An assorted thread of DeFi data analytics problems I've been kicking around in an attempt to answer the general question:

What constitutes as actual healthy DeFi protocol fundamentals growth?

Apologies in advance for the very abstract, dense rhetoric that sounds like I'm going on some existential rant. This is very much a long string of thoughts - not a formal dissertation.

That said, I look forward to hearing thoughts from those who make it to the other side!

Something to keep in mind while sorting through this thread:

Maybe a lot comes down to how you define the difference between growth & adoption, or if they're intersecting sets, or if one is a subset of another. But even then, what points to the best, most reliable metric?

1. A lot of DeFi tokens' liquidity are mostly paired to ETH on an AMM LP.

If ETH moons and no ETH/ALT trade occurs, the valuation of the ALT naturally must rise proportionally as well.

No one had yet decided that the ALT became more valuable, it just happened due to the math.

Following that, the valuations can just tend to be reflexive off the notion that "my ALT is pumping" when in reality no one may really be buying in, until someone does actually believe it and executes - usually to buy more.

A self-fulfilling price prophecy memed by ETH value

This is well known within the community when looking at individual token valuations, but I hear significantly less about this conundrum when we start looking at macro DeFi and its metrics.

Volume, TVL, Fees, etc all have some inherent inflation because of this.

So should these metrics be denominated ETH terms for better accuracy in real protocol fundamentals growth?

2. Take a typical fundamentals metric like TVL in a lending protocol or AMM. A bunch of DeFi assets could be sitting in the protocol. Suddenly some ALTs (not ETH) - instantly inflating TVL, and also inflating fees & volume after a trade/interest fee accumulation/etc.

These numbers/metrics all go up, but did the protocol itself provide any *extra* value if the assets sitting on it just inflated in value - for something the protocol had next to nothing to do with?

Denomination in ETH does not catch this if ETH itself did not pump.

So should these metrics be denominated in each asset's own self-value in some "normalized" sense?

Would this "normalized" approach basically be an asset-by-asset net-inflow (compared to prev day/week/month) metric?

Right here, you realize that the asset-by-asset value normalization literally just means the base number of tokens, not the underlying $ or ETH value.

So, we find ourselves at token net TVL, fees, volume, etc. I'll call this "net-token metrics" henceforth.

3. But then the small-value/new assets in the protocol metrics would show massive net-token metrics compared to other large-value assets - without necessarily moving the needle within the overall intuitive protocol fundamentals.

So... what if these net-token metrics were weighted by their overall % value in the protocol from the previous hour/day/week/month (depending on how precise of a time period you want)?

Ex: Take a protocol that last month had only 2 assets: 75% value in ETH, 25% value in LINK.

This month, we saw a TVL net-token growth of 50% in LINK, and a TVL net-token loss of 10% in ETH.

Weighted net-token TVL growth = prev_value_weight*(1+new_growth)

So, that results in:
ETH monthly net-token TVL growth = 0.75*(1-0.10) = 0.675
LINK monthly net-token TVL growth = 0.25*(1+0.50) = 0.375

Add the two net-token TVL growths together to get your protocol agg net-token TVL growth! (100% or 1.00 is no change)

0.675+0.375 = 1.05, or 5% overall protocol growth!

What did we just do? We managed to capture TVL on value-weighted basis, while not being sensitive to the token PA in $ or ETH terms!

This could also be extended to the Volume and Fee metrics as well, this is just an example.

What's it good for? Well, isolation from PA in $ or ETH terms (or just normalization) allows us to see true value-weighted adoption, in this example the form of the TVL metric.

To me, these "weighted net-token metrics" get much closer to the real core of protocol fundamentals - it isolates out the noise of massive crypto/underlying asset volatility.

Adoption metrics in the form of address counting isn't Sybil-resistant, but these metrics are!

Now, there's much further for me to go. Many more existential-like thoughts. But I'm gonna take a break for now and let both you and myself stew on this for a bit.

What's next on my mind:
-Native token metrics vs non-native token metrics
-Organic vs subsidized growth metrics in the form of token emissions (aka yield farming, liquidity mining, etc)

Give me some feedback - I really want to hear what you have to say! Chances are, I may have had a lapse in thought somewhere.

Also, I'm like 95% sure there's a TradFi-like equivalent to what I've done with Weighted Net-Token Metrics, but I can't remember it.

Please cc all the potential fundamental analysis big brains out there, I want to hear input from them all!

@OnurESucu @Shaughnessy119 @n2ckchong @santiagoroel @MapleLeafCap @SamuelShadrach4 @0x_Osprey @DeFi_Ted @0xBoxer @dcfgod