CIP-34: Testing Fee Models for CoW Protocol

Governance Closed Proposals
, ,
1

The purpose of this CIP-Draft is to gather feedback from the community about a plan to start testing incentive-aligned fee models for CoW Protocol in (roughly) the first half of next year. The draft text of the CIP is below for review and discussion.

Summary

This proposal asks whether CoW DAO should embark on a 6-month period to test promising fee models for the purpose of generating revenue from CoW Protocol.

Motivation

One of CoW DAO’s long-held objectives for CoW Protocol is to make sure that it is financially self-sustainable, so it does not rely on external funding for growth or development. To achieve sustainability in the long term, CoW Protocol must start experimenting with generating revenue in the short term.

The fees that CoW Protocol charges today are estimated to cover gas costs alone; they are collected by the protocol and remitted to solvers. Most revenue kept by CoW DAO to-date is from over-collecting gas fees. This is neither intentional nor desirable (as users occasionally overpay for transaction costs), and it is likely to be addressed by updates to the rules of the solver competition.

Rather than rely on unintentional and unpredictable revenue, CoW Protocol should develop revenue streams that are reliable and sustainable.

Fortunately, CoW Protocol is in a unique position to implement fee models that are incentive-aligned with its users. CoW Protocol leverages intents and uses batch auctions as its price finding mechanism. This allows the protocol to generate surplus and improve prices for its users – in other words, to give them more than they asked for. Taking a fee from this price improvement is one example of a model that is incentive-aligned with users and uniquely achievable by CoW Protocol.

Charging fees can, of course, be controversial. And it is possible that implementing new fees will have a negative impact on user behavior. It is therefore a good idea to focus attention on fee models based on added value (i.e. where CoW Protocol takes fees from the savings it creates for users) and to test and monitor the impact of various fee models before settling on a long-term solution.

Specification

The core team proposes that CoW DAO embarks on a 6-month testing period to collect data that will inform CoW Protocol’s long-term fee model.

In that time, CoW Protocol may test the following fee models:

  • Quote improvement fees: charging a fee on the “extra” amount received by a trader, thanks to the performance of CoW Protocol
  • Surplus fees: charging a fee on surplus (surplus = amount received - quoted amount - slippage) for order types that competing platforms cannot generate a surplus from (e.g. dutch-auction-based models cannot generate surplus for out-of-market limit orders)
  • Volume-based fees: charging a small % (such as one basis point) on the total trade amount of certain token pairs (i.e. not stable-to-stable trades) can be rather lucrative, though less incentive-aligned
  • Fixed fees: charging a fixed amount for every trade
  • Any new idea for a fee model that presents itself during the testing period

Within the scope of this proposal, the core team should test a range of options regarding fee percentages, caps, and volume-based tiers, while closely monitoring market share, retention rate, and other relevant user metrics. They should start by testing the models expected to have the least negative impact on user behavior.

Consistent with CoW DAOs structure and values, fees should always accumulate to CoW DAO and be applied in the ecosystem to the benefit of the DAO through future CIPs. Implementation and communication of the fee mechanics will be performed by the core team, who will also maintain a public notice of current fees on cow.fi. The core team will also provide updates, periodically, on the results of the fee model testing.

Rationale

Additional revenue can be used for multiple purposes: to accelerate the development of CoW Protocol, to acquire new users, to fund token utility, and more. Ultimately, the goal is to make CoW Protocol self-sustainable.

Taking a test-and-learn approach is essential to maximizing the positive impact of new fee models. Passing one CIP to cover 6 months of testing allows the team to move fast and nimbly. Needing to pass a new CIP for each new test would be unnecessarily restrictive.

Execution

If this CIP passes, the core team will begin testing fee models immediately and keep COW token holders updated on learnings periodically.

Fees will accrue to a new Safe that is created for the purpose of revenue model testing. The Safe will belong to CoW DAO be operated by the treasury team with a threshold of 1/2.

Thank you!

10 Likes
2

This proposal is well thought-through. Adopting a test-and-learn approach will allow CoW Protocol to maximize the upside and minimize the downside. I am bullish. And I have confidence in the ability of the core team to generate clear learnings from a thorough and systematic testing program. Excited to see the results!

3 Likes
3

Thank you for the proposal!

In the best case you want to benefit from the bull markets and the degens/the mania, so it might be a good idea to think about different fee models for different market conditions. For example:

The protocol sets a specific limit for the 24 hour trading volume, if this limit is passed, the protocol switches to the bull market fee model and introduces volume-based fees. If the volume falls below the line, you switch back to an incentive aligned fee model. I think it is absolutely fine to charge higher fees during a bull market, because the utility of the protocol rises dramatically (much more MEV etc.)

You can also give the COW token more utility, we can bring back the fee reduction mechanism and people who lock COW get lower fees, or COW holders only get incentive aligned fees etc. etc.
This should increase the price of the COW token, which also benefits the sustainability of the protocol.

4 Likes
4

Personally very excited to see fee initiatives being worked on. The long-term sustainability of CoW DAO is important to continue growth and improvement for its CoWSwap and MEVBlocker products.

My only comment is to consider rebates/trader incentives and if they should also be explicitly listed under the “carte blanche” positioning. I suspect post-trade rebates (only a rolling weekly basis/etc) against a flat fee may be easier for the team to implement vs. protocol updates that adjust fees at the time of trading. Even if not, it may be an interesting tool to allow the team to use in this exploration phase.

4 Likes
5

Hello there so just giving my thoughts on this proposal

Turning on fee switch is a decent idea !

CowSwap is such a nice product, the metrics are talking for themselves.

However, fees shouldn’t be higher than 0,15% - 0,20%. Indeed, the users should not realize or care about paying the fees. In a bull market this won’t even bother anyone, trust me. Ethereum’s gas fees will do the job :'D. In this sht bear market, cowswap has done 1B in volume and november didn’t end yet… Imagine now what happens in a bull run, expect atleast 3 - 5B in traded volume per month. At a 4b$ volume per month, this could be a juicy 6M trading fees earned. What about giving back 30 - 50% of it to the holders ?

I would go with the fixed fees for every trade.

Actually, holders should also be rewarded if this goes ON, there is many ways to do that :
-fee discounts for holders
-distribute fees to stakers, pay them for staking
-incentive programs for holders idk

I think CowSwap team is doing great and turning on fees would be a super banger for this nice product of yours. Definitely okey with that one ! But please do not forget about the token price and your valuable holders.

Honestly sharing fees with holders is a must… Have a nice day

2 Likes
6

What are the proposed fees going to be denominated in? COW? Or the swap token

1 Like
7

For users, the fees will be collected in swap tokens.

8

To update the community on the status of the fee testing …

The first test we ran – a surplus fee on out-of-market limit orders (set at 50% with a cap at 1% of total volume) – generated 130 ETH for the protocol and had no observable impact on retention between Jan 23rd and March 26th.

In this time, we saw a normal number of small volume spikes during highly volatile market days. This does not indicate anything unusual. https://dune.com/cowprotocol/cow-fees

In March, we saw a notable rise in the popularity of limit orders on CoW Swap. Specifically, volume increased from $128M in January to $760M in March. This was somewhat unexpected and likely not correlated with the fee test. Nevertheless, the observed behavior is that adding fees did not depress limit order usage. https://dune.com/queries/1715231/2830822

Over the testing period, ~50% of limit orders come from new users, and month-over-month retention was over 30%. This data is consistent with previous months. We found no meaningful evidence of users moving away from CoW Swap due to the fee model. https://dune.com/queries/3563774/5997236

Take all together, these data points lead us to believe that this first fee test was successful.

The next major fee test (which is already live, as announced on X last week), is a quote improvement fee on market orders. Quote improvement fees are incentive-aligned in that the more CoW Protocol gets for its users, the more CoW Protocol gets for itself.

Initially, the fee will be set at 50% of the difference between the final received amount and the initial quoted amount, with a safety cap at 1% of volume – which it is unlikely to hit. (NB: the fee is only charged when the value is positive, slippage does not impact the calculation.) In other words, if CoW Protocol can find a way to give a user more than they asked for, it will split the extra 50/50 with the user.

As always, anyone can refer to the CoW DAO docs to see which fees are currently live: Fees | CoW Protocol Documentation

Thank you!

3 Likes
9

The testing mode of this GIP was scoped to 6 months. Now it’s 6 months since the last update. Are there any updates or planned updates to the fee model testing? Or could we consider these fees “final”?

Some additional questions:

  1. There seem to be two additional fee types of Volume Fees and Partner Fees that aren’t mentioned in this thread or update post by @marshy - I guess that’s another iteration of the fee model test?
  2. Why are the volume fees only active on Gnosis Chain? Due to a much bigger ecosystem and MEV ecosystem on Mainnet, I don’t think a test on Gnosis Chain alone is very meaningful.

Overall I’m very happy about experimentation and also about Cow DAOs path to economic sustainability. That attitude is making me most bullish on Cow DAO actually. At the same time it is a bit distressing to see that a fee model is in “testing mode” when considering building on top of or integrating Cow Swap. Changes to the fee structure might make an integration prohibitively expensive and it would help to carve them in stone as much as possible - while I do understand that you want some flexibility as market cycles unfold and the ecosystem matures.

10

As can be seen in CIP-49, the term for testing fee models was extended for another six months.

In light of the above I wouldn’t consider the current fee model as final.
(Having said that, if it wasn’t working, it would have been replaced already… so it appears to be a good candidate for the final proposal)

Partner Fees are not part of the core protocol fee model and are decided by integration partners, therefore they aren’t mentioned or considered as part of the protocol fee experimentation.

Volume fees were considered a bigger deviation from the status quo and therefore deemed better to experiment with in a “lower stakes” environment.
Having said that, this fee model seems to be generally more suitable for faster/cheaper chains and L2s, so it is likely to be extended to the Arbitrum and future L2 deployments as well.

:heavy_heart_exclamation:

I agree this is an important point to consider!
Generally I think it is fully in CoW DAO’s benefit to make sure CoW Protocol’s fee model remains competitive across most/all users, chains, tokens and market environments.
(This is the reason that stable pairs don’t take the same fee even on Gnosis)
The final decision will be made publicly through a CIP - so every integrator and partner can rest assured they can participate in this process. (just like they could down the line even after the fee model is considered “final”)

1 Like