# Marinade LM Research

As a part of the Meta-DAO's report to Marinade, we recommended that Marinade use liquidity mining to attract stake to the new delegation strategy. We also recommended that Marinade opt to spread the incentives over a long period of time, even if that means lower APR. This is the research that we used to come to the latter conclusion.

## Methodology

Simply, we wanted to ascertain whether it was better to have smaller incentives over a longer period of time ('long and small') or bigger incentives over a shorter period of time ('short and big').

To do so, we used two main strategies:
- **First principles reasoning**: determine exactly what would make for a 'good' liquidity mining program and form hypotheses on which length is better.
- **Empirical evidence**: use data from past liquidity mining programs, both on Ethereum and Solana, to test these hypotheses.

Unfortunately, the empirical evidence wasn't as robust as we'd like. For example, some of the longest DeFi LM programs have been run by early protocols. It's unclear whether their growth was due to their liquidity mining or simply being early. The other challenge was collecting good data: although protocols generally announce when they *start* liquidity mining programs, they scarcely announce when they end them.

Still, some data is better than no data! But we will contextualize it, using metrics like overall DeFi growth over the period of the LM program to determine how much of the sticky capital the LM program was responsible for.

## First principles reasoning

> Traditional liquidity mining has been criticized as encouraging mercenary capital that flees the moment rewards dry up.
>
> [Balancer Labs](https://medium.com/balancer-protocol/unlocking-the-vemodel-f363d2d7bd91)

> A successful “Trade Mining” program requires the two following statements to hold true:
> - Sybil Resistance - The incentives program must be difficult or impossible to exploit
> - Inelastic Users - Users tend to be “sticky” and continue using the product after incentives are removed
>
> [Gauntlet](https://www.gauntlet.xyz/resources/uniswap-incentive-design-analysis)

When done correctly, a liquidity mining program attracts sticky users. When done incorrectly, a liquidity mining program attracts mercenary users who leave the second the program is done. While the latter group may improve a protocol's metrics (e.g., TVL) in the short-term, the protocol is by definition a net-loser to them because those users wouldn't have shown up unless the LM (and thus the dilution) was greater than any required fees. So we want to attract future sticky users.

So let us ask: what does a sticky user look like? We think that these traits are correlated with stickiness:
- **Low engagement in DeFi**: highly-engaged users tend to always know which farms are the most profitable and to exploit those opportunities. Users without the time or a strong enough interest will not know about as many of these opportunities and will be more inclined to 'set it and forget it.'
- **Relatively low sophistication**: while any interaction with DeFi requires a basic level of sophistication, highly-sophisticated users are less likely to stick with a given brand based on trust alone.
- **Price-insensitive**: those who are less price sensitive are by definition okay paying more fees in order to gain convenience.

From this archetype, we can hypothesize that 'long and small' would be better for this type of user. Since they are price-insensitive, they will be okay receiving less rewards. And since they have low engagement in DeFi, they may not even see a 'short and big' program.

On the other hand, we can also see how a 'short and big' strategy would be perfect for attracting the type of mercenary capital that we don't want. Since the mercenaries are highly-engaged, they will be likely to see the announcement. Since they are price-sensitive, the big APR boost would significantly impact their behavior.

Gauntlet agrees that short-term programs are likely to attract less sticky TVL:

> One possible theory for a sustained lift from a liquidity mining program is through forgetful LPs. If liquidity is supplied to capture incentives and left unmonitored, it may remain unmoved for some time after the program ends.
>
> This may be observable in some long-lasting liquidity mining programs but is unlikely to be observed in such a short-term (several weeks) program where the end date of the program was known from the start.
>
> [Gauntlet](https://gov.uniswap.org/t/uniswap-liquidity-mining-analysis/21416)

## Empirical data

We were able to collect data on 8 protocols. Here is that data:

| Protocol | Start | End | Pre-LM TVL | Post-LM TVL | post-LM TVL as a percentage of pre-LM TVL | DeFi % change from pre to post | present TVL | present TVL as a % of post-LM TVL | present DeFi as a % of post-LM DeFi |
| -------- | ------ | --- | ------ | ----------- | -- | -- | -- | -- | -- |
| Yearn | 07/20 | 08/20 | $8M | $250M | 3025% | 200% | $353m | 141% | 1000% |
| Compound | 06/20 | ongoing | $100M | $2.4B | 2400% | 3667% | $2.4B | N/A | N/A |
| Aave | 04/21 | 06/22 | $3.3B | $8.2B | 251% | 69% | $7.4B | 90% | 110% |
| Balancer | 06/20 | ongoing | $8m | $980m | 12250% | 3667% | $980m | N/A | N/A |
| Synthetix | 08/19 | 12/23 | $30m | $740m | 2403% | 13,333% | $732m | 99% | 108% |
| Solend | 10/21 | 10/22 | $61m | $230m | 284% | 12.5% | $173m | 75% | 107% |
| Convex | 07/21 | 12/22 | $3b | $20b | 667% | 206% | $1.7B | 8% | 35% |
| Uniswap ([Optimism Test](https://gov.uniswap.org/t/uniswap-liquidity-mining-analysis/21416)) | 11/22 | 01/23 | N/A | N/A | 230% (avg) | 121% | N/A | N/A | N/A |

This data seems to confirm our hypothesis.

Both of the short LM programs (Uniswap Optimism and Yearn) were able to attract significant TVL while the program lasted, but the TVL didn't stick. Since the end of Yearn's program, its TVL has increased 41% while all of DeFi grew 900%.

On the other hand, the multi-year programs appear to be successful. Although it's hard to determine how successful the Compound and Balancer programs are at attracting sticky liquidity given that the programs are currently ongoing, both have tracked overall DeFi growth - slightly underperforming it in Compound's case and outperforming it by 4x in Balancer's case. Synthetix is a better datum because they finished their LM program. Since then, only 1% of TVL has left, albeit in a time that DeFi TVL grew 8%.

And when we look at the medium-term liquidity mining programs, it's a mixed bag. It looks like Solend and Aave's programs were moderately successful at attracting sticky TVL. Convex, on the other hand, has performed exceptionally poorly, with their current TVL at only 8% of their post-LM TVL (DeFi TVL is currently 35% of what it was when they finished their program).

Essentially:
- Both short-term programs were unsuccessful at attracting sticky TVL.
- The three multi-year programs were successful at attracting sticky TVL.
- The three medium-term programs were moderately successful to completely unsuccessful at attracting sticky TVL.

So there appears to be a positive relationship between the length of the program and the stickiness of attracted TVL.

## Conclusion

We set out to determine whether it would be better for Marinade to have a longer LM program with proportionately smaller rewards or a shorter LM program with proportionately larger rewards. Having reasoned through this question and analyzed empirical data, it is our determination that a longer LM program with proportionately smaller rewards would be better for Marinade.

## Resources
- https://gov.uniswap.org/t/uniswap-liquidity-mining-analysis/21416
- https://ar5iv.labs.arxiv.org/html/2105.13891
- https://x.com/Rewkang/status/1415870042307276803?s=20
- https://governance.aave.com/t/arc-extend-aave-liquidity-mining-rewards/4852
- https://medium.com/blockchain-capital-blog/uniswap-liquidity-mining-analysis-8a14be698eb0
- https://alexkroeger.mirror.xyz/HoTLzeiTUBn7c-uZoVcZ6PO9AlGrVQI_4WYDSeJFTiA
- https://mirror.xyz/tapiocada0.eth/CYZVxI_zyislBjylOBXdE2nS-aP-ZxxE8SRgj_YLLZ0
- https://sips.synthetix.io/sips/sip-276/
- https://www.nansen.ai/research/all-hail-masterchef-analysing-yield-farming-activity
- https://www.gauntlet.xyz/resources/uniswap-incentive-design-analysis
- https://hackernoon.com/the-death-of-fcoin-a-tale-of-bad-token-design-261d64a8116f
- https://multicoin.capital/2020/08/13/exploring-the-design-space-of-liquidity-mining/
- https://docs.google.com/presentation/d/1Eh4LQ4R7D0wpwAUOb3Tr1Z83GKl2s9AQSvZdH05b8LQ/edit#slide=id.p
- https://sips.synthetix.io/sips/sip-8/
- https://x.com/synthetix_io/status/1733849092688859288?s=20
- https://blog.solend.fi/liquidity-mining-2-0-phase-2-4938edcd1a4c
- https://blog.solend.fi/retroactive-liquidity-mining-e917311143bf
- https://dune.com/Marcov/Convex-Finance
- https://x.com/0xSolarcurve/status/1398590535745552384?s=20
- https://x.com/defipulse/status/1267097016586375170?s=20
- https://governance.aave.com/t/arc-continue-liquidity-mining-program-on-aave-v2-ethereum-market-and-introduce-liquidity-mining-on-aave-arc-market/7189
- https://governance.aave.com/t/arc-extend-aave-liquidity-mining-rewards/4852
- https://governance.aave.com/t/arc-round-5-liquidity-mining-aave-v2-ethereum-market/8380