Pioneered by IDEX in October 2017, refined by Synthetix in July 2019, and implemented at scale by Compound in June 2020, liquidity mining (LM) has captured the imagination of dozens of protocols as a better way to distribute tokens.
A non-exhaustive list of liquidity mining programs from June-September 2020
The effect on the DeFi sector has been electrifying — Total Value Locked is over $10 billion at the time of this writing, up from just over $1 billion on June 16th, 2020. It has also put a strain on the Ethereum network, with gas prices and transaction fees at all-time highs as users rush to capture and realize profits. While the fervor is reminiscent of the 2017 ICO bubble, the fundamentals are stronger.
In this piece, I will explain what liquidity mining is, what has worked well, and what could be improved. While the space is evolving rapidly, I hope to capture most of the interesting developments that could inform both protocols who want to implement such programs and users who want to participate in them.
Liquidity Mining 101
Liquidity mining is a network participation strategy in which a user provides capital to a protocol in return for that protocol’s native token.
This term was alluded to a few years ago by Jake Brukhman of CoinFund, who discussed “generalized mining” in the context of supply-side network participation. The nuance with liquidity mining is that the network has a specific need, which is liquidity provisioning, and users don’t need to purchase tokens but are rather rewarded with tokens, which is often a governance token that allows holders to vote on protocol parameters, including value capture mechanisms. Many often refer to this as “yield farming”; while these terms are often used interchangeably, yield farming does not necessitate a token (e.g. liquidity providers could earn yield on Uniswap solely via transaction fees).
Not all liquidity mining programs are created equally. Looking at the last few months of launches, three categories have emerged:
Non-exhaustive, not fully to scale. Source.
- Fair launches: The primary goal is to distribute the majority of tokens via some objective criteria other than a direct sale (e.g. being an active user of the protocol) and ensure that everybody has equal access to that distribution. Think of this as Uber being owned by its drivers and riders from day one.
- Programmatic decentralization: The primary goal is gradual community ownership and minimizing treasury management. Think of this as Uber signing a legally binding agreement to distribute the majority of its stock to drivers and riders over the next few years.
- Growth marketing: The primary goal is to incentivize specific user behavior over a pre-defined period. Think of this as Uber rebating a portion of their customers’ rides in Uber stock.
Each of these categories has its pros and cons, protocols could fall into multiple categories (e.g. Uniswap hard-coded 2% inflation for long-term programmatic distributions), and the “best” methodology will depend on the goals of the protocol.
Liquidity mining is important for several reasons:
- Broader distribution: The 2017-era ICOs made a lot of retail investors angry. Private rounds with large percentages of the token supply sold to investors resulted in lots of retail pain as those investors exited their positions. Liquidity mining attempts to even the playing field, giving both institutional and retail investors an equal chance of owning a protocol’s native token.
YAM and YFI are highlighted as “fair launches”. Source
- Closer alignment: The benefit of a liquidity mining program is that a token holder is more likely to be a protocol user. 0x had the insight below in late 2019 after analyzing their token holder base, and LM effectively forces this Venn diagram together:
Data is from 10/30/2019 and does not reflect the current distribution. Source
- More inclusive governance: Users who have ownership in a protocol are motivated to help it succeed. By sharing the potential financial upside as early as possible, LM programs strengthen community engagement and help protocols either launch or transition into DAOs.
- Faster experimentation: In DeFi, liquidity = usability. The reflexive nature of liquidity mining programs leads to more capital inflow as the token appreciates, creating a flywheel that lowers the barrier to entry for teams to bootstrap new projects and gain traction in the market. This also results in a downward spiral in the opposite direction — just as bitcoin miners shut off their rigs if the BTC price drops below a certain threshold, so too will liquidity miners withdraw their capital from an AMM or lending pool if the economics no longer make sense. This cycle increases the pace of innovation which ultimately benefits the industry.
What has worked well
The last four months have seen dozens of experiments, and while it is easy to notice the failures, many design choices were successful and should be incorporated into future iterations.
Rewarding long-term liquidity
Much of the liquidity in current LM programs comes from “mercenary capital” that has no loyalty to the protocol and instead pursues opportunities that are most profitable at that time. It’s like your friend who signs up for every meal delivery startup just for the free food. The issue here is that short-term liquidity is less valuable than long-term liquidity, and LM programs should be adjusted to reflect this.
Ampleforth did a good job addressing this issue with the “time multiplier” mechanism in their Geyser program, which granted rewards based on deposit length. The rewards increased (and were awarded retroactively) from 1x on day 1, to 2x on day 30, to 3x after day 60. Because of this, many people were willing to wait two months before they withdrew funds.
The retention figures were mixed but promising. Based on a team update on August 4th (43 days after the program started), about 6,036 unique users have tried Geyser, with 4,242 users still active on that date (a ~70% retention rate). Based on an unconfirmed internal source, on September 8th (78 days later), Geyser (specifically the AMPL-WETH Uniswap pool) had 7,318 unique users and 3,193 active users (a ~44% retention rate). The user dropoff is expected given the launch of many other LM programs, but the liquidity drop off is sharper — as of September 8th, the AMPL-WETH pool has about $9.5m in liquidity out of about $83m in total deposits (~11% liquidity retention).
Liquidity mining programs should not be a “set it and forget it endeavor” — while protocol teams do their best in predicting the behavior of these programs in the wild, they need to be ready to adjust on the fly.
Balancer did a great job fine-tuning its LM program over the initial few weeks by adding five additional parameters aimed at rewarding specific types of liquidity, such as:
- ratioFactor: Penalizes uneven pools for providing less useful liquidity to traders.
- feeFactor: Penalizes high trading fees because they make the pool less attractive to trading
- wrapFactor: Penalizes pairs of highly correlated tokens because they attract less useful liquidity for Balancer
Balancer’s quick and continuous adjustments proved to resonate with liquidity providers. Before the start of the program on June 1st, unique liquidity providers were in the 1–15 range. That number jumped to 71 on June 1st and hasn’t looked back. The month of September saw unique LPs range from 861–1,517.
Source: Balancer Dashboard
Cross-protocol community engagement
Liquidity mining programs do not operate in a vacuum — liquidity providers actively assess their opportunity costs of participating in a program and an effective way of getting them involved is by aligning with the communities that they’re currently a part of.
YAM did a great job here by launching with eight liquidity pools targeting the largest and most active token communities in DeFi:
YAM v1 interface
YAM’s growth was impressive — before finding the contract bug, it reached over $500M in total value locked in 24 hours:
Source: Yam Dashboard
Continuous product innovation
A liquidity mining program for a subpar protocol doesn’t make it a better protocol. Compound, Curve, and Uniswap all did great here by having a functioning and useful protocol before the launch of their LM programs, which made it easier for people to want to participate in the liquidity mining program in the first place.
Furthermore, protocol forks should not focus on merely removing founder and investor allocations, but rather on meaningfully adding utility to the protocol in a way that differentiates itself from the competition. Pickle Finance is doing a good job thus far with a product roadmap that includes several novel yield-generating investment strategies and, ultimately, a stablecoin arbitrage strategy that aims to bring stablecoins back to their pegs. Based is also actively developing their roadmap, which includes a DEX and fair-launch platform.
Shorter program duration
Liquidity mining programs that are too long lose flexibility in responding to shifts in market dynamics and protocol strategy. While one could argue that longer programs are better for token distribution, that distribution could also happen on the open market based on the conviction and time preferences of buyers and sellers.
The other benefit is that you introduce enough float in the market to enable more efficient price discovery. Having a low float at the beginning of a multi-year liquidity mining program for a highly anticipated project runs the risk of damaging the community if the market capitalization starts out too high and early holders lose money on their investment.
Yearn’s YFI token launch is an extreme example, with 100% of the total supply being distributed over 9 days. From a market structure perspective, there was little sell-side pressure because there were no prior holders, which created a virtuous cycle where aligned holders who got in early benefitted most from the financial upside. The token is currently held by 13,507 addresses and has one of the most passionate and engaged communities in the industry.
For teams who opt for longer program durations, one way to strike a balance is to front-load emissions, since earlier liquidity is more valuable than later liquidity. Similar to bitcoin block reward halvings, you could have a decay function where rewards are greater in the first few days/weeks and progressively taper off. SushiSwap did a great job here by accelerating emissions by 10x for the first two weeks, which at the peak allowed them to amass $1.5B of assets, about 73% of Uniswap’s liquidity at the time.
For longer duration LM programs, there exists an economic attack vector where other yield-generating protocols (e.g. Yearn yVaults, Harvest Finance) can participate in the program without the intention of holding the token long-term. This reduces the rewards for participants who are more aligned with the long-term vision of the protocol. A vesting schedule of rewards could reduce the potential for this attack, as mercenary capital would think twice before participating.
Vesting also gives more time for information to propagate across the market, which helps price discovery by letting token holders decide whether this is a viable long-term project (e.g. clear token value accrual, functioning governance system, active community).
DODO made a brave decision in their LM program by locking tokens until one week after they supply the initial liquidity on their AMM and having them vest linearly over the next 6 months thereafter. Even with these restrictions, DODO was able to attract over $90M in liquidity from 3,105 addresses.
More performance metrics
Many protocols are likely starting liquidity mining programs without clear goals of what specific results they want to incentivize, or metrics around how useful those programs are. Ideally, teams should understand that “distributing X% of token supply over Y weeks results in $Z of additional liquidity for the protocol”. And ideally, you get metrics around dollar-cost per unit of liquidity and the duration of liquidity within the protocol, effectively crypto-native versions of CAC and LTV.
- “What % of farmers sell rewards immediately?”
- “What % of farmers vote with their rewards?”
- “How broad is distribution?”
The program was quite successful, at one point attracting ~$20M in locked ETH and providing some important data points for the team, such as the “daily cost-of-liquidity” which varied from under $1,000–4,500 per million.
Most LM programs today disproportionally benefit those with large amounts of capital, which hurts community engagement and token distribution. Based tried to address this by putting a cap of $12k per address on the amount that could be staked in their initial liquidity pool. Pickle also tried to address this by implementing quadratic voting to prevent whales from gaining asymmetric influence over governance decisions. While we don’t know if whales created multiple addresses to circumvent the staking and voting restrictions, it is a step in the right direction.
On supply caps
I believe that long-term oriented projects should not have a capped supply. These protocols are more like companies than currencies and no company would limit their ability to issue shares. In addition, not having the ability to create new liquidity mining programs renders the protocol more susceptible to vampire attacks.
At the same time, a consistently high rate of inflation could destroy value for all token holders. In addition, high inflation could exacerbate governance-related attack vectors, which could have repercussions for the wider DeFi ecosystem. For example, if token X with an uncapped supply and adjustable inflation is accepted as collateral in Compound, malicious actors could vote to mint an infinite amount of token X and steal all of the collateral in Compound. One solution is to hard-code a low-inflation tail that goes into a community-governed treasury or hard-code the option to include terminal inflation while initially setting it to 0% along with an inflation cap.
In addition to the suggestions above, there remain several issues that liquidity mining programs need to address:
- Loopholes: While not intended, LM programs might have the potential for users to “game” the incentives. On Compound, for example, recursive borrowing & lending likely resulted in “fake” volume and crowded out real users. By some unconfirmed estimates, this could be over 30% of Compound’s reported supply value (i.e. if there is ~$1B supplied, ~$700M of that is nonrecursive value). This user behavior doesn’t provide much value for Compound because much of the liquidity in the protocol isn’t accessible by other users.
- Technical risk: Security audits are expensive and teams who want to do a fair launch often don’t have the resources to complete one beforehand. This has resulted in bugs found in mainnet contracts, leading to losses of user funds. This also gives an advantage to those with the technical expertise or resources to check the veracity/safety of contracts. Fair Launch Capital is trying to address this issue by making a no-strings-attached grant to cover the costs of audit and launch.
- “Rug-pulling”: Even if unintended bugs aren’t present, the fact that most liquidity mining programs today are started by pseudonymous founders makes them the perfect breeding ground for scammers. These malicious actors could exploit the contracts (e.g. by calling the “mint()” function like Hotdog or simply selling tokens like Yuno) with little to no repercussions. More technical users could understand these attack vectors by using tools like Diffchecker, but LM remains a dangerous game for retail participants.
- Information asymmetry: While the aim is fair distribution, insiders likely have a head start in the first few minutes/hours of a LM program, which leads to an unfair advantage relative to retail participants. One way of addressing this is by giving sufficient notice that a LM program will be starting.
- Gas costs: High ethereum gas fees tend to “price out” small participants, leaving LM programs to those who could afford to pay the gas fees. This hurts token distribution and lower-value projects like those focused on NFTs and gaming.
While there’s been a ton of experimentation and we likely haven’t arrived at the optimal distribution model, make no mistake — liquidity mining is here to stay. Furthermore, while many of the liquidity mining programs were successful at the time of this writing, readers should be aware that we don’t know the long-term implications. I look forward to another retrospective 6–12 months from now. In the meantime, if you are working on a liquidity mining program for your project then please feel free to reach out!