Results and Analysis: Uniswap Arbitrum Liquidity Mining Program

Executive Summary

Over a nine-month period, Gauntlet executed a liquidity mining campaign on the Arbitrum network, utilizing 1.8M ARB tokens (approximately $1.7M USD at the time of commitment). The campaign aimed to create a self-sustaining flywheel effect by attracting liquidity, improving price execution, and increasing volume market share on various Uniswap pools. To maximize impact, we shifted our pool selection methodology from a market share-driven to a simulation-based approach.

Overall, the campaign saw positive results: 

• $15.5M in market share-adjusted TVL added to targeted pools during the program
• $10.6M of TVL maintained in targeted pools post-incentives
• $9.11 of TVL added per $1 of incentives during the program
• $5.99 of TVL per $1 of incentives post-incentives

The program generated notable increases in volume and fee revenue: 

• $823M in market share-adjusted volume added during the program, equating to $259 of volume added per $1 spent 
• $318M in volume maintained post-incentives, reflecting $119 per $1 spent
• $725K in market share-adjusted LP revenue
• $114K in LP revenue maintained post-incentives, amounting to $1.37M in projected LP fees over the next 12 months. 
• $187 in additional volume for every $1 spent on incentives

A strong correlation (r=0.93) was observed between TVL and liquidity within +/- 2% of the current tick, driven by Merkl’s reward distribution favoring fee-generating liquidity. Some LPs engaged in automated, anticipatory rebalancing during reward distributions, and this activity continued post-incentives, indicating sustained engagement due to market share gains achieved during the program. Increased liquidity led to improved price execution, attracting more trading volume. 

Gains in volume-market share were moderately correlated (r=0.58) to optimal-routing market share (discussed in detail below). During the campaign, optimal-routing market share increased by +27.3%, adding approximately $570M in volume, and by +24.8% post-campaign, adding approximately $147M in volume.

The results show that the liquidity mining campaign on the Arbitrum network targeting Uniswap pools successfully boosted TVL, trading volume, and LP revenue, demonstrating a strong ROI and sustained impact even after removing incentives. The strategic shift to a simulation-based approach for pool selection and the active optimization management to sustain a tight feedback loop for pool adjustments were key to maximizing the program’s success.

We thank the Uniswap DAO for funding this initiative. To view our original proposal click here.

Background

Program Details

As mentioned above, over the past nine months, we executed a liquidity mining (LM) campaign on the Arbitrum network using 1.8M ARB tokens (approximately $1.7M USD at the time). The primary objective was to create a self-sustaining flywheel effect, where additional incentives attract liquidity, improve price execution, and increase volume market share. This cycle was expected to attract even more liquidity providers (LPs), sustaining growth in liquidity and trading activity.

The program was designed to target specific Uniswap pools on the Arbitrum network to foster growth and stability. During the campaign, significant revisions were made to the pool selection methodology, shifting from a market share-driven approach to a simulation-based approach. This strategic shift was based on the assumption that improved price execution would capture more market share from competitors.

Methodology

Pool Selection

Initially, pools were selected based on their market share. However, over the course of the program, the methodology evolved to focus on the impact of additional liquidity on price execution. The simulation-based approach aimed to identify pools where increased liquidity would lead to better price execution, expecting to attract more trading volume and improve market share. For more information about our methodology, please refer to our Uniswap Arbitrum price execution analysis.

Metric Calculations

Performance metrics are calculated based on average values during three periods: "pre-incentive," "during incentives," and "post-incentive." The pre-incentive and post-incentive periods include data from 30 days before and 30 days after the last incentive, respectively.

Reported incremental changes in TVL, volume, and LP revenue are derived by multiplying the change in market share by the total market TVL, volume, and LP, respectively, which includes both the incentivized pool and its competing pools. These figures can be annualized by multiplying by 12 and assuming prevailing market trends remain the same for the next year. This approach is necessitated by the need to normalize out market effects where price fluctuations and changes in demand may affect aggregate values between periods.

We use a shorter time window to determine statistically significant changes in pools: 15 days before incentives compared to 15 days during incentives. The rationale is that LPs typically respond almost immediately to appealing incentives, so any change in liquidity should be apparent quickly. We employ a one-sided Mann-Whitney U-test to compare TVL share values, examining whether there is a statistically significant positive shift in liquidity towards the incentivized pools.

How do we measure impact and ROI?

To accurately measure the impact and ROI of the incentives, we used several key metrics that account for market fluctuations and changes in the trading environment. Here’s how we did it:

1. Incentive Spend: This metric represents the USD value of the ARB tokens used as incentives. It provides a baseline for calculating ROI and assessing the incentive program's overall expenditure.
2. TVL Added: To estimate the impact on TVL and control for extraneous variables, we considered the TVL share, which is the proportion of total TVL across both incentivized and competing pools (such as those on Camelot, Traderjoe.xyz, Sushiswap, Balancer and could include other fee tiers of Uniswap V3 pools for a given pair). This involves comparing the TVL share during and after the incentives to the pre-incentive period, and then multiplying the change in share by the total market TVL. This method helps isolate the impact of the incentives from other market factors, providing a clearer picture of how much additional TVL was generated due to the incentives.
3. Volume Added: Similar to TVL, trading volume added was normalized by market share. This is done by taking the difference in market share between periods (during and post-incentives) and multiplying by the total volume during or post-incentive periods. By analyzing changes in the market share of trading volume during and after the incentives compared to the pre-incentive period, we can determine the additional volume attributable to the incentives.
4. LP Revenue: The revenue generated for Liquidity Providers (LPs) was also normalized by market share. The calculation is done identically to volume, albeit using fee marketshare as opposed to volume marketshare. This calculation involved comparing the LP revenue share during and after the incentives to the pre-incentive period and multiplying the change by the total market LP revenue.
5. Projected Future LP Revenue: To estimate the long-term impact, we projected the future revenue generated over the next 12 months. This projection was based on the assumption that current market trends would continue, allowing us to annualize the incremental changes in TVL, volume, and LP revenue. The calculation is simply multiplying the post-incentive market-normalized LP revenue by 12 (since post-incentive periods comprise of 30 days following the last incentive distribution).
6. Overall ROI: The overall Return on Investment (ROI) was calculated by dividing the projected future LP revenue by the incentive spend. This metric provided a comprehensive view of the incentive program's efficacy, illustrating the return generated for each dollar spent on incentives.

By normalizing these metrics against market share, we were able to control for external factors such as price fluctuations and changes in market demand, ensuring a more accurate assessment of the incentives’ impact and ROI.

Performance Results

Overall

• Incentive Spend: $1.7M USD at the start of the program.
• TVL Added: During the program, targeted Uniswap pools on Arbitrum gained $15.5M in market share-adjusted TVL, with $10.6M maintained post-incentives. This translates to $9.11 of TVL added per $1 of incentives during the program and $5.99 per $1 spent post-incentives. The impact varied by pool, ranging from -$41 to $141 during the program and -$48 to $145 post-incentives.
• Volume Added: We observed an $823M increase in market share-adjusted volume during the program, with $318M maintained post-incentives. This equates to $259 of incremental volume added per $1 spent during the program and $119 per $1 spent post-incentives. The relative impact on volume varied by pool from a minimum of -$6K to a maximum of $4.7K during the program and -$2.6K to $4.2K post-incentives.
• LP Revenue and Projected Future Revenue: The program generated $725K in market share-adjusted LP revenue during its course, with an additional $114K maintained post-incentives. Annualized, this amounts to $1.37M in projected LP fees over the next 12 months, assuming stable market conditions.

Total Value Locked (TVL)

Below, we break down changes to TVL during-incentives and post-incentives (30 days since the last incentive):

DURING:

• TVL USD: Percentage change Pre vs. During: +14.13%
• TVL Share Percentage change Pre vs. During: +19.3%
• Net Market-adjusted TVL (USD) added During incentivization: +$15.53M

POST:

• TVL USD: Percentage change Pre vs. Post: +9.8%
• TVL Share change Pre vs. Post: +15.3%
• Net Market-adjusted TVL (USD) added Pre vs. Post incentivization: +$10.19M
• On average, it took 23.9 days for pools’ TVL to settle at a new TVL baseline.
• The ROI for TVL (USD) is approximately $5.99 in additional TVL for every $1 spent on incentives ($1.7M total).

During the incentive distribution period, there was a significant positive impact on TVL across the targeted pools. The TVL in USD saw a percentage increase of +14.13% when comparing the pre-incentive period to the during-incentive period. Additionally, the TVL share experienced an uplift of +19.3%, indicating a substantial gain in the proportion of TVL relative to competing pools. The net market-adjusted TVL added during this period amounted to +$15.53M, demonstrating the effectiveness of the incentives in boosting liquidity.

Post-incentive TVL growth slowed relative to the during-incentive period, but remained 9.8% higher than the pre-incentive period. The TVL share also rose by +15.3%, further solidifying the gains made during the incentive period. The net market-adjusted TVL added post-incentivization was +$10.19M.

One key characteristic is the time it took for pools’ TVL to settle at a new baseline after the incentives were removed. TVL is defined as settled or stabilized when the day-to-day changes in TVL converge to zero, as determined by fitting a convex polynomial to the TVL data and identifying the inflection point where these changes reach their minimum. On average, it took 23.9 days for TVL to stabilize, indicating a relatively quick adjustment period. Notably, the ROI for TVL in USD was approximately $5.99 in additional TVL for every $1 spent on incentives, with a total incentive expenditure of $1.7M.

In summary, the incentives drove significant increases in TVL during and after the incentivization period, with sustained gains observed even after the incentives were removed.

Volume

DURING:

• Volume USD: Percentage change Pre vs. During: +12.4%
• Volume Marketshare Change Pre vs. During: +19.93%
• Net Market-adjusted Volume (USD) added During incentivization: +$823M

POST:

• Volume USD: Percentage change Pre vs. Post: +6.28%
• Volume Marketshare Change Pre vs. Post: +15.6%
• Net Market-adjusted Volume (USD) added Pre vs. Post incentivization: +318M
• The ROI for Volume (USD) is approximately $187 of additional volume for every $1 spent on incentives.

Next, we examine the program's impact on market share in terms of fees and volume to give us a sense of how the incentivized pools performed relative to similar competitors with minimal confounding effects from price fluctuations. The best and worst performing pools in terms of market share changes between pre-incentive and post-incentive periods are shown in the table below:

Interestingly, all BTC-denominated pools performed exceptionally well, largely owing to their very low initial market share and not necessarily due to any unique traits of BTC. The incentives showed a remarkable ability to revive “dead pools” (pools with less than $10 of TVL) and leave a lasting positive impact after their removal.

For the worse-performing pools, additional context is warranted. For instance, while the BOOP/WETH 1% pool showed a decline in market share, it now has over twice the volume observed before incentives. Additionally, Camelot was competing for market share around this time. We closed out our incentives on BOOP while maintaining a dominant share of TVL at 97%. Similarly, for the TIA.n/WETH 1% pool, it was competing with Traderjoe.xyz for TIA.n volume, with some farming opportunities specifically around bridging TIA to Traderjoe.xyz pools, as highlighted by various market activities.

LP Revenue

DURING:

• LP Revenue USD: Percentage change Pre vs. During: +12.4%
• Fee Marketshare Change Pre vs. During: +20.51%
• Net Market-adjusted LP Revenue (USD) added During incentivization: +$725k

POST:

• LP Revenue USD: Percentage change Pre vs. Post: +6.28%
• Fee Marketshare Change Pre vs. Post: +14.55%
• Net Market-adjusted LP Revenue (USD) added Pre vs. Post incentivization: +$114ksome text
  • Annualized, and assuming market conditions remain, this annualizes to $1.37M of additional LP Revenue over the next 12 months.
• The ROI for LP Revenue (USD) is approximately 23% ([$1.37M + $725K] / $1.7M spend).

Proportional to volume, LP Revenue increased 12.4% during incentivization and 6.28% post incentives. Cumulatively, throughout the campaign, LPs earned an additional $839k in fees. Assuming current market conditions hold, LPs are expected to earn an additional $1.37M over the course of the next 12 months.

Other Learnings, Observations, and Best Practices

Optimal-routing Volume (Price Execution)

Our pool selection methodology was based on the assumption that increased liquidity would lead to improved price execution, which in turn would attract more trading volume. This hypothesis was supported by our observations: gains in volume-market share were moderately correlated (r=0.58) to optimal-routing market share. Optimal-routing market share is determined by how historical trades would have been routed to the optimal DEX destination in terms of price execution. Utilizing our RPC simulation tooling for these calculations, this correlation lends credence to our pool selection methodology, which assumes that volume is driven by price execution quality.

Throughout the campaign, we observed a significant increase in optimal-routing market share. During the incentivization period, the optimal-routing market share increased by +27.3%, equating to approximately $570M worth of additional volume routed away from competitors.

After the incentives ended, there was still a substantial gain with an increase of +24.8% in optimal-routing market share, translating to approximately $147M of additional volume on targeted Uniswap pools. These metrics indicate that incentivized pools captured an additional 24%-27% of optimally routed volume, resulting in a total of $717M being re-routed towards the incentivized pools, as they offered superior price execution relative to competitors.

Liquidity Concentration

During the liquidity mining campaign, we observed a strong correlation (r=0.93) between TVL and liquidity within ±2% of the current tick. This is expected given Merkl’s reward distributions were designed to be proportionate to the fees each LP generates. However, we also noted significant reductions in depth coinciding with reward distribution phases, suggesting some LPs were engaging in automated, anticipatory rebalancing. The charts below show normalized values for visual convenience:

While this phenomenon was present in only a handful of pools, it suggests that LPs were actively rebalancing their positions in response to reward distributions. For several pools, this effect continued even after incentives were terminated. This suggests that LPs chose to maintain and manage their liquidity positions due to the market share (volume) gains achieved during the program, which offered an appealing yield opportunity. This continuous engagement by LPs reinforces the positive impact of improved price execution on maintaining trading volume and liquidity in the pools.

The Arbitrum LM program used a 98/1/1 parameterization, heavily favoring fee-generating liquidity. Merkl’s reward distribution parameters determine what proportion of total incentives are distributed based on (1) fees generated by the LP, (2) the quantity of tokenA in the active range, and (3) the quantity of tokenB in the active range. This 98/1/1 configuration was chosen to encourage LPs to actively manage and allocate liquidity efficiently around the current tick, aiming to maximize price execution improvements and drive volume growth for the pool.

By promoting such active liquidity management, this parameter setting aimed to create a self-sustaining flywheel effect. In contrast, other settings, such as 40/30/30, where rewards are more evenly distributed, might enable LPs to park their liquidity with less active management, potentially failing to enhance price execution as effectively.

Nevertheless, this parameter setup also raised concerns about potential exploitation, where LPs might drive additional volume themselves by exploiting Merkl’s reward structure. This issue is more pronounced in new chains or pools with low volume, where LPs could wash trade in small bursts to absorb all the rewards. While Arbitrum is not a nascent chain, we conducted due diligence to identify any signs of exploitative practices by the LPs.

Initial observations suggest that self-driven swaps by LPs were not major contributors to the surges in volume during incentivization periods for most pools. We measured that they accounted for an average of 3% of the volume across all pools, with a high of 13% and low of 0%.

We must note that some level of LP-driven volume is expected as LPs actively manage their positions through rebalancing. However, the percentage of LP-driven volume should not dominate the total trading volume, as this would imply exploitation via wash trading or potentially Just-In-Time (JIT) liquidity practices.

Upon examining the on-chain records, we found no evidence of JIT liquidity behavior. The frequent rebalancing observed is characterized by mints and burns that are, on average, significantly larger than swaps, leading to spikes in volume during rebalancing periods. Additionally, there were no signs of flash loan usage related to any JIT behavior, nor was there any observed increase in sandwich-MEV activity for the incentivized pools.

While LPs did contribute some trading volume, this did not appear to significantly undermine the program's goals.

Incentives Duration

Examining the relationship between the duration of incentives and TVL ROI reveals a weak negative correlation (r=-0.18). We had expected that incentives distributed over a longer period of time would have diminishing returns, as TVL tends to jump quickly after incentives start. However, the weak correlation is not convincing of diminishing returns.

Notably, several pools were terminated early, especially in cases where TVL did not respond to incentives within the first two weeks. This early termination highlights the importance of promptly assessing the effectiveness of incentives and adjusting the strategy accordingly to maximize capital efficiency.

Conclusion

Summary

The Arbitrum Liquidity Mining program has been relatively successful in meeting its objectives of attracting liquidity (added $15.5M and $10.6M of TVL during and after incentives, respectively), improving price execution (24-27% increase in optimally-routed volume market share, accounting for an additional $717M of volume re-routed), improving volume market share (a 15.6% increase in market share, adding a cumulative $1.1B of additional volume), and increasing LP revenue (an additional $725K earned during incentives and an expected additional $1.37M over the next 12 months). The nine-month campaign yielded significant insights and valuable outcomes that point toward the efficacy and potential of such programs.

The sustained improvements in TVL and market share post-incentive periods indicate that the benefits of the LM program can extend beyond the duration of the incentives. This suggests that such programs can create a foundation for long-term growth and stability. The transition to a simulation-based pool selection methodology has proven beneficial in targeting pools with a high potential for capturing market share through improved price execution. This approach can be further refined and applied to future programs to maximize their impact.

While some pools did not respond to incentives and were actively cut early, this indicates that our pool selection process is still imperfect at identifying opportunities for lasting growth relative to competitors. Moreover, it still highlights the need for active incentive management and to cut incentives to pools early so incentive spend remains efficient.

What’s Next?

Considering the program’s overall positive outcomes, we will continue to invest in refining our pool selection and allocation methodology. We recognize that these refinements may need to be tailored to the specific goals of the program or ecosystem, whether they aim to bolster TVL, capture market share, or optimize for another KPI.

Future efforts will place a stronger emphasis on the parameterization of Merkl distributions. For this program, it was logical to distribute rewards to maximize volume and price execution. However, other programs may require a more nuanced approach, targeting specific types of liquidity provisioning, such as one-sided liquidity, and encouraging desired LP behavior, such as boosting incentives for assets bridged from other chains.

Looking ahead, Gauntlet is excited to apply these learnings and strategies to future LM efforts. This combination of strategies enables us to effectively achieve these incentive programs' goals and maximize their impact. We are currently supporting the Arbitrum UADP/LTIPP incentivization program, where select pools will systematically be allocated incentives to encourage TVL growth.

Alongside strategies like Protocol-Owned Liquidity (POL), Liquidity Mining (LM) remains a vital tool in our arsenal for enhancing on-chain liquidity. LM provides significant leverage by bootstrapping a large amount of liquidity and TVL relative to the amount of dollars spent in the short term.