Note: This post is an abridged version of our full Uniswap protocol fee report, which contains more details on the background, methodology, results, and conclusions of this analysis.
A protocol fee would redirect a fraction of the fees from liquidity providers (LPs) to Uniswap, generating revenue for the Uniswap DAO. However, this would reduce LP yield, leading to decreased liquidity, higher slippage for traders, and a potential loss of trading volume to competitors.
The Uniswap Foundation has engaged Gauntlet to model the effects of a protocol-level fee on revenue, liquidity, and trading volume for certain deployments of the Uniswap protocol. This report investigates the effects on Uniswap v3 specifically, and may serve as a reference for delegates when they consider potential future votes to implement a protocol fee.
Using a data set comprised of all swap transactions on the Ethereum mainnet deployments of Uniswap, Curve, Balancer, and Sushi between August 2023 and January 2024, Gauntlet has built a simulation engine that can predict how a protocol fee would impact the liquidity and slippage of Uniswap pools. Based on those changes, the engine predicts how swap transactions would be impacted for two different types of orderflow: core and MEV. Core volume represents retail and institutional traders and is assumed to migrate to exchanges offering the lowest slippage. MEV volume primarily consists of toxic arbitrage trades, and is assumed to decrease in line with liquidity. The simulation also considers the potential exacerbation of the negative flywheel effect of reduced LP revenue from trading volume causing further reductions in liquidity.
Using the above methodology, Gauntlet considered the impact on core volume, MEV volume, and protocol revenue across hypothetical protocol fees from 1% to 99% for pools that Uniswap Labs charges an interface fee (those that trade between stablecoins, ETH, and WBTC). While Uniswap V3 only supports protocol fees ranging from 10-25%, projections from outside of that range may be useful for designing future versions of the Uniswap protocol. The revenues are based on Uniswap’s usage from August 2023 to October 2024 and should be seen as ballpark numbers. Actual revenues could be substantially higher or lower depending on macro factors such as overall DEX trading volumes and crypto market prices.
As observed, the impact on volume, TVL, and revenue depends significantly on the fee applied. In the most conservative case allowed by the v3 fee contracts, Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 4.25-14.78% in liquidity, a 4.25-14.78% reduction in MEV volume, and a 0.22-1.44% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 1.93-10.87% for Uniswap with 93-96% of that lost volume coming from a reduction in toxic MEV volume.
From a revenue perspective, if this 10% fee is only applied to the pools for which Uniswap Labs charges an interface fee, Uniswap would earn $10.3-10.8M annually based on market conditions over the time period analyzed or $40m annually based on the peak market volume observed in December 2021.
It’s important to note that the impact on revenue and trading volume varies across different pools, and further work will be required to dynamically set fees to maximize efficiency.
If the Uniswap governance community were to consider implementing protocol fees, Gauntlet recommends an incremental approach to rolling them out. This should begin with a low fee on select pools, followed by an extension to additional pools and a gradual increase in the fee on existing ones, in order to validate the projections presented in our report. Under current market conditions and during an upswing, a conservative fee on carefully selected pools would generate a significant amount of revenue with a limited impact on non-MEV volume.
The community choice of whether or not to institute a protocol fee comes down to whether or not the long-term revenue gained outweighs the losses in volume and liquidity.
Our analysis shows that the losses to TVL and toxic MEV volume may be significant with even a conservative fee switch. Still, we expect the impact on core, non-MEV volume to be very minor under all but the most extreme fees. The only means to identify if a fee switch is an optimal strategy for the Uniswap DAO and protocol community is to perform calculated experiments with a live fee switch, which Gauntlet is keen to help craft.
Note: This post is an abridged version of our full Uniswap protocol fee report, which contains more details on the background, methodology, results, and conclusions of this analysis.
A protocol fee would redirect a fraction of the fees from liquidity providers (LPs) to Uniswap, generating revenue for the Uniswap DAO. However, this would reduce LP yield, leading to decreased liquidity, higher slippage for traders, and a potential loss of trading volume to competitors.
The Uniswap Foundation has engaged Gauntlet to model the effects of a protocol-level fee on revenue, liquidity, and trading volume for certain deployments of the Uniswap protocol. This report investigates the effects on Uniswap v3 specifically, and may serve as a reference for delegates when they consider potential future votes to implement a protocol fee.
Using a data set comprised of all swap transactions on the Ethereum mainnet deployments of Uniswap, Curve, Balancer, and Sushi between August 2023 and January 2024, Gauntlet has built a simulation engine that can predict how a protocol fee would impact the liquidity and slippage of Uniswap pools. Based on those changes, the engine predicts how swap transactions would be impacted for two different types of orderflow: core and MEV. Core volume represents retail and institutional traders and is assumed to migrate to exchanges offering the lowest slippage. MEV volume primarily consists of toxic arbitrage trades, and is assumed to decrease in line with liquidity. The simulation also considers the potential exacerbation of the negative flywheel effect of reduced LP revenue from trading volume causing further reductions in liquidity.
Using the above methodology, Gauntlet considered the impact on core volume, MEV volume, and protocol revenue across hypothetical protocol fees from 1% to 99% for pools that Uniswap Labs charges an interface fee (those that trade between stablecoins, ETH, and WBTC). While Uniswap V3 only supports protocol fees ranging from 10-25%, projections from outside of that range may be useful for designing future versions of the Uniswap protocol. The revenues are based on Uniswap’s usage from August 2023 to October 2024 and should be seen as ballpark numbers. Actual revenues could be substantially higher or lower depending on macro factors such as overall DEX trading volumes and crypto market prices.
As observed, the impact on volume, TVL, and revenue depends significantly on the fee applied. In the most conservative case allowed by the v3 fee contracts, Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 4.25-14.78% in liquidity, a 4.25-14.78% reduction in MEV volume, and a 0.22-1.44% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 1.93-10.87% for Uniswap with 93-96% of that lost volume coming from a reduction in toxic MEV volume.
From a revenue perspective, if this 10% fee is only applied to the pools for which Uniswap Labs charges an interface fee, Uniswap would earn $10.3-10.8M annually based on market conditions over the time period analyzed or $40m annually based on the peak market volume observed in December 2021.
It’s important to note that the impact on revenue and trading volume varies across different pools, and further work will be required to dynamically set fees to maximize efficiency.
If the Uniswap governance community were to consider implementing protocol fees, Gauntlet recommends an incremental approach to rolling them out. This should begin with a low fee on select pools, followed by an extension to additional pools and a gradual increase in the fee on existing ones, in order to validate the projections presented in our report. Under current market conditions and during an upswing, a conservative fee on carefully selected pools would generate a significant amount of revenue with a limited impact on non-MEV volume.
The community choice of whether or not to institute a protocol fee comes down to whether or not the long-term revenue gained outweighs the losses in volume and liquidity.
Our analysis shows that the losses to TVL and toxic MEV volume may be significant with even a conservative fee switch. Still, we expect the impact on core, non-MEV volume to be very minor under all but the most extreme fees. The only means to identify if a fee switch is an optimal strategy for the Uniswap DAO and protocol community is to perform calculated experiments with a live fee switch, which Gauntlet is keen to help craft.
thanks so much for hopping on the community call and presenting your findings!
this was super informative, and I'm excited to see what steps are taken next as we continue to explore the possibilities.
thanks so much for hopping on the community call and presenting your findings!
this was super informative, and I'm excited to see what steps are taken next as we continue to explore the possibilities.
We recently made adjustments to reflect the most up to date methodology and data, if you have any further questions please let us know
We recently made adjustments to reflect the most up to date methodology and data, if you have any further questions please let us know
Thanks for the report!
Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 10.71% in liquidity, a 10.71% reduction in MEV volume, and a 0.75% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 5.24% for Uniswap with 93% of that lost volume coming from a reduction in toxic MEV volume.
Thanks for the report!
Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 10.71% in liquidity, a 10.71% reduction in MEV volume, and a 0.75% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 5.24% for Uniswap with 93% of that lost volume coming from a reduction in toxic MEV volume.
These are promising numbers. Questions / thoughts:
You say that MEV trades "make up between 40 and 80% of volume on Uniswap and other major DEXs", which obviously is a very wide range. What % did you assume to get the predictions listed above?
Does the perfect elasticity model have some more support behind it? It seems there's actually a circular dependence between liquidity and volume. For instance, in the example above the LPs that decide to stay in the pool actually would lose 4.49% of their fee income, since the volume of pool does not stay constant, but decreases. It would be interesting to see this number (the losses of the LPs) quantified for the other protocol fee values... Or update the model so that liquidity is removed iteratively, until the remaining LP loss converges to zero.
Thanks for the report!
Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 10.71% in liquidity, a 10.71% reduction in MEV volume, and a 0.75% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 5.24% for Uniswap with 93% of that lost volume coming from a reduction in toxic MEV volume.
Thanks for the report!
Gauntlet predicts that a flat 10% protocol fee would lead to a loss of 10.71% in liquidity, a 10.71% reduction in MEV volume, and a 0.75% decrease in core trading volume when factoring in the flywheel effect. Combining the loss in MEV volume and Core volume would result in a total volume loss of 5.24% for Uniswap with 93% of that lost volume coming from a reduction in toxic MEV volume.
These are promising numbers. Questions / thoughts:
You say that MEV trades "make up between 40 and 80% of volume on Uniswap and other major DEXs", which obviously is a very wide range. What % did you assume to get the predictions listed above?
Does the perfect elasticity model have some more support behind it? It seems there's actually a circular dependence between liquidity and volume. For instance, in the example above the LPs that decide to stay in the pool actually would lose 4.49% of their fee income, since the volume of pool does not stay constant, but decreases. It would be interesting to see this number (the losses of the LPs) quantified for the other protocol fee values... Or update the model so that liquidity is removed iteratively, until the remaining LP loss converges to zero.
Nice chatting with you on the governance call today, to follow up:
You say that MEV trades “make up between 40 and 80% of volume on Uniswap and other major DEXs”, which obviously is a very wide range. What % did you assume to get the predictions listed above?
Nice chatting with you on the governance call today, to follow up:
You say that MEV trades “make up between 40 and 80% of volume on Uniswap and other major DEXs”, which obviously is a very wide range. What % did you assume to get the predictions listed above?
Estimates of MEV volume differ from source to source, but for our analysis we used Nansen’s address tagging dataset to identify MEV accounts, and labeled all swaps from those accounts as MEV volume. Overall, we identified 45.6% of volume as MEV within our dataset, but this proportion differed from pool to pool. The table below, from our full fee report, shows a per pool breakdown for the top targeted pools:
Does the perfect elasticity model have some more support behind it? It seems there’s actually a circular dependence between liquidity and volume. For instance, in the example above the LPs that decide to stay in the pool actually would lose 4.49% of their fee income, since the volume of pool does not stay constant, but decreases. It would be interesting to see this number (the losses of the LPs) quantified for the other protocol fee values… Or update the model so that liquidity is removed iteratively, until the remaining LP loss converges to zero.
We view the perfect elasticity model as a conservative approach to estimating the response from an LP: in the worst case scenario, LPs facing a 10% loss of fee yield would start removing their liquidity from the pool until the fee yield is restored to where it was before, which comes out to a 10% reduction in liquidity. However, LP yield is highly volatile on a day to day and month to month basis, so it is possible that this reduction in yield will be insignificant compared to the fluctuation that LPs are already used to. As such, it is likely that the loss of liquidity will be less than projected.
On the point about the circular dependence, this is something we are modeling directly as a feedback loop within our sim. While we were light on the technical details of this within this tl;dr summary of our fee report, we discuss it in depth in our full fee report as the flywheel effect. Let us know if you have further questions about the approach!
Nice chatting with you on the governance call today, to follow up:
You say that MEV trades “make up between 40 and 80% of volume on Uniswap and other major DEXs”, which obviously is a very wide range. What % did you assume to get the predictions listed above?
Nice chatting with you on the governance call today, to follow up:
You say that MEV trades “make up between 40 and 80% of volume on Uniswap and other major DEXs”, which obviously is a very wide range. What % did you assume to get the predictions listed above?
Estimates of MEV volume differ from source to source, but for our analysis we used Nansen’s address tagging dataset to identify MEV accounts, and labeled all swaps from those accounts as MEV volume. Overall, we identified 45.6% of volume as MEV within our dataset, but this proportion differed from pool to pool. The table below, from our full fee report, shows a per pool breakdown for the top targeted pools:
Does the perfect elasticity model have some more support behind it? It seems there’s actually a circular dependence between liquidity and volume. For instance, in the example above the LPs that decide to stay in the pool actually would lose 4.49% of their fee income, since the volume of pool does not stay constant, but decreases. It would be interesting to see this number (the losses of the LPs) quantified for the other protocol fee values… Or update the model so that liquidity is removed iteratively, until the remaining LP loss converges to zero.
We view the perfect elasticity model as a conservative approach to estimating the response from an LP: in the worst case scenario, LPs facing a 10% loss of fee yield would start removing their liquidity from the pool until the fee yield is restored to where it was before, which comes out to a 10% reduction in liquidity. However, LP yield is highly volatile on a day to day and month to month basis, so it is possible that this reduction in yield will be insignificant compared to the fluctuation that LPs are already used to. As such, it is likely that the loss of liquidity will be less than projected.
On the point about the circular dependence, this is something we are modeling directly as a feedback loop within our sim. While we were light on the technical details of this within this tl;dr summary of our fee report, we discuss it in depth in our full fee report as the flywheel effect. Let us know if you have further questions about the approach!
