Product Performance - Surfacing Gas Costs

Authors: @cavalier_eth, @dylan

Summary

Gas costs for Index Coop product launches & product maintenance are currently primarily paid for by Set, who will not be able to pay for them forever. Currently, gas costs are not considered by IC or Methodologists, obscuring the true performance of our products, and misaligning incentives.

This post outlines:

  1. A breakdown of the current costs incurred for product launch & maintenance
  2. The work in progress to solve the current issues across the Coop, Set and methodologists
  3. A plan and IIP to reimburse Set for future-looking product launch & maintenance costs
  4. Suggested future work that may help mitigate maintenance costs

The goal of this post is to share the process under way on this important issue, and the plan to resolve it. The proposed framework of forum post + IIP has been shared the with the WG leads.

Motivation

During the early months of Index Coop, there was no clear agreement on who would be liable for gas costs, and as the performers of rebalances and in the spirit of a startup, Set Labs bore those costs. In the months since, AUM and product maintenance costs have grown considerably, and Set is no longer able to responsibly bear the cost of rebalances.

The cost of rebalances is not currently considered, or paid for, by methodologists or Index Coop, and so is currently unaccounted for in profitability assessments of our products. For Index Coop to launch and maintain profitable products, the true costs need to be known, transparent, and shared with methodologists.

Product Launch & Maintenance Cost Breakdown

1. Contract Deployments

Over the lifetime of a product, contract deployments should comprise a relatively small piece of costs. Every product that the Index Coop launches requires a minimum number of contract deployments on mainnet. If a product uses sufficiently novel functionality, a series of contracts may also need to be deployed on test nets and mainnet for thorough testing & review.

For a single standard product launch with no novel functionality, contract deployment costs are somewhere between 2 to 4 ETH (8,000 to 16,000 USD) depending on the gas environment.

2. Automated Index Rebalancing Costs

After contract deployment costs, automated Indices like ETH2x-FLI and BTC2x-FLI have maintenance costs from daily rebalancing events and emergency deleveraging events.

Daily Rebalancing Events

Our current automated index products (ETH2x-FLI & BTC2x-FLI) define a rebalance interval of 24 hours. This means that the product will, at minimum, rebalance the amount of collateral posted/borrowed on Compound on a daily basis. These daily rebalances cost anywhere from 400 USD to 800 USD per day depending on the gas environment.

Emergency Rebalancing Events

Emergency deleveraging events are required during periods of extreme price volatility. Periods of extreme price volatility typically coincide with high gas volatility, so costs can be extremely difficult to predict. Transactions have been submitted with up to 2000 gwei in gas fees in the past. High gas costs must be tolerated because if these products fail to rebalance, they face risk of liquidation.

When emergency deleveraging is needed, rebalance bots will submit several rebalance transactions in succession. In extraordinary cases, a ripcord must be pulled to prevent imminent liquidation. May 19th 2021 was the first and last time Ripcord was pulled for ETH2x-FLI. Set Labs spent more than 50k USD rebalancing ETH2x-FLI and BTC2x-FLI that day.

3. Composite Index Rebalancing Costs

In contrast to Automated Indices, Composite Indices (DPI, MVI, BED, DATA) have maintenance costs more severely affected by component token DEX liquidity than price volatility. The primary metric that drives Composite Index rebalancing costs is the total number of trades required to rebalance the product over a given period of time. Three factors primarily drive rebalance trade counts:

DEX Liquidity Moving

This is by far the biggest risk for rebalancing composite indices, the source of greatest stress at Set, and an area for Index Coop to level up it’s understanding and become true experts. Set’s current rebalancing system is architected to support Kyber, Balancer V1, Uniswap V2/V3 and Sushiswap. If a component’s liquidity moves to another DEX we do not support, it can become extremely costly to rebalance a component token. Liquidity moving to a Layer 2 or side chain is even more difficult, as arbing against L1 DEXs becomes much more risky without the ability to use flash loans. There are long term ways we can mitigate this, but being able to track and anticipate liquidity movement risk must be an essential competency of the Coop.

Token Additions & Removals

Any time a token is added or removed from a Composite Index there is a high likelihood of a large number of trades being required to rebalance the product. Most components in a composite index only rebalance their allocation slightly month to month. Adding or removing a token requires trading in to or out of a 0% allocation which likely consists of a higher delta in asset allocation than otherwise.

As an aside, any new token additions also represent an ongoing & compounded risk of “DEX Liquidity Moving” from above. For a Composite Index composed of 100 different tokens, the risk of DEX liquidity moving on any one of them presents a risk to the maintenance cost & profitability of the entire product.

Rebalance Cadence

Simply put, products that rebalance more often will typically have higher maintenance costs. DPI, MVI, and BED rebalance on a monthly basis whereas DATA only rebalances on a quarterly basis. DATA’s single quarterly rebalance may require a larger number of trades than any single DPI or MVI rebalance but in total we expect it’s rebalance costs to be significantly lower.

A single MVI or DPI rebalance typically requires 100 - 200 trades costing 20,000 to 40,000 USD in ETH. In total, automated & composite index rebalances have cost more than $350k over the past 12 months with the trend continuing upwards.

Work Currently In Progress

For Index Coop to build a 100 year organization, the products we launch, their maintenance cost risks, and their paths to profitability need to be well understood. This will require concerted effort across Index Coop Working Groups, Methodologists, and Set.

Better Understanding Current Costs & Liquidity Risks

EWG and Wizards have created a tool to better understand token liquidity profiles for composite indices. This is under active development and will be used by PWG, Methodologists and Set to better understand how liquidity profiles are changing leading up to and on the day of a rebalance.

AWG is working on setting up analytics to track product revenue relative to product maintenance costs over time to get an accurate picture of the path towards true product profitability.

Working With Methodologists

PWG is working with methodologists to share data and tools on rebalancing costs, shifting liquidity profiles, and providing continual feedback on projected costs of token additions, removals, or significant reweightings. They are also exploring ways product methodologies can be updated to help reduce product maintenance costs.

Any new fee split agreements will be post-gas, so that incentives are aligned, and the split is of the “gross profit” of the product. It’s worth noting that IC pays for all gas costs until products are breakeven.

Increased PWG Ownership

PWG is taking the lead role in rebalance discussions with methodologists rather than Set. It will also be conducting product performance reviews using data gathered from AWG and Set.

PWG has at its disposal a handful of levers that can impact Composite Index rebalance trade count outside of DEX liquidity conditions and methodologist target asset allocations. It is actively working on better understanding the relationship between product performance and profitability with these levers. Those are:

  • Slippage tolerance on rebalance trades
    • This has been historically set to 50 bps, but was increased to 100 bps or more for the October rebalance.
  • Maximum gas price on rebalance trades
    • This has been historically set to 100 Gwei, but was increased to 150 Gwei for parts of the October rebalance.
  • Rebalance kickoff date and target duration
    • This has been historically the first business day of the month, but was extended for parts of the October rebalance.

As an aside, there’s a conversation to be had here about what liberties PWG can take when implementing methodologies to save on costs while faithfully reproducing a defined methodology for our customers. There is also an open question on whether or not that discretion lies solely with PWG or with Methodologists. To date, these decisions have been made by both parties together.

Working With Set

Set is providing historical rebalance data to Index Coop teams to compile and learn from. Set is also continuing to improve existing rebalancing infrastructure to increase the flexibility of rebalance parameters, and improve monitoring, so that rebalances can be performed over longer periods of time with less manual oversight required.

EWG is not yet in the position to consider in-housing rebalancing due to the complexities of the process and the expertise & organizational maturity needed to react to unexpected black swan events (e.g. May 19th). However, in the long term it is essential that EWG owns the product rebalancing process. This has been raised in the Set → Index responsibility transition.

For the time being Set will take only an operational role in executing rebalances with slippage tolerances and max gas prices being decided by PWG.

Product Maintenance Reimbursement Planning and IIP

Finance Nest is working with Set to design a product launch & maintenance reimbursement model for the existing rebalancing system. This system of reporting & reimbursement will be public so any member of the Index Coop can audit gas costs independently.

An IIP is in the works to propose reimbursing Set for rebalancing gas costs each month from December onwards. The IIP can be expected in the week beginning Nov 15th.

Future Work & Maintenance Cost Mitigation

Some longer range options are available to help mitigate product maintenance costs across the board. These include:

1. Moving to an Auction Based Rebalance System for Composite Indices: Moving to an auction based system may significantly reduce the amount gas costs required to rebalance a product, potentially at the detriment of product performance. In this scenario, arbitrageurs are incentivized to build integration to new DEX protocols & impact of the current gas fee environment is also baked into bids.

2. Funding Leveraged Indices via COMP or AAVE Rewards: For Leveraged Indices, gas costs could be socialized by claiming COMP or AAVE rewards gained from each product’s utilization of lending protocols, or increasing fees.

3. Moving existing products to post-gas fee splits: Launching any product without a post-gas fee split agreement is sure to produce conflicting incentives. For products that are already live, shifting to a post-gas fee split will better align incentives. Methodologist and Index Coop’s vision of what constitutes a profitable and sustainable product should converge with the adoption of post-gas fee splits across the board.


This week we’ll be working on the IIP and invite you to share any commentary and feedback

15 Likes

Many, many thanks Cav. This will be very helpful for Methodologists to design more profitable projects. @prairiefi and I will update the product profit model to include the gas fee split arrangement, the move to 100 bps trades, and adding/removing components.

I have one suggestion. I think we have a duty to disclose the NAV decay of our composite index products to our customers. With the move to 100 bps trade sizes (which I support), the NAV decay will now be substantial. For example, for a project with 25% monthly turnover, the NAV decay will 0.25%/month or 2.96%/year because the decay is compounded. This is much larger than a typical streaming fee. For easier communication the NAV decay could be combined with the streaming fee into an “Expense Ratio” as is typically done for equity funds. For the above example, if that product had a streaming fee of 1.95% then the Expense Ratio would be 4.91%/year.

6 Likes

Are there plans to add support for DEX aggregators like Matcha?

This is incredible - great work team! Looking forward to seeing continued improvements and development of this tool!

Completely agree - this is the way it should be.

Can you elaborate on what this means? Have PWG and EWG considered pulling rebalancing with both CEX and DEX liquidity?

I agree with this. To me, this should increase the urgency of implementing mint/redeem for composite index products which should both decrease costs to holders and increase product revenue by ~50%.

cc: @DarkForestCapital @verto0912 @snasps

7 Likes

I support splitting at the “gross profit” level but it’ll be important to specify which costs are included in calculating “gross profit”. Does it only include gas-related costs (rebalancing and recomposition) or also, for example, LM incentives and market-making fees etc paid for by IC? If the latter, then most projects are unprofitable during that period and it will be important to manage Methodologists’ expectations that they won’t receive any income during that period. That’s fine but it just needs to be clear.

Also, a product can be profitable and then have an unprofitable month, for example, due to a recomposition or reintroduction of LM incentives. In that case, is the product still considered “profitable” for the purposes of the fee split? In other words, is profitability determined once or on a month-by-month basis? Again, either is fine but I can imagine a lot of teeth-gnashing if it isn’t clear upfront.

4 Likes

I’ve modeled what I understand this would look like assuming “gross profit” uses the “all-in” product costs. This is for a typical product (not JPG) with $2M AUM, 25% turnover, 100 bps monthly rebalancing, $4k/mo LMI for 12 months, 3% streaming fee, and 70/30% profit share. The regular dips are for adding or removing a component every 4 months. There may be other costs I’m unaware of.

4 Likes

I understand why ETH/BTC2x-FLI needs daily rebalancing but is there rationale why other indices need Monthly Rebalancing ? We had the same rebalance cost issue on a Tradfi fund and received plenty of push back to move to quarterly rebalance.

6 Likes

The short answer is ‘no’. We’re working on a set of recommendations to improve product profitability and one of them is to rethink the rebalancing frequency.

4 Likes

I was told we can’t use Matcha (or any other order aggregator) for rebalancing even though it would be less expensive b/c the route selection is off-chain. I don’t understand the reasoning considering that Set uses Matcha by default for their other products, we can use Matcha for exchange issuance, and the rebalancing process is already off-chain. Maybe @overanalyser or @edwardk can comment.

2 Likes

My understanding (!!!) is that we are limited to DEX liquidity. The goal is to make rebalances etc permissionless / automated. i.e. the methodologies would provide weights, the coop agrees and uploads, then keepers / bots / (/ etc) implement.

Including a centralised party in that system is bad as it:

  • makes us less DAO like.
  • introduces a centralised failure mode.
5 Likes

The rational has been that crypto moves quicks / is developing at speed so we need to be able to adjust product composition.

$DPI started with monthly rebalance and a 6 month project life before inclusion. A number of tokens have been added close to 6 months while a 6 month + 3 month rebalance frequency would produce 9 months lead time.

In addition, token release / vesting / burn mean that weights are still being adjusted even without addition / removal.

$MVI, and $BED have Monthly rebalance.
$DATA launched with a quarterly schedule.

Other product proposals have requested a more frequent rebalance schedule.

5 Likes

My understanding (and I may be wrong here) is related to the failure mode:

  • Exchange issuance gets price quotes and issuance costs and compares them to the secondary market, then asks the user to confirm. If Matcha gives a stupid price the holder should notice and not proceed.
  • Rebalance is automated and designed to just run. If an external 3rd party falls over it won’t be clear until after that transaction (and we need to monitor and react).

Also, if issue / redeem fails / gives a bad price, then it’s that users problem. If a rebalance fails, it’s the coops problem.

The last $DPI rebalance was ~ $15 M of trades (~7.5% of the fund value) with a few over $500,000.

The trade off is between lower costs (either less gas or less NAV decay) or higher complexity / risk in using an aggregator.


Note, while sushi Trident will creat multiple pool types (xy=k, concentrated etc) with Tines acting as a trade router, I would expect that our rebalances would continue to use a single pool.

2 Likes

The more mature asset classes I saw had 3m and even 12m on some LT funds.

Totally get your point on Crypto and maybe hard to shoehorn a standard for all indices.

Increased DEX liquidity and possible economies of scale may mean pictures changes but the newer index’s whilst interesting probably increase the problem.

1 Like

I see. One automated long-term solution could be to get the quotes from an aggregator and then check and execute the orders on the underlying Dex’s. That’s an engineering project though so we can leave it for another day.

1 Like

I think the aggregators provide 3 key benefits:

  1. Identify the best pool to carry out a specific trade (due to differences between pools)
  2. Split trades across multiple pools when the reduced price impact out weights the gas.
  3. Acess market maker RFQ / side pools.

We have a trade splitter implemented for the FLI products, this hits multiple DEX pools in a single transaction. So #1 and #2 are possible. #3 can also have significant benefits. e.g. for $50,000 ETH to LRC (Still 1% lost on the trade):

2 Likes

Given we already monitor rebalances, how would monitoring rebalances done through an aggregator be different? Given the cost, to the Coop and the holders of our products, we might have a decent margin of error here.

The market marker point is an important one. Aggregators are the only way we can work with market makers, like Wintermute, on increasing liquidity for some tokens specifically around the rebalancing window.

Another option to reduce the number of trades is to do direct token swaps, instead of routing everything through ETH. So we can swap MANA for ILV directly if both have their primary liquidity on the same venue. 1x price impact, 1x slippage, 1x swap fee - instead of 2x all of them.

On NAV decay disclosure, yes, it needs to happen but we need data. NAV decay will depend on the actual execution of the rebalance. From price impact, to swap fees depending on whether 0.3% pool or 1% pool was used and actual slippage of those trades. Just saying NAV decay is 100bps is inaccurate, even though it’s probably better than not disclosing anything.

On monthly vs quarterly rebalancing, quarterly would be quite challenging to do for the metaverse. There’s just too much going on, a lot of new tokens get launched, liquidity goes in and out, etc. We wouldn’t quite be able to properly capture the theme with quarterly rebalances, especially if we can only make 1 addition or removal per rebalance.

9 Likes

To help inform the upcoming IIP, I’ve taken the liberty of listing the costs which will need clarification as to whether or not they’re considered direct costs, i.e., included in the definition of “gross profit”:

Included?
Rebalancing gas costs
Recomposition gas costs (ie, adding/removing components)

Unclear?
Liquidity mining incentives
Moving DEX liquidity

Not included?
Product-related contributor rewards
Gas fees for managing concentrated liquidity
Gas for arbitrage bot
Market maker fees
Impression mining rewards
Seed funding
Anything not listed above

For activities which are not initiated by the Methodologist (moving DEX liquidity or providing/removing LMI) there will probably need to be a decision-making system to avoid costs being incurred on a product against the Methodologist’s wishes and vice versa.

I don’t think we have monthly project finance reporting yet so the new arrangement will probably also require additional effort from Finance.

cc: @Thomas_Hepner @verto0912 @DarkForestCapital @Kiba @snasps @Matthew_Graham

4 Likes

1000% agreed on the need for data particularly since the decay will increase around recompositions. Since the NAV decay will be at least 100bps x Turnover perhaps we can start with a lower bound estimate until more data are available.

More frequent rebalancing is preferable but the tradeoff, as we all know, is the effect on profitability and value decay. We need better tools and data to evaluate these questions.

4 Likes

The first step is for us to define and understand the profitability of each product . Broadly we are comfortable with loss-making products as we are oriented towards growth, but need to make those decisions with eyes wide open, and consider at what point products should be sunset.


I won’t go into extensive detail on these, as more will come out shortly, but quick thoughts:

For clarity, this level of detail won’t be in the next IIP, it will only decide wether we should reimburse Set for gas fees.

[Personal perspective for now] I would agree

[Personal perspective for now] These would not be included in gross profit, but provided by Index Coops’ liquidity solution - ideally a Liquidity Pod with a fixed remit, that you can expect a post on very soon.

Happy to say this is on the way, lead by @ElliottWatts and a crucial tool for PWG and community to make informed decisions :slight_smile:

1 Like

Thanks for sharing the detailed review.

I have a few questions ;

  1. “Simply put, products that rebalance more often will typically have higher maintenance costs. DPI, MVI, and BED rebalance on a monthly basis whereas DATA only rebalances on a quarterly basis. DATA’s single quarterly rebalance may require a larger number of trades than any single DPI or MVI rebalance but in total we expect it’s rebalance costs to be significantly lower.”

Aside from the frequency of rebalancing, isn’t the liquidity of Tokens also a major contributor to rebalancing cost ?

  1. “Set is providing historical rebalance data to Index Coop teams to compile and learn from.”
    Where can we access this data?
1 Like

“Another option to reduce the number of trades is to do direct token swaps, instead of routing everything through ETH. So we can swap MANA for ILV directly if both have their primary liquidity on the same venue. 1x price impact, 1x slippage, 1x swap fee - instead of 2x all of them.”

second this