mTVL: Finding a North Star

TL;DR

  • organizations benefit from having a single metric that guides growth (i.e. the North Star metric);
  • for a yield aggregator, TVL has too many shortcomings to be that main growth metric;
  • by heavily calibrating TVL with performance, a strong candidate for NSM is born.

There is something powerful about having a single metric that measures the health of a product. The genius behind DeFi Pulse’s push for total value locked (TVL) was that market cap had become an imperfect measure of growth for the likes of Compound, Maker, and Uniswap, even post-token. TVL became the new North Star metric (NSM) of these DeFi protocols.

At Pickle, we face a similar albeit more nuanced dilemma. As a yield aggregator, there are quirks to our business model that make some refinement of the TVL metric necessary. For Pickle, TVL is money deposited in jars. Since we operate on a performance fee basis (for now), not all TVL is created equal, as jars perform differently. We need a way to tell apart this good and bad TVL in order to turn it into an NSM.

Let us first evaluate the properties of a good NSM:

  • the growth of this metric is a proxy for the growth in users who derive the most value from the protocol. Ideally, certain “power users” who for whom Pickle has significant “must-have-ness” are driving the NSM. Regular TVL isn’t discerning here.
  • should be a metric with no “upper bound”. We should be able to chart the metric and see it go upwards across time. TVL works like this. A modified TVL should retain this too.
  • it is a proxy for revenue. Unlike revenue, however, a good NSM should be a leading rather than a lagging indicator. All else equal, a focus on growing the NSM is more sustainable than focusing on growing revenue.

Given the above-mentioned properties, the NSM becomes a good focus for the core team, the community, and the mission of Pickle.

The NSM: monetizable TVL (mTVL)

Rather than focusing on the sheer value of the deposits made on the protocol, Pickle should focus on how those deposits are performing, which in turn is the source for the growth of each depositor’s position as well as the best way for the Treasury to grow. If Pickle does this at scale, then our protocol is truly making the most of those locked deposits.

Thus, we modify TVL by a factor that represents how it is performing, to arrive at the new TVL.

mTVL = TVL * performanceFactor

What is performanceFactor?

  • The performanceFactor is a multiplier that let us know what TVL is “good” and what TVL is “better” (the only “bad” TVL is 0) based on a target investment performance.
  • When investment performance is on target, performanceFactor = 1, meaning that TVL is weighed at face value.
  • When investment performance is below-target, that TVL is penalized (performanceFactor < 1), and when it is above-target, that TVL is rewarded performanceFactor > 1.

This can work with any target. I propose, however, that the target puts Pickle an order of magnitude above not using Pickle. That is, as Pickle’s main advantage is to auto-compound, when using Pickle to get an apy that is at least 10x the underlying apr, the requirement for “must-have-ness” is satisfied. We can then derive that:

performanceFactor = (apy/apr) / 10

Since,

apy = (1+(apr/365))^365 - 1

Then, we can see that for a performanceFactor of 1, we would need an apr of 365%, approximately. Pretty degen.

mTVL in action

Let us take a snapshot of current jars and use mTVL to analyse them.

While we can see that, the sum of mTVL and TVL appear similar, the picture looks quite different jar-to-jar.

  • In fact, the mTVL of mic/usdt is around 10x larger than that of 2nd place bac/dai, and on average 100x of other jars. Because of its very high APR, this jar’s underlying strategy makes using Pickle an extremely attractive proposition (must-have, indeed) and performance fees more than worth the benefit of robo-farming.
  • down to the revenue numbers, mic/usdt brings $12,000/day to our Treasury, and bac/dai some $10,000/day, whereas the other 3 large jars of 3pool, wbtc/eth, and renBtc bring $450/day, $115/day, and $21/day respectively.
    • since the mTVL of mic/usdt is 10x that of bac/dai, it appears as if this difference isn’t justified on the numbers. However, it is important to keep in mind that as these jars compound, they grow at very different speeds and so does our revenue growth. For example, in 30 days, mic/usdt would be contributing $18,000/day to the Treasury (vs. $13,000/day for bac/dai); in 90 days the mic/usdt would be contributing $45,000/day (vs. $20,000/day for bac/dai). All else equal, a jar like mic/usdt is more sustainably profitable and has higher built-in growth, so its value on the present should be much higher.
  • mTVL more clearly shows the futility of pursuing low-yield strategies. Under the current monetization model (performance fees only), jars that provide a little yield must have massive volume, in order to make decent contributions to the Treasury. Considering development and operational costs, we would be better served looking for high-yield opportunities. Should the monetization model change, the modifiers to TVL would change as well(e.g. , the performanceFactor could be tweaked — with a lower target — or another factor could be added to better calibrate TVL coming from large-caps seeking decent yield).

Final thoughts

deliver massive value to users
=> get sustainable (high) revenue & growth
=> consistent and significant positive price-pressure
=> (cont.) deliver massive value to new & retained users

An NSM like mTVL is measuring the long-term health and sustainability of the protocol as a whole. Together with well-designed value-capture mechanisms like the Smart Treasury and DILL tokenomics (vote-locking, boosted rewards, and staking rewards), a growth in the NSM means a growth in the fundamentals that drive the price of PICKLE upwards.

8 Likes

truly awesome initiative to define a north star for pickle.
some comments:

are we looking at customer-APY or treasury-APY (taking into account vested SUSHI)?

with the autocompounding “stacking” mTVL we run danger to become ever narrower in our focus on specific jars, maybe it makes sense to introduce some kind of decaying mechanism for this, because looking JUST as mTVL one could come to the conclusion to discontinue all but one Jars.

also take into account that there might be need to continue the “bread and butter” jars with some incentive as it creates “stickiness” - having one place to do your degen stuff while also place your “boring” liquidity still makes sense because you can earn all your incentives in the same token (and dont have to keep track of pickle, badger, sushi…).

It would be easy to see how many of our users use more than one product (# of different Jar LP tokens per ETH address) and which these are so we can adjust our offerings to attract a broad base of degen, and part-degen TVL. Cross-pollination between Jars etc.

2 Likes

Hey, amazing comments. Thanks for taking the time to come here and enrich this discussion:

For the exercise above, these are the customer-APYs, but not taking into account the removal of performance fees and swapping with SUSHI (under a tenous assumption they’re equivalent).

Yes, this can be mitigated with a better approximation of APR, for example, with a decay function as you mention. One may also change the “performance target” from 10x to something like 5x or even 2x. However, where’s the fun in that? :stuck_out_tongue: I believe setting realistic but tough targets (that you can hit 60-70% of the time is more useful for forging ahead.

You’re right in that we should continue these jars, under the present scenario. However, when seeking out new jars, increasing the NSM can always be seen, all else equal, as Pickle doing right for their users.

This is true as well, and under the current exercise, no assumptions are made about whether mTVL can be useful on which jars to reward. One could argue, high APY in itself negates the need for a high reward and that it is useful to attract low APY but “reputable” (or “boring”, as you say) large-cap deposits with high rewards. Conversely, one could argue the opposite. In fact, should we introduce AUM fees, we could calibrate the mTVL to make it more favourable to large-caps, and subsidize their APY to attract them. However, that is predicated on Pickle having strong value, so we end up asking the same questions that led us to define mTVL favouring performance.

We can experiment with both approaches and observe. I would argue that the mTVL is defining your new “bread-and-butter” benchmarks – for the users you want, and for what best serves sustainably income to the treasury.

Yes. Mind you, we will have other metrics besides the NSM. We can and will measure retention, user satisfaction, and user activation (including jars per wallet). These can all be part of a more comprehensive health scorecard. Adding that with UX research and informal conversations with users gives the team the whole picture for vision, mission, and roadmap. From there we build a product backlog that we tackle in deliverable increments at a regular cadence. What makes the NSM useful is its simplicity, its capacity to abstract a lot of complexity into something subtle, so we can discuss that instead of the whole picture when appropriate to do so and spend more time executing / less time analysing.

1 Like

I think the idea of productive TVL as a NSM is perfect. There are so many instances of projects that market their TVL as the be all and end all, but ultimately it can be just idle money. TVL that creates both competitive returns for the user and revenue for the protocolshould be held in higher regard than those that only fulfil one or other, or neither, of those goals.

Before this thread I would have argued that revenue rather than TVL be the NSM … however I like the fact that this approach highlights those strategies that are currently under-utilised and should receive greater focus.

3 Likes

LKJ, thank you for the extensive analysis and clear explanation of your thought process. This is valuable information and much appreciated. From a layman’s perspective my only question is if there is any value to include some form of risk adjustment to the numbers. the performance factor calc is a metric to measure return performance, and thus directly (i’m assuming) risk of the underlying etc. If we just focus on mTVL the numbers will spotlight the riskiest strat/underlying.

I’m curious if ‘stickiness’ also matters in the sense that some less risk-loving degens would tend to just park their funds in the ‘less-risky’/lower yielding strats (USD/ETH @41% is still impressive) for long-term vs faster money chasing returns. both valid strats for eking out max yield but just different styles.

again, thank you for your time to analyze and provide the opportunity for this discussion.

:pray:

2 Likes

Normally, the best thing when you want something to be considered is to measure it. If you want to reward less risk, to an extent, then we could consider adding a riskFactor, at the expense of perhaps overcomplicating the metric. High-yield strategies aren’t inherently more risky, even though in traditional personal finance that is usually the advice. It may be the two are correlated, but we do not know at what degree, so rather than using one as proxy for the other, we’d include both.

I think, from the latter part of your message, that your consideration may be rather us not excluding relatively high returns for strategies that use large-cap, high-reputation assets like bitcoin, ether, and the more prominent stablecoins. Certainly, we can. In fact, it’s something I alluded to myself in this thread. We have one strat under audit just like that, and once we get our house in order re: goals and best use of resources there will be room for at least one strat in every three we make to be “less degen”, I gather.