Use of $COW in protocol economics

[m0xt]

Cow Protocol and its flagship product, CowSwap, have built a strong position in the market over the years, consistently maintaining around 30% market share. This solid foundation allowed the protocol to begin charging protocol fees starting early last year.

In addition to trading fees, the protocol also earns revenue from MEV Blocker, though that currently accounts for less than 10% of total revenues.

Over the past 12 months, Cow Protocol has generated $16.2M in protocol revenue.

Source: Dune

Monthly revenues have averaged around $1.3M, closely tracking broader market sentiment and trading volumes.

With its expansion to new chains, new integrations and ongoing work on cross-chain functionality, Cow is well-positioned to further grow its market share and strengthen its position in the ecosystem.

The team has consistently delivered, and I’m confident they’ll continue executing at a high level and pushing the project forward.

However, there’s one aspect that makes the business model more complexand arguably less effective, which could lead many potential investors to pass on the opportunity.

Currently, CowSwap users don’t pay any fees upfront since fees are instead “deducted” from the sell token if and only if the trade is successfully executed.

The protocol charges a fee equal to 50% of the user’s surplus, taken in the sell token. However, this protocol fee doesn’t go directly to the protocol itself. Instead, it’s first collected by solvers. At the end of each accounting period, these collected fees are converted into $COW tokens and then transferred to the protocol.

The key point here is understanding what truly drives protocol revenue.

We could discuss whether it is trading volume or surplus generation. Because stable-to-stable trades typically produce little to no surplus and are thus less profitable for the protocol. In contrast, trading volatile assets tends to generate higher surplus and therefore more revenue.

But that’s not the focus of today’s discussion. The key is:
the more people trade, the more revenue the protocol generates—which is directionally correct, even if the details are a bit more nuanced.

The benefits of converting all protocol revenues into $COW are fairly clear—it creates constant buy pressure on the token. However, it also introduces a risk: the protocol is adding a highly volatile asset to its balance sheet, one that’s tightly linked to its own success.

We’ll come back to that point later.

For now, let’s introduce another key piece of the system: the solvers.

Solvers are the behind-the-scenes operators who find the best possible prices for users. They’re a core reason why traders prefer Cow over traditional AMMs, because they consistently get better quotes (and MEV protection).

Of course, solvers deserve to be rewarded for the valuable service they provide. But this is where, in my view, a structural issue arises.

Currently, solvers receive fixed rewards in $COW—for actions like submitting bids, winning auctions, etc. This creates a negative feedback loop for the protocol: the higher the price of $COW, the higher the cost of incentivizing solvers.

To be fair, the $COW reward budget is reviewed and adjusted every six months. Still, the system introduces unnecessary risk for both the protocol and the solvers.

Cow Protocol’s goal is to offer rewards that are high enough to keep solvers engaged and competing. It ultimately leads to better prices for users.
Solvers, on the other hand, aim to cover their costs and make some profit.

But when the medium of exchange used in this relationship is an 80-volatility asset like $COW, it becomes harder for both sides to plan, price, and operate efficiently.

Here’s a chart comparing the protocol’s monthly revenues to the rewards paid out to solvers. It highlights the growing tension between revenue generation and incentive costs—especially as $COW price fluctuates.

Note: this chart isn’t perfectly precise, as I used average monthly $COW prices to better illustrate my point.

Source: Dune

Cow recorded peak revenues in December 2024, as the broader crypto market rallied following Trump’s election and the perceived positive sentiment it brought to the space.

But the rise in overall market prices (including $COW) also drove solver rewards to a record high of $1.9M in December, matching protocol revenues for the same month.

So while December marked a record month for revenue, the protocol effectively broke even—highlighting the cost side of the current incentive structure.

Again, the data might be slightly off due to averaging, but the overall point still stands.

To understand the full picture: solvers first send the collected protocol fees to the protocol. Then, the protocol calculates how much is owed to solvers and distributes $COW tokens to them as rewards.

But from the protocol’s (or an investor’s) perspective, these rewards are purely costs and they need to be accounted for as such. While we don’t have visibility into the exact cost structures of solvers, it’s reasonable to assume they’re not simply holding these rewards. Most will likely sell the $COW tokens shortly after receiving them to cover operational costs and lock in profits.

Let’s assume similar trading volumes in Month A and Month B. That would result in roughly the same protocol revenues and solver activity in both months.

However, in Month A, the price of $COW was $0.30, while in Month B, it was $0.90. Even though everything else stayed flat—including revenues—the protocol’s expenses tripled, purely because the price of $COW went up.

Yet the solvers’ real operating costs didn’t change between the two months. This highlights a core issue: the current reward structure doesn’t accurately reflect the underlying economics—it introduces volatility and inefficiency where there shouldn’t be any.

As mentioned earlier, the current design creates a negative feedback loop for the protocol: the higher the price of $COW, the more expensive it becomes to reward solvers—regardless of whether their actual costs have changed. This misalignment adds unnecessary pressure on the protocol’s economics and makes long-term sustainability harder to model.

What I’d like to open for discussion is this: why not reward solvers in stablecoins instead?

Some argue that paying rewards in $COW helps align solvers with the protocol’s success. That sounds great in theory, but does it actually work in practice?

If true alignment is the goal, there are other, potentially more effective ways to achieve it.

For example, the protocol could offer solvers token options, vesting-based incentives, or discounted $COW allocations—methods that create long-term alignment without exposing the protocol to unnecessary volatility or cost inflation.

The protocol and solvers could sit down together and work out a cost structure based on USD terms. This would give solvers more stable and predictable revenue, while allowing the protocol to better manage and forecast its expenses. It’s a win-win that reduces risk for both sides.

Based on the margin this structure would leave the protocol with, a portion of the remaining revenue (after deducting these solvers costs) could still be allocated to $COW buybacks. That way, if buy pressure is a goal, it can still be achieved in a more controlled and sustainable manner.

And before I wrap up, I have one request that I think would benefit everyone.

Right now, we can track protocol revenue by chain, but we have no visibility into how solver rewards are distributed across chains, since all rewards are currently settled on mainnet.

It would be helpful to see chain-level economics, including both revenue and associated costs, to better understand where the protocol is most efficient and where it has room to improve.

I hope this post helps spark a thoughtful discussion around the current token design and the role of $COW within the broader ecosystem.

[tanglin]

I agree with the proposal, but love to see the team feedback on it as well.

If the COW solvers receive is instantly sold to USD then it doesn’t make sense to pay in COW in the first place, probably a mix of vested COW + liquid USD would be ideal?

[Curia]

We appreciate you putting this proposal together. The main question raised here is about changing the solver rewards from COW tokens to stablecoins, motivated by concerns over the volatility of COW’s price potentially impacting the protocol’s revenue.

However, there seems to be an important additional consideration regarding the mechanism of how solver rewards are funded, specifically:

Where does the protocol source the funds to convert into COW tokens to pay solvers?

If the rewards come directly from the treasury or an allocated reserve, then fluctuations in the COW token price could indeed significantly affect protocol revenue and sustainability. However, if the rewards come from surplus revenue generated by the protocol, the volatility in COW’s price has less impact on overall protocol revenue because:

• The surplus represents excess funds after operational costs.

• The protocol uses this surplus to purchase COW tokens on the open market to pay solvers.

• Hence, the actual revenue and surplus remain intact regardless of COW price fluctuations.

As you mentioned:

@m0xt

The protocol charges a fee equal to 50% of the user’s surplus, taken in the sell token. However, this protocol fee doesn’t go directly to the protocol itself. Instead, it’s first collected by solvers. At the end of each accounting period, these collected fees are converted into $COW tokens and then transferred to the protocol.

It appears that solver rewards are funded from surplus revenue rather than directly from the treasury.

If our understanding is incorrect and the rewards are funded differently, please clarify. We believe it’s important to resolve this question before deciding whether changing the reward token is necessary.

[Hasu]

Thanks for the writeup @m0xt , this is a topic that I have broached several times over the years with the Cow contributors as well, and we had good conversations about it.

I see two separate issues:

1. solvers incentives should be independent of $COW price.

You can easily use a PID controller to address that: Set a floor target (e.g. x competitive bids/auction), then adjust incentives to stay over that floor.

Can be done off-chain as well, no need to do on-chain.

2. paying solvers in $COW token, irrespective ov the amount, is wasteful.

The second point is worth inspecting on its own. Even if your system allocates e.g. $100/day but does so in $COW, you are still wasteful because the solvers will sell $COW and cross the spread in the process.

A counterargument often given is “we don’t have enough ETH/stables to use for incentives” but this is easily dispelled because by paying solvers in $COW, you are “perpetually ICOing” in the public spot market – once you acknowledge that, you can more deliberately sell the same $COW into e.g. the private VC market instead of the public spot market.

A simple rule in mechanism design is to that efficiency is maximized when the mechanism currency used has the least regret for both parties, which in this case is probably ETH or stablecoins.

The same can be argued for the bond solvers have to post – if you require that bond in $COW, at the limit you will get fewer solvers, and the solvers you get will quote you less efficient prices, because some of their revenue is needed to compensate them for the unecessary risk of holding $COW. So using $COW in any mechanism costs the DAO money and makes the service worse, at the limit.

It is one of my longest-running issues with Defi projects that they try to shoehorn their token into their mechanism design, making these mechanisms worse in the process. It was understandable for many years because regulation was so hostile that it was rational for teams to give their tokens some economic linkage even at the cost of making their systems worse. But I think its now becoming rational to graduate from that folly.

[valloderbabo]

From the recent discussion, the following core questions came up for me:

1. Do solver payments in COW come directly from the treasury, or are they sourced from the market? In both cases it seems similar to a form of “continuous token sale.”

2. Why is the reference price for solver payments only updated every six months, instead of being adjusted more dynamically?

3. If incentives are funded from the treasury, has the DAO considered paying solvers in ETH or stablecoins instead, to reduce risk exposure and improve efficiency?

Looking forward to hearing the team’s perspective on this

[notsoformal]

Hello @m0xt , @tanglin , @Curia , @Hasu and @valloderbabo ,
Thank you for bringing this discussion to the forum and following up from the Discord chat, I think it is very beneficial to have this conversation here.

Summary:

1. Are rewards correlated to $COW → No. From the moment COW exceeds a few cents, rewards are capped to a native token amount, hence, the amount of COW rewards does not increase with the price of COW - see below for more details.

2. It is wasteful to pay solvers in $COW → No, because:
   a) This mechanism fosters stronger alignment between the Protocol and Solvers, ensuring that those who contribute to the Protocol’s efficiency and success are also able to share in its long-term growth, while directly influencing its future through active participation in governance. This commitment is evident in the behavior of leading solvers, many of whom continue to hold a substantial share of their COW rewards as a signal of long-term alignment.
   b) The current buyback strategy has had a positive PnL from the Treasury perspective

I will go in more detail on responses below:

However, this protocol fee doesn’t go directly to the protocol itself. Instead, it’s first collected by solvers. At the end of each accounting period, these collected fees are converted into $COW tokens and then transferred to the protocol.

Without getting too technical, the flow of funds is mostly:

1. In each settled batch, solvers “leave in the settlement contract” the slippage (including all fees charged) in surplus token.
   Note: Applicable for non-colocated solvers, as for those they might decide to transfer funds back to the protocol instead of #1 - currrently all solvers behaving the same way on this front*).*
2. On a certain cadence (currently daily, but will be increased in L2s), these tokens are swapped to native token to a Payout Safe.
3. Each week on or around Tuesday, the Solver team runs an accounting script that leads to the payment of solvers (both buffer changes in native token and COW rewards, both quoting and solving), partners (in native token) and the treasury (in native token).
4. The Treasury then initiates a token buyback (TWAP) to cover the amount of tokens emitted / paid to solvers, and sends the remaining native token to CoW DAO Treasury for treasury management.

So, to clarify, all fees are accrued at protocol level (settlement contracts), are withdrawn to one of the entities managed by CoW Foundation and used to pay external parties, with the remainder being used for treasury management.

This means that currently, on an ongoing basis CoW’s balance sheet slightly increases in COW terms (per buyback analysis post here) and in native token.

Currently, solvers receive fixed rewards in $COW—for actions like submitting bids, winning auctions, etc.

This is not the case - documentation.

Solvers are rewarded in two parts:

1. Quote reward:

   1. Valid rewards (as per CIP-72) are paid on a per trade (not batch) basis a capped amount of min [6 COW, X amount of Native token].
      1. Example 1 - Gnosis chain - the payment would be min [0.15 USD or 6 COW], where the cap of 0.15 USD applies.
      2. Example 2 - Base - the rule is min [0.00024 ETH = 1.02 USD , 6 COW = 2.1 USD]. so the native token cap applies
   2. This means that as $COW price rises, there is not a guaranteed higher amount being paid in USD.

2. Solving / competition reward:

   1. Payment to solvers is the additional total surplus provided by a solver compared to the competition, measured in the native token.
   2. This calculated Payment (in native token) is then converted to COW using a Dune price feed for the day prior to the Payouts.
   3. This means that solving rewards are dependent on overall surplus generated, normalised in native token price. Actually, an increase in $COW versus the native token would lead to lower COW rewards being distributed.

Overall, the drivers for the solver rewards are not $COW price, but actually underlying are Native token prices (either because of the caps on quoting, or on the surplus generated on the solving). As these are then just converted to $COW for payment , there is not a direct link between solver performance and rewards to be paid and $COW price increase.

The above explains why in moments of added volume / volatility, there are more Protocol Fees generated (higher price improvements or surplus), but there is also more activity (more quotes and batches being executed), hence more rewards to pay (calculated in native token, converted to COW).

Most will likely sell the $COW tokens shortly after receiving them to cover operational costs and lock in profits.

This is the basis for the CoW Buybacks in the market, to absorb any / potentially all the sell pressure. Also, the solvers bonded under CoW DAO bond are mandated to keep 25% of tokens “soft-locked” to participate in the mechanism (see here).

Even some non-CoW DAO bonded solvers are holding significant amounts of COW.

However, in Month A, the price of $COW was $0.30, while in Month B, it was $0.90. Even though everything else stayed flat—including revenues—the protocol’s expenses tripled, purely because the price of $COW went up.

Assuming that $COW and Native prices are kept constant, any variation on the USD amount paid to solvers corresponds to them generating more value (more quotes, or more surplus). What I mean here is that CoW Rewards cannot be evaluated from a USD perspective, but from a COW amount perspective relating to volume / Fees to benchmark performance.

Right now, we can track protocol revenue by chain, but we have no visibility into how solver rewards are distributed across chains, since all rewards are currently settled on mainnet.

This is a fair request, and we’ll see to add this to the Analytics roadmap.

---

---

Regarding @Hasu ‘s comment, there is no “perpetually ICOing” as rewards paid to solvers are not 100% sold in the market, there is a buyback motion to keep supply from increasing.

solvers incentives should be independent of $COW price.

As per the notes above, from the moment $COW is above a few cents, the native token price caps are reached. Hence, rewards are mostly independent from $COW price

paying solvers in $COW token, irrespective ov the amount, is wasteful.

Several considerations here:

1. Some colocated solvers are holding the token, hence it’s a good strategic alignment having the rewards paid in COW.
2. Colocated solvers are part of CoW DAO bond and required to hold 25% of all rewards paid.
3. The current buyback strategy has actually been PnL positive, where considering that there was an overpurchase of 2.6M COW, at current price level, there is a ~570k USD gain on the treasury operation.

4. We also think that if Solvers were paid in native tokens, they would not buy the same degree in the open market, given friction and, potentially, inertia.

Let’s continue the discussion, thank you for all the comments.

[valloderbabo]

Thanks for the detailed explanation on the solver rewards mechanism and buybacks — it really helped clarify a lot.

I’d like to now shift the discussion to the recent CIP‑72 changes:

Context from X; @rrohit689:
1/ CIP‑72 on @CoWSwap is a great mech‑design lesson: if you reward quotes without demanding execution, you get overbids and toxic flow.
No solve, no payout, fixes this.
2/ Early data shows the meta is shifting toward pure efficiency. Quote rewards (≈50% of solver income) reportedly fell ~95% last week.
Gas/MEV discipline, batch size, and routing now decide which solvers succeed.
3/ Big executors by settled volume/activity — Barter, Rizzolver, Portus, PLM — seem well-positioned for this new regime.
Expect fewer overbids, cleaner fills, and consolidation around the most efficient solvers.
4/ Short-term: quotes may look less aggressive.
Long-term: healthier auctions and more honest pricing.

Question:
It looks like, as a result of CIP‑72, COW payouts to solvers will be much lower because quote rewards dropped so sharply. If that’s the case, how do you expect this to affect COW emissions and buybacks?

For example, with lower payouts and the weekly 20% buyback buffer now easier to cover, would it make sense to adjust buyback targets upward, if this can be reasonably anticipated?

[tanglin]

thanks for your–very detailed–response, this is why after so many years i’m still a big supporter of Cowswap! this makes much more sense now, and I do see the search for aligment in the strategy (even if they only keep 20% of the COW, or even the 10% as part of their treasury it would be net positive vs them getting the entire allocation in USDT)

curious to hear @m0xt thoughts, been following his analysis for years and usually very accurate

[karolak]

The current system works as intended: more COW is being bought back than paid out as solver rewards, which reduces the free float. This is similar to a stock buyback program instead of paying dividends. That means if the protocol is net positive in revenue it already SHARES that $ with $COW holders without any tax complications and staking mechanics needed!

With 180M COW unlocked to fund the next five years of the Cow DAO budget, we could even consider introducing a token burn strategy. If buybacks continue to be positive, the excess COW could be burned directly.

@obodur If you don’t understand the basic concept of revenue strategies like token/share buybacks, you probably shouldn’t be dealing with crypto in the first place.

The more COW held in the treasury, the more valuable each holder’s share becomes, since governance rights give you a say in what happens with those treasury tokens. For example, if 100M COW is removed from circulation and placed in the treasury, and we vote to burn it, your holdings would represent a larger percentage of the market cap, effectively increasing your share of value.

This kind of healthy tokenomics is extremely rare in crypto. There’s no reason to implement short-sighted stablecoin conversions—especially since solvers may prefer to hold COW themselves. Weakening the native token by converting it to stables makes little sense. We could start from 20% of excessive COW buyback burn. Easy to implement in the contract.

The worst what can be done is to let some crypto amateurs bend cow’s longerm strategy for short term gains (in their opinion, cause in reality what is being proposed is less healthy for the price than current system which works and it is AMAZING to see COW buyback bigger than solver rewards.) Also the original OPs proposal has many misconceptions and OP clearly shows ignorance and lack of understanding how the system works.. But gladly that was well explained by @notsoformal which I am thankful for. COW token buyback - An update on 1 year of execution

[karolak]

So you are implying you can not as a company pay its employees or contractors with stocks?

“The companies you mentioned in your post title grant RSUs to employees at hire. These grants vest over time, typically over 4 years. There is a waiting period before any of the grant starts to vest, that waiting period is usually on the 1 year anniversary of employment for a cliff vest of 25%, with the remaining grant vesting in equal increments at some time interval (depends on the company if it’s every month, every quarter, every half year) over the next 3 years. Employees who are identified as people the company wishes to retain will also receive annual refresher grants which vest on a similar schedule as the new hire grant. Googling “company name” + “RSU vesting schedule” will give you the specific details for the bigger name companies.” But probably ur chatgpt didnt help you out there

Stock/Cow Token is a form of payment as it holds certain value, academic discussions does not matter here, currently 20% more of COW is guaranteed to be bought back than paid to solvers. Which means COW token holders are gaining value as long as revenue allows for that excessive buybacks. That should end discussion here.

[notsoformal]

I think that decision will need to be assessed with @kpk and the wider treasury team.

There is still research to be done on the dynamics of COW rewards, e.g. post implementation of CIP-67: Switching to the Fair Combinatorial Auction, and the dynamics of throughput / # of trades settled on solving rewards.

Taking just mainnet values from the solver payouts Dune dashboard and correcting caps for CIP-72 passing, the reduction on Quote rewards (~80% decrease on a week basis), led to a reduction of “just” 20% of total rewards.

So the Buyback target is easier to achieve, but there should be some consistency in the treasury strategy and further analysis prior to signaling an increase in buyback volume.e

[m0xt]

Thank you very much for the detailed explanation.

Apologies for the earlier confusion and incorrect assumptions on my part. The strong correlation I observed between net $COW buy pressure on secondary markets and the token price initially led me to believe there was a direct impact on protocol economics.

The December spike in COW price to $1.10 coincided with a sharp reversal in net buy pressure, which fell to -1.7 million, resulting in net negative buy pressure.

This suggests that during that period, there was heightened protocol activity, likely leading to an increase in both Quote Rewards and Competition Rewards. Given the high market volatility at the time, it’s reasonable to assume that Competition Rewards played a significantly larger role in driving total rewards.

This raises an important question: if we were to see another major market rally, would we observe a similar dynamic—where rewards distributed far exceed the tokens bought back—leading once again to negative buy pressure?

I also noted a comment by @notsoformal pointing out that after the passage of CIP-72, Quote Rewards were reduced by approximately 80% on a weekly basis, yet this “only” resulted in a ~20% decrease in total rewards. This further supports the idea that Competition Rewards are the dominant component.

Another factor that comes to mind is the ability for users to set their own slippage tolerance, which can have a direct impact on protocol profitability. While the default is typically set at 0.5%, users do have the option to lower it, let’s say to 0.1%.

From a UX standpoint, most users likely stick with the default setting, but in theory, if a significant portion of users (or the frontends they use) were to reduce their slippage tolerance, it could materially lower the protocol margins.

Ultimately, these are the key questions any serious investor would be looking to answer, as it’s not possible to underwrite the token without a clear understanding—and confidence—in the protocol’s future profitability and the sustainability of its margins.

Overall, I think Cow is performing exceptionally well—the $11.5B in trading volume during August is a strong testament to that momentum.

Well done to the team. Keep up the great work!

[m0xt]

The period from September 2024 to March 2025 raises some concerns.

Volumes were trending up nicely, in line with expectations for a bull market, so one would assume the protocol performed well. However, that wasn’t the case.

Do we have any insights into what was happening during that time?

It appears that, during this period, actual buybacks (revenues) failed to keep pace with the upward trend in volumes, while solver rewards remained largely flat.

Was there any change in how the protocol charged fees during that period or any structural shift that could explain the divergence between rising volumes and relatively stagnant buybacks?

[notsoformal]

Hi @m0xt , thanks for the follow-up and the nice message!

This raises an important question: if we were to see another major market rally, would we observe a similar dynamic—where rewards distributed far exceed the tokens bought back—leading once again to negative buy pressure?

The negative buy pressure came from an ad-hoc option from the treasury team / KPK on pausing the buyback during the price spike to evaluate where the market is going. I can’t pronounce myself on KPK behalf, but would say that in my opinion the “buyback mandate” should be aspired to be fulfilled in a longer period of time (e.g. that there is zero emission from solver operations in a quarterly basis).

This would alleviate the treasury from allocating too much capital in case of a temporary price spike.

Another factor that comes to mind is the ability for users to set their own slippage tolerance, which can have a direct impact on protocol profitability. While the default is typically set at 0.5%, users do have the option to lower it, let’s say to 0.1%.

That is a reasonable topic if a significant number of users would move to tighter slippage, all rest equal, it could lead to less fee generation on average - at least for limit orders. Nonetheless, this is not the behaviour seen. Also, with dynamic slippage implemented and the fact that users tend to apply sufficient slippage so that trades get executed (given market movements) I wouldn’t call this an “expected outcome”.

Was there any change in how the protocol charged fees during that period or any structural shift that could explain the divergence between rising volumes and relatively stagnant buybacks?

That period was when COW was listed in most centralised venues, which led to needs on allocating capital, as well as, revisiting the Treasury strategy allocation. Going back to my first message, during this period the Treasury committee took the decision to reduce buybacks temporarily to assess were the market would be going.

Note: The following Dune query has been what has been used to manage Buybacks, from a quarterly perspective, buybacks have always been above solver emissions.

https://dune.com/queries/5273980/8661207?time_frequency_e15077=month

Well done to the team. Keep up the great work!

Appreciate the involvement and it’s great to have community members helping build CoW’s future!

cc @tanglin as per request on Discord