Discussion Post: Chainlink Transaction Value Enabled (TVE)

Oracles and Data
,
1

Tldr; Chainlink Labs reports TVE of ~7tn YTD [1]. Chainlink Labs seems to overstate its TVE by ~13x, leaving an unexplained $6.45tn TVE discrepancy. We highlight and respond to 2 possible errors in our methodology and 3 unconvincing responses thus far. We hope to motivate a more sensible and transparent discussion of Chainlink TVE. This is a living document, and we welcome feedback/comments.

Research Question: Do Chainlink Labs’ TVE figures accurately represent Chainlink’s true TVE?

Screen Shot 2022-12-20 at 2.04.53 PM

Definitions:

  • Transaction Value Enabled (TVE): TVE is the total US dollar-volume of transactions facilitated by Chainlink. TVE includes Avalanche, BNB Chain, Ethereum, Fantom, Gnosis, Heco, Optimism, and Polygon.
  • Total Value Secured (TVS): TVS is the total US dollar value of assets deposited into applications secured by Chainlink oracles.

Methodology:

  • We should expect similar ratios of Chainlink TVE / TVS to DeFi Volume / TVL.

    • TVE should be proportional to DeFi Volume because it reflects the aggregate sum of volumes for transactions invoking Chainlink oracles.
    • TVS should be roughly equal to DeFi TVL since it is the aggregate sum of TVL for applications using Chainlink.

  • We can therefore estimate Chainlink TVE by calculating Chainlink TVS x DeFi Capital Efficiency, where DeFi Capital Efficiency = DeFi Volume / DeFi TVL.

  • Note 1: In fact, the ratio of Chainklink TVE/Chainlink TVL should likely be lower than the ratio of DeFi Volume/DeFi TVL. While TVS includes all TVL for applications consuming Chainlink Price Feeds, Chainlink’s TVE is only a small fraction of total volume for price feed-consuming applications. These applications tend to use only a few Chainlink price feeds, and they may even use several different oracle providers.

  • Note 2: We are solving for expected Chainlink TVE. We know Chainlink TVS and DeFi TVL with reasonable confidence; however, DeFi Volume is somewhat trickier.

Results:

  • Chainlink TVS: Chainlink’s YTD average TVS is $23bn [2].
  • DeFi Capital Efficiency: First we narrowly define DeFi to include DEXs and lending protocols. This will simplify our volume and TVL calculations.
    • DEX Capital Efficiency = YTD DEX Volume / Average YTD DEX TVL = $1,250bn / $36bn = 35 [3], [4].
    • Lending Capital Efficiency = YTD Lending Volume / Average YTD Lending TVL = 8 [5], [6].
    • DeFi Capital Efficiency = (YTD DEX Volume + YTD Lending Volume) / (Average YTD DEX TVL + Average YTD Lending TVL) = ($1,250bn + $206bn) / ($36bn + $25bn) = 24.
  • Chainlink TVE: We can now solve for TVE using Chainlink TVS and DeFi Capital Efficiency:
    • Chainlink TVE = Chainlink TVS x (DeFi Volume / DeFi TVL) = $23bn x ($1,456bn / $61bn) = $549bn.
    • Therefore, we should expect Chainlink TVE in the ballpark of $549bn YTD.
    • Said differently, if Chainlink represents 38% of average DeFi TVL, then we should expect that Chainlink TVE equals something like 38% of YTD DeFi Volume. This yields $549bn.

Conclusion:

  • Chainlink Labs’ reported TVE of ~$7tn is nearly 13x the expected $549bn in TVE YTD. In other words, expected TVE is only 8% of reported TVE. The difference between Chainlink’s reported TVE and expected TVE is $6.45tn!
  • Reported TVE reflects 5x the aggregate YTD volume of DEXs and lending applications.
  • Chainlink’s reported TVE implies Capital Efficiency of 292. This is 12x what we should reasonably expect given DeFi Capital Efficiency of 24.

Possible Errors:

  • DeFi Volume: It’s possible that we substantially underestimate DeFi Volume.
    • We narrowly define DeFi in terms of lending and DEX protocols. These categories represent 67% of total DeFi TVL, and 65% of Chainlink TVS. However, for the overall DeFi Capital Efficiency Ratio to equal Chainlink’s Capital Efficiency Ratio of 292, the Capital Efficiency Ratio for non-DEX and non-lending DeFi applications must exceed 836. This is nearly 35x the Capital Efficiency Ratio for DEXs and lending applications. While it’s possible that other DeFi categories exhibit higher Capital Efficiency Ratios, it seems unlikely that we will find ratios sufficiently large.
    • Popular estimates of DeFi volume can differ. For example, Venus Protocol’s YTD Borrow Volume is $39bn according to Messari vs. $7bn according to Token Terminal [7]. Moreover, Borrow Volume only represents 22% of all Venus Protocol volume according to Messari [8], which means Messari’s Total Volume estimate for Venus Protocol is 30x Token Terminal’s Borrow Volume estimate. This is particularly important given that Lending TVL represents 63% of Chainlink TVS. We are confident that TokenTerminal’s data is more accurate than Messari’s, especially as Messari’s Protocol Metrics dashboard is still in beta, but is there a better source for DeFi Lending Volume that explains $6.45tn in missing volume?

(So far) Unconvincing Responses:

  • Derivatives: In theory, DeFi derivatives could explain much of the $6.45tn discrepancy. dYdX is the largest DeFi derivatives platform and a Chainlink Price Feed consumer. However, Chainlink enables a very small percentage of dYdX volumes. dYdX only uses one Chainlink Price Feed, LINK/USD, which represents $2.52bn of $457.14bn (0.55%) in dYdX YTD trading volumes. Therefore dYdX’s consumption of Chainlink Price Feeds can only explain ~0.05% of the $5.22tn TVE discrepancy [9].
  • CEX Volume: CEX spot and futures volume is >$27tn YTD. The $6.45tn discrepancy would disappear if it were the case that Chainlink enabled 24% of CEX spot and futures volume. However, Chainlink CEX TVE is negligible given that Huobi, Coinbase, Gemini, and Bybit only use 7 feeds collectively [10]. Moreover, Binance, which represents 63% of monthly spot and futures volume, is not a user.
  • NFTs: NFT sales volume is ~$21bn YTD, so NFTs could only explain at most 0.33% of the discrepancy, assuming Chainlink were responsible for enabling 100% of NFT sales volume [11].
7 Likes
2

Appreciate the interest in the Total Value Enabled (TVE) metric! James here from the Chainlink Labs Data Science team.

In terms of methodology, TVE is measured as the total US dollar volume of transactions enabled by Chainlink oracles, specifically every transaction where a Chainlink Price Feed reference contract was called across 11 blockchains Chainlink is deployed on. TVE is calculated across millions of transactions using on-chain data and resolved across thousands of token prices. These transactions include a wide variety of DeFi functionalities beyond trading and lending that require Chainlink oracles for accurate price updates and that account for a significant portion of TVE, such as rebalancing, automated yield farming, and stablecoin minting.

Some of the assumptions in your analysis are considerably narrowing the scope of what TVE measures, which is resulting in the perceived discrepancy that you’ve noted, particularly the assumption that TVE should be proportional to DeFi Volume. TVE encompasses much more than DeFi volume, measuring not only input and output values but how Chainlink oracles are used within transactions to facilitate value securely. More info: Chainlink FAQs | Chainlink

5 Likes
3

Thanks for your response, James!

I’m a firm believer in “trust but verify,” as I’m sure you can appreciate given Chainlink’s mission of trust-minimization. Is it possible for you to share any or all of the following?

In order of least to most demanding:

  1. The FY22 TVE figure for a single protocol (e.g., Synthetix).
  2. A breakdown of FY22 TVE by DeFi category (e.g., DEXs, lending/borrowing, derivatives) or a subset thereof.
  3. Historical TVE figures for FY19, FY20, and/or FY21.
3 Likes
4

This seems like a great opportunity for the web3 industry at large to level set on terms/methodology. With seemingly two different approaches to calculating TVE, it would be interesting to dig into the assumptions and rationale for each approach. What is the value here of taking the more narrow approach as presenting in the original post versus the approach being put forward in the response? Are there other approaches?

I’m interested in this discussion both as it specifically applies to the OP analysis, but also analysis in general. My initial thought here is that a more narrow approach might be easier to track and verify whereas a broader approach might be significantly more difficult to validate but a “truer” measure of the TVE concept.

1 Like
5

Thanks for your response! I’m glad you similarly find this discussion important.

To clarify, I agree with Chainlink Labs’ definition of TVE, but I believe their figures are overstated (maybe due to a bug in their code…) based on my approach to ball-parking true TVE.

In a perfect world, I would scrape TVE data on-chain (as Chainlink Labs might). In a less than perfect world, I would ballpark TVE based on all categories (i.e., including rebalancing, automated yield farming, etc. in my calculation).

Our world is neither of those, so I ballpark TVE based on DEXs and lending/borrowing. I think this is reasonable since rebalancing, automated yield farming, and stablecoin minting explain a small fraction of total TVE compared to DEXs and lending/borrowing. I eagerly await evidence to the contrary.

Chainlink Labs should open source their TVE methodology so the community can double check it. Trust-minimization is paramount.

6

Btw TokenTerminal recently added a dashboard for DeFi GMV (the most expansive definition), which apparently totaled $9.78T in FY22. Given that average Chainlink TVS was 25% of DeFi TVL in FY22, that implies Chainlink facilitated at most $2.45T in TVE, leaving a discrepancy of $4.55T (65%) unexplained.

Of course, Chainlink oracles likely enabled substantially less than $2.45T in TVE.

While $2.45T in volume might have flowed through protocols that use Chainlink price feeds in some capacity, the reality is that Chainlink price feeds are only used in a fraction of these transactions (and therefore responsible for securing only a fraction of that value), since these protocols often only use a couple of Chainlink price feeds.

For instance, it seems likely that Chainlink Labs is counting 100% of dYdX volume in Chainlink TVE, but as mentioned previously, dYdX uses one single Chainlink price feed, LINK/USD, which represents only 0.55% of dYdX YTD trading volumes.