Offboarding Unprofitable Collaterals

Background

Maker faces difficulties with maintaining vault types that have low usage. While our stability fee revenue directly correlates with the total amount minted against a collateral asset, our primary operating expense of maintaining oracle price feeds does not - rather it is dependent on prevailing ETH and gas prices. This creates a situation where we need a certain minimum amount of DAI minted to operate vault types profitably.

The minimum required usage depends on a variety of factors, including the vault stability fee, ETH price, gas price, and oracle sensitivity (eg does the oracle update price on 1% variation, or less frequently on eg 3% variation in price). So while the exact cutoff for when a vault type becomes unprofitable varies, there is general agreement among the core units that collaterals requiring their own price feed with less than 10 million DAI minted are operated at a loss. In addition, Maker also bears lending risk on the associated vaults.

With this in mind, MakerDAO can consider offboarding assets with very low usage that lack a realistic prospect of becoming profitable in the future. This could improve financial performance with minimal impact on DAI supply and user experience, and allow MakerDAO to direct these resources towards more impactful goals.

Choosing Assets to Offboard

First, Maker must decide which assets will be offboarded. We propose a three part test:

  • Is total DAI minted against the collateral asset (and all others sharing the same price feed) more than 10 million DAI?
  • Is there a realistic prospect of the total amount minted reaching 10 million DAI within 6 months?
  • Does Maker gain a significant benefit for retaining the collateral type that offsets the costs?
    • Eg. Is the oracle price feed used for other purposes, or sold to clients outside of MakerDAO? This may make it worthwhile to maintain the asset.

If we answer no to all 3 questions above, then the asset should be offboarded.

There are other cases where an asset should be offboarded even if we can answer yes to one or more of the above tests:

  • If the asset has been migrated to a new token (eg KNC)
  • If the asset would present unacceptable risk at a total debt exposure of 10 million or greater

Offboarding Actions

For assets that MakerDAO chooses to offboard, the following actions will be taken in sequence (subject to governance approval):

Stage 1 -

  • Publicly communicate that the asset will be offboarded
  • Set debt ceiling to 0 DAI
  • Increase stability fee to 5% (if currently less than 5%)

Stage 2 -

  • Set liquidation penalty to 0%
  • Set liquidation ratio LERP to increase to 500% over 3 months
  • Lower oracle sensitivity to 5%

Stage 3 -

  • Set liquidation ratio high enough to liquidate all outstanding debt
  • Deactivate oracles for affected collateral type

Change log:

24 August 2021 -

  • Removed “stability fee increase to 100%” from Month 6 Offboarding Actions
  • Added context to “offsetting benefits” condition in Choosing Assets to Offboard section

7 September 2021 -

  • Changed offboarding actions from specifying “months” to “stages”
  • Consolidated offboarding actions into fewer steps, and with fewer stability fee increases
7 Likes

Thank you for putting this together! Some thoughts:

  • I think the time horizon for plausibly reaching $10M should be > 6 months. If there is plausible profitability in say 1 year (to be conservative), isn’t better to wait?
  • Related, and especially with a 9-month offboarding process, there should be some proviso about what happens if the offboarding is no longer a good idea (like “revert to a regular DC/SF by executive vote”).
  • On the other hand if offbording is a risk decision, the timeline should be shorter (eg rapidly and drastically increase SF).
2 Likes

Good stuff—but in certain jurisdictions this might be considered Loan Sharking. Here’s Canada :maple_leaf: who has a threshold of 60% per annum:

https://www.justice.gc.ca/eng/rp-pr/csj-sjc/crime/rr02_3/p34.html

1 Like

Frank makes a great point and many, if not most, US states have specific regulations about usury. Now, I am NOT saying by any stretch of the imagination that the activity occuring in Maker vaults is lending in the legal sense but if we can avoid the easy optic of a potentially usurious SF, we should.

Case in point: so-called journalists from publications like Coindesk love to write click bait with ill-informed and manipulative titles (Ex 1:
'A Loan Shark Situation': MakerDAO Is Leaving Crypto Borrowers With Rising Bills - CoinDesk).

Since we don’t have a communications CU to head off journalists, we should aim to avoid providing people like Brady Dale an easy target for tarring and feathering.

4 Likes

I’d be fine removing the 100% rate increase, ultimately the most important part is raising the liquidation threshold high enough to close all of the outstanding vaults. Although if people close out the position themselves, it will likely work out better for them (even if we remove the liquidation penalty, keepers will bid at least a bit below market price so they can pay for gas and still profit). So we’d still want to incentivize repayment and avoid people allowing their vault to be liquidated.

2 Likes

These tests are meant to be open to interpretation and more of a general guideline for governance. We should definitely use judgement when offboarding assets, particularly if we expect adoption in longer than 6 months but still relatively soon.

Agree 100%. Each step in this process will require being passed in an executive vote, so it won’t be on autopilot in any way. If an asset becomes more appealing during the offboarding process, governance will be able to change course and reverse the rate increases/debt ceiling reductions.

Agreed. Risk team will still make any necessary recommendations about stability fee changes as part of the parameter proposal process, so if we need to move faster on rates or liquidation ratios for safety reasons this process will not hold us back.

4 Likes

You might want to add a 4th question to your test: “Is the Oracle feed being used by anyone else than MakerDAO?” I’m not sure if your first question completely captures that, as you can use the Oracles for other than lending services. If there are other paying customers for the feed, we shouldn’t offboard it, just to lose market share to other providers. Which leads me to my next point which is that I believe that there’s more that can be done in terms of selling our Oracle services to external customers to turn unprofitable feeds profitable. Furthermore I think there’s still a lot that can be done in terms of gas optimizations - especially for low use feeds.

My view is that if we want our Oracle services to be succesful we should expand the number of feeds, not start cutting our offerings. In a wide product offering, some products will do better than others, but the network effects of having a large “one-stop” Oracle offering in the end cannot be overstated. So I think we need to look at the profitability of the Oracle services as a whole rather than individual feeds.

However I do think that we should probably think more carefully about which assets we onboard in the future, to ensure that overall the Oracle services stay profitable. Not entirely sure how this is assessed in the onboarding process of today.

Edit: TL;DR - In this stage of the project I think we should focus our attention on growing our userbase and building new profitable feeds, rather than focusing our attention on how to cut costs.

12 Likes

@monet-supply thanks for putting this together. Two QQs - will an individual or committee periodically (eg quarterly) assess assets based on the framework and submit a governance vote to enter the tokens into the delisting process? Also, which tokens are good candidates as of today given the criteria?

2 Likes

Is there any provision for increasing global debt limit to account for debt in ilks with a debt ceiling set to 0? We saw earlier this year that this can effect the ability of vault users or PSM users to mint DAI

4 Likes

I expect that mandated actors (oracles and risk core units) will take the lead on proposing to offboard assets that fit the criteria. But asset offboarding is handled primarily through the signal request governance framework, so it’s open to community members who would like to make offboarding proposals.

In addition to KNC-A, several other vault types have been discussed for offboarding (primarily the ones voted on in this previous signal request).

Currently, USDT based vault types (USDT-A, UNI-V2-DAIUSDT-A, and UNI-V2-ETHUSDT-A) are barely used and seem to be good candidates to offboard.

Other Uniswap LP vaults (LINKETH and AAVEETH) were also voted on to be offboarded in that signal request, but it’s unclear how much cost savings Maker will get from offboarding these assets (as we’ll still maintain the AAVE and LINK price feeds), so these assets require a bit more research before moving forward.

And a few other lesser used assets have also been considered (eg BAT, LRC, MANA, and ZRX), but Nadia from Growth suggested we may want to keep these vault types to allow more time for usage to increase:

Great point. If we set debt ceiling to 0 for an asset with substantial outstanding debt, we can also increase the global debt ceiling by the same amount to avoid this problem.

2 Likes