at 15% for an effective interest rate of 11.25% on the total 14eth loan. Borrower B elects to borrow the maximum 28eth which entails 7 eth loans across all 4 pools resulting in a 28 eth loan with an effective interest rate of 20.625%. In this case, there is a clear distinction between the two loans and the rates charged to each based on their level of risk


For lenders: Under this framework, lenders can allocate their eth deposits based on their appropriate level of risk appetite. Lenders seeking a secure, low risk rate of return can opt to allocate exclusively to Pools and A and B while those with greater risk appetite can lend into Pools C-D to earn higher yields at risk of impairment.

Why does this matter?

  1. In a crisis such as the one currently unfolding in front of us - this structure should limit mass withdrawals that would kickstart a bank run:
    1. Lenders in Pools A and B have knowledge that they have a significant safety cushion on their deposits and remain incentivized to keep funds in the pool which ensures continuity on borrowing activity (vs the current situation where the spike in rates and lack of liquidity have pushed borrowing to a standstill)
    2. Lenders in Pools C and D will withdraw capital and potentially suffer losses, but there is a much better understanding of the potential risks involved from the outset
    3. For borrowers - this provides a much more effective way to manage their loans. In the event of a bear market where borrow rates on the riskier pools spike, credit worthy borrowers can opt to pay down the riskier portions of their loan and bring their interest rate down to more manageable levels vs withdrawing their otherwise good loans from the platform altogether
  2. Longer term, this model allows depositors to set the protocols appetite for different levels of loans
    1. in NFT bull markets, yield seeking depositors may allocate more eth to the riskier pools and enable borrowers to borrow at 20-40% LTVs with low interest rates
    2. In NFT bears, depositors become more risk averse and reduce their exposure to these pools, this reduces the amount of high LTV borrowing that can be done, proactively limiting the risk of incurring further bad loans

If you guys have made it this far, thanks for taking the time to read this and I’m looking forward to getting everyone’s feedback and furthering the discussion.


I think I can see what you mean.
Different rates are given based on different Health Factor.
High HF, low rate; low HF, high rate.
Depositors choose the pool they want to deposit in, based on their risk appetite.
Very good idea.

This can lead to the following problems.
1, because the health value of collateral changes dynamically, this can lead to frequent switching of collateral between different pools, which may result in a very large calculation of the interest rate by the agreement.
2, Different pools with different interest rates may reduce the utilization of depositors’ funds. (This is the downside of segregated lending pair agreements, which isolate risk but also reduce the utilization of funds)

This seems to be more difficult in terms of implementation.
Let’s see what others have to say.

Translated with DeepL

This is because the Collateral Rate of the collateral will vary with the floor price.
If we only calculate the Collateral Rate at the time the borrowing occurs and calculate the interest based on this Collateral Rate. Then, there seems to be some problem that a collateral with low Collateral Rate will also become a collateral with high Collateral Rate when the floor price falls, when the risk has gone up, but he still have a low risk interest rate.

1 Like

I think long term BendDAO will go to something like this, which aave v3 is going to make.
But for now, people even concerns about some of the bluechips going to zero.
There is not much NFT bluechips now, so maybe it’s not a good timing?

1 Like

This is a great point. one of the biggest challenges from an implementation perspective will be what happens when the FP of the underlying collateral changes. I don’t think its practical to move funds across different lending pools as collateral values change but there could be two other ways to resolve this:

  1. Do nothing - everyone from each respective lending pool simply owns the risk
  2. Keep the funds in the same pools, but charge higher interest rates for the portions of the loan that are now at a higher risk category to compensate lenders for the elevated risk.

For example: Frank borrows 40eth against his BAYC when the FP is 100 eth and pays the composite interest rate of 20.625% as he has 10 eth loans across all 4 pools. When the BAYC FP falls to 80eth his effective new loan capacity falls to 32 eth, so he can now only borrow 8eth across each of the four pools. While his loans aren’t reshuffled across different pools to rebalance for the change in risk his interest rates are adjusted to compensate existing lenders for the increased risk profile. In this case - it would result in him paying a 7.5% base rate on first 8eth of the loan, 15% on the second 8eth, 25% on the third 8eth tranche and 35% on the last 16eth outstanding. This would increase his effective interest rate from 20.625% to 23.5% and potentially much more if depositors from pool D begin pulling out funds due to market conditions. For Frank’s lenders, Pool A would earn 7.5% on 8eth of his loan and 15% on the 2eth that he still owns him in the pool with similar outcomes for his lenders in pools B and C who are also compensated wit higher yield. This creates a organic incentive for Frank to pay down some of his loan to reduce interest (thereby making the overall loan safer) while lenders on the lower risk pools continue to retain a healthy safety cushion while earning a higher rate.

I agree with this - the first step for the DAO should be to stabilize the deposit / borrow base and ensure orderly liquidation of high risk loans. I want to try and start this conversation early because successful implementation of these changes could be very complex and it would be great to leverage the experience and knowledge of our giga brained community early in the process.

Yes, I agree with you, maybe we can start this work by open an issue on github, to discuss more there with programmers? Issues · BendDAO/bend-lending-protocol · GitHub or just keep working on this post?

1 Like