Enso Token Launch Auction Design: A Data-Driven Approach

Published in

EnsoFinance

19 min readJul 17, 2024

1. Introduction
2. 2017–2020 Dutch auctions
3. Performance comparison of Dutch auctions with Fjord LBPs
— 3.1 Fjord Foundry introduction
— 3.2 LBP data exploration
— — 3.2.1 Data quality and aggregates
— — 3.2.2 LBP record raises deep dive
— 3.3 Price discovery, long-term sustainability, and auction fairness: Dutch auction vs. LBP as token launch mechanism
— — 3.3.1 Coingecko, DefiLlama, Dexscreener data
— — 3.3.2 Dune data
— — — 3.3.2.1 Aggregates analysis
— — — 3.3.2.2 Analysis of LBPs with positive performance
— — — 3.3.2.3 Analysis of LBPs with negative performance
— — 3.3.3 Comparative analysis of Dutch auctions vs. LBPs for fundraisers in the range of $10M
— 3.4 User distribution in auctions
4. Conclusion
5. Appendix
— 5.1 Google sheet data collection process
— 5.2 Dune data process
— 5.3 Fjord Foundry team as LBP Curator analysis
— 5.4 Note on indexed pools on fjordfoundry.com
— 5.5 Sources
— 5.6 Links to the Google Sheet and Dune dashboard

Introduction

Enso is embarking on an exciting new chapter with our upcoming token launch. In preparation for this significant event, we analysed various auction types to ensure optimal price discovery, fair distribution, and a sustainable long-term strategy. Our goal is to provide a thorough and transparent analysis to our community, showcasing that our decisions are rooted in comprehensive data analysis and strategic planning.

In this article, you will find a detailed data analysis of the different mechanisms employed in past fundraisers and our rationale behind choosing the best auction type for Enso’s token launch. We believe that transparency and data-driven insights are crucial for the success and sustainability of our project, and we are excited to share this journey with our community.

This article focuses on Dutch auction data from 2017 to 2024. We show that historically, Dutch auctions were the most prominent auction type for token launches, with fair access and lasting success relative to other auction types like Fjord Foundry’s (formerly Copper Launch) Liquidity Bootstrapping Pools (LBPs).

We determined that Dutch auctions historically led to a +97% median performance from 2017–2020. In comparison, we found that the 2021–2024 Fjord Foundry curated LBPs have a median performance of -87.63% namely a buy-and-hold strategy of $100 into LBPs would be worth $12.37 today (July 2024). That buy-and-hold strategy deployed into these Dutch auctions would be worth $197 today, a 15.92x difference relative to the LBPs’ strategy.

Overall, curated LBPs saw losses of more than half their value, with typical performances ranging from -56.56% to -96.17% Only the top 10% of LBPs, based on performance, managed to break even or achieve a positive return, with performances of +0.30% or higher.

In aggregate, the median number of participants in Dutch auctions is about 5.3x higher than in curated LBPs. Looking specifically at Dutch auctions and LBPs in the same fundraising range, we see that on average, token holdings are distributed to 3.83x more participants in Dutch auctions compared to Fjord LBPs.

With a larger base of token holders, the influence of any single individual on the project’s price stability is reduced, thereby promoting a fairer distribution and contributing to more sustainable and resilient price action for all token holders.

2017–2020 Dutch auctions

In the table below we will find some of the biggest historical fundraises via a Dutch auction, held on-chain or on a centralized thirdparty:

All three of Gnosis, Polkadot, and Raiden Network were held on-chain with the Dutch auction deployed as a Smart Contract. Whereas, Algorand, Solana, Celo and Flow were held on Coinlist.

The median performance of all seven Dutch auctions to date is +97%, namely half of the Dutch auctions have a higher performance than +97% and the other half, is lower than +97%.

Note that because returns do not scale evenly, we can’t directly compare a -99% return with a +10,000% return. Large positive outliers (like +10,000%) can distort the average, making it less accurate for showing typical performance, which is why we mainly use the median.

In our case, due to the massive Solana performance, the average performance would not be a good metric as it is highly influenced by the extremes and does not provide an idea of the distribution of the performance in our data sample. The average would be +11,200% (i.e. 112x), which does not further our understanding of the Dutch Auction success relative to a +97% median performance.

The auctioned supply varied from a very low 0.25% for Algorand to as high as 50% for Polkadot and Raiden Network.

Looking at the Participants’ statistics we have the following observations:

The median Participant-count is 3,476
Most of the Dutch auctions Participants range from 1,801 to 4,190
The top 10% of Dutch auctions have more than 8,627 Participants

Performance comparison of Dutch auctions with Fjord LBPs

Our analysis relies on the LBP-auction-ending price, namely the token price of the last valid transaction during the LBP auction, as a benchmark to compute the LBP performance as of today. In a follow-up analysis, we will focus on each of the LBP buy and sell by parsing on-chain event logs to use the average price weighted by tokens bought as a benchmark. The latter should give a more realistic picture of the price at which holders acquired the tokens by the end of the LBP auction.

Note that in this comparative analysis, we do not take into account the following metrics:

Auction parameters: starting, ending and reserve prices, the resulting fully diluted valuations (FDVs) at each price point, the amount of supply auctioned, the price decay function (how fast the price descends), the token vesting schedule
Project-specific parameters: the fundamentals of the project, the underlying technology, the founding team, past fundraisers, blockchain on which it is deployed, if the project is infrastructure or application […]
The market state: the crypto industry maturity at auction time, the stage of the market e.g. bull vs. bear, the total market cap of the crypto industry […]

Fjord Foundry introduction

In 2021, the decentralized token launch marketplace Fjord Foundry (formerly Copper Launch before the September 2022 rebrand) launched the Liquidity Bootstrapping Pool (LBP) as a token launch mechanism. LBPs are price-descending auctions enabling users to buy and sell the token during the auction. The buying pressure can locally overcome the price decrease trend and increase the token value momentarily until the price reverts to decaying.

We do a comparative analysis of the Dutch auction performance relative to the LBPs. The count of curated LBPs scrapped from Fjord Foundry is 150.

We fetch today’s token prices (July 2024) from the Coingecko API, the Defillama API, or the Dexscreener API using the token contract address and blockchain. If none of the three APIs yields a match, we manually fill in missing data using the Coingecko website. We leave the field empty if the four solutions have no data available.

We define as Performance the rate of return of the Price from the auction-end to today, it is computed as follows:

Performance = Price_{today} / Price_{auction-end} -1.

This measures the percentage change in price from the auction end to today.

Curated LBPs, as opposed to non-curated LBPs, are the ones that are validated by Fjord Foundry curators from the applicant pool. As per the site wording: “Warning: Curation is a third party process but does not guarantee the long performance of a project. Always DYOR.”

We focus our comparative analysis on curated LBPs since they already went through a first due diligence process from the Fjord Foundry curators, thus making the comparison fairer relative to Dutch auctions, some of which were listed on Coinlist following a selection process.

Starting from September 2023, the Fjord Foundry team curated 6 LBPs themselves, shifting away from the approach of solely relying on third-party curation. The median performance to date of their recommended LBPs is -76.72%. You can find more details in the Appendix section Fjord Foundry team as LBP Curator analysis.

LBP data exploration

Data quality and aggregates

The following section gives an overview of the data of curated LBPs, when they were launched, achieving which fundraising levels, what their Performance is to date and how many Participants they had.

You can find the Google Sheets document aggregating LBP and Dutch auction data here.

With a sample of the first five rows here (sorted by descending Funds Raised ($)):

Given the count of LBP auctions grouped by years, we have the following table and observations:

The curated LBPs’ start dates range from November 2021 to June 2024.
The total sample is 150 curated LBPs,
41 of which have NULL performance namely the price as of today is missing from all three of Coingecko, DefiLlama, and Dexscreener APIs given the token contract address and chain, as well as from the Coingecko website.
This is usually because the token no longer exists or hasn’t seen any price action in months or years.
The majority of curated LBPs are in 2024.
2023 had very few LBP-curated launches.

Looking at the distribution of Funds Raised ($) across the years and LBPs with the following observations:

In 2021, a total of $462 million was raised, surpassing the combined total of all subsequent years
In 2021, the median funds raised were the highest by far, with half of the twenty LBPs raising over $12 million, and even the smallest raise exceeded the median of 2024
The significant increase in LBP count in 2024 likely contributed to a lower median raise compared to 2022 and 2021
2023 had the lowest median funds raised, with its median being almost half that of 2024

We have the following curated LBP Performance metrics and observations:

The performance of LBP auctions varies widely, with a range from -100% (total loss) to +27,507.20%. The DUEL curated LBP with a +27,507.20% performance is an outlier since we used the LBP auction-end price of $0.00002, which is 139.36x lower than its average weighted auction price of $0.0027872. The average weighted auction price is a more accurate measure of the token acquisition cost for holders. Using the average weighted auction as the reference price for DUEL would result in a performance of 2x to date
The 90th percentile performance is +0.30%, indicating that only the top 10% of LBPs managed to break even or achieve a positive return
The interquartile range, from the 25th to the 75th percentile, spans from -96.17% to -56.56%, highlighting that the majority of LBPs experienced losses amounting to more than half of the investment

We now focus on the Participants’ distribution across all LBPs and have the following observations:

The 2024 curated LBPs have 549 median Participants, almost one-third of the 2021’s at 1500 Participants.
2024 has the highest total Participants across all years, despite raising less than half the funds of 2021 with $218M against $462M. Note that we do not differentiate if the same participant bids in several LBPs.
2024 also had the curated LBP with the lowest amount of participants across all years with 20, 25x less than 2021’s minimum participants of 512.
Most curated LBP participants are between 224 and 1,490.

LBP record raises deep dive

See below the table depicting the top 20 LBPs ordered by the largest Funds Raised ($):

2021 saw the largest fundraises: $139M for GuildFi and $105M for Merit Circle, with a total of 13 fundraises above $10M in that year.

Participants range from 823 for Spool DAO to 9,712 for GuildFi. Analysing each auction transaction in the future analysis will shed light on the distribution of the holdings among this participant set. The median performance of the top 20 LBPs by Funds Raised ($) is -96.44% and the median number of participants is 3,374 in this sample.

We see that most of the record raises happened in 2021:

We observe that none of these thirteen LBPs that were raised >$10M in 2021 has a positive performance in our sample, with a median performance of -98.55%. This indicates that, for half of these investments, the performance was worse than -98.55%. To illustrate, if one invested $100, a median performance of -98.55% means the investment would now be worth only $1.45. To recover the initial investment, the current price would need to increase by 68.96x.

Price discovery, long-term sustainability, and auction fairness: Dutch auction vs. LBP as token launch mechanism

Coingecko, DefiLlama, Dexscreener data

We aggregated the 150 curated LBP pool data over the years and computed metrics such as the median performance across groups such as the whole sample, and buckets of funds raised for fundraises above $10M+.

We compare the five Dutch auctions that raised $10M+, namely Gnosis ($12.5M), Polkadot ($145M), Raiden Network ($33M), Algorand ($60M), Celo ($10M) as follows

with the 22 LBPs that raised $10M+ (from $10M to $139M):

We first observe that only one of the twenty-two LBPs has a positive performance in our sample, with a median performance of our sample of -96.08% (-84.47% average performance). This indicates that, for half of these investments, the performance was worse than -96.08%. To illustrate, if one invested $100, a median performance of -96.08% means the investment would now be worth only $3.92. To recover the initial investment, the current price would need to increase by 25.51x.

In comparison, the five token launches conducted via Dutch auction that raised $10M+, namely Algorand, Polkadot, Gnosis, Celo, and Raiden Network, have a median performance to date of -41.23% (+188% average performance).

The table below details metrics of LBP performance across the years including for the $10M+ bucket:

We observe that the 2021 LBPs have a median performance of -99.16%, namely for $100 invested, the current value of the investment would be $0.84, and the underlying token would need 119x to break even.

Further, 2024 has seen 93 curated LBPs in this sample, more than all the previous years combined.

Dune data

In parallel with the above findings, we wanted to confirm our previous set of results with Dune data. We uploaded our same raw dataset to Dune, and then fetched and cleaned the most recent secondary market price available in these three tables:

prices.usd, dex.prices, and dune.dune.result_dex_prices_evm_daily.

You can find the Dune dashboard here.

Aggregates analysis

As seen in the table of aggregates below, and given the 125 pools (out of 150) for which we have price data:

We confirm a similar order of magnitude for LBP performance, namely the median performance sourced with Dune data is -87.63% across the 125 pools, versus -89.46% across all 150 pools for the pools with today’s prices matched with the Coingecko, DefiLlama, or Dexscreener API.

Out of 150 LBPs to this date, we see that 13 pools (8.66% of the sample) have a positive performance as of 2024–07–01.

Among these 13 pools, 5 raised more than $1M, and 8 raised less than $1M. The latter group usually largely outperforms due to a small auction-end FDV, for instance, AxonDAO Governance’s token launch led to a +7,533% return (75x) to date from a $118k raise at $2.3M FDV in January 2024, with a current FDV of $183M as of 2024–07–02.

Analysis of LBPs with positive performance

For axis-scaling purposes, we plot below the LBPs with positive and negative performance separately.

The chart below depicts the Fjord LBP pools with positive performance, with the start date of the fundraiser on the x-axis and the performance (%) on the y-axis. The dot size is scaled with the raised funds, the bigger the dot the bigger the raised funds.

The dark blue dot at the top right is $DUEL with a +27,890% (278.9x) performance for $868k raised. In practice, it is closer to 2x if we were to take the average price weighted by the tokens bought using the @GoodData_BadGuy Dune dashboard: the average price at which holders bought $DUEL during the sale is $0.0027872, and given the latest price of $0.005598 on 2024–07–01 12:54, the performance is on average +100% (2x) instead of 278.9x.

AxonDAO did a +7,533% performance for a $118k raise while Autonolas did a +1885% performance for a $547k fundraiser. The remaining 10 tokens have a performance varying from +16% to +503%.

Analysis of LBPs with negative performance

The chart below depicts the remaining 112 LBPs with negative performance, with time on the x-axis and Performance (%) on the y-axis, and where the dot size is scaled with the raised funds, the higher the funds raised the bigger the dot size.

The two biggest dots, in blue for GuildFi ($139M raised), and orange for Merit Circle ($105M raised), have negative performances of -94.26% and -48.57% respectively. Polkadot, which raised $145M via Dutch auction, has a performance of +943% in comparison, as seen in the table below:

The above results indicate that relative to LBPs, Dutch auctions lead to better price discoveries given the sustainable price action years after the TGE and the positive performance to date, even for very large fundraise ($100M+). Note that the count of public fundraises above $100M is small (3), thus we are constrained to compare the Polkadot Dutch auction of 2017 with LBPs from 2021, despite the expectedly different market microstructure and crypto industry maturity.

Comparative analysis of Dutch auctions vs. LBPs for fundraisers in the range of $10M

We now explore results for fundraisers in the $10M range.

The following table highlights the performances of Flow and Celo, each having raised around $10M in 2020 via a Dutch auction on Coinlist, with the three Fjord LBPs that raised similar amounts:

We see that the median auction performance for comparable raises (between $9M and $11M) for the Dutch auctions of Celo and Flow is +28.07%, against a median performance of -95.34% for all seven LBPs (one that has no data since the token cannot be found).

Based on this data sample, we conclude that the price discovery enabled by Dutch auctions relative to other mechanisms like Fjord LBP, leads to more sustainable price action in the long run for comparable token launch sizes, here for the $10M range.

User distribution in auctions

The table summarizes the distribution of participant numbers in Dutch and curated LBP auctions, highlighting key percentiles and extremes:

Median participation difference: the median number of participants in Dutch auctions (3,476) is approximately 5.3x higher than in curated LBPs (655).
25% to 75% range: most auctions fall within the 25th to 75th percentile range. For Dutch auctions, this range is 1,801 to 4,190 participants, while for curated LBPs, it is 224 to 1,490 participants. The range for Dutch auctions is significantly broader, showing a larger and more varied participant base than curated LBPs.
Minimum Participants: the minimum number of participants in Dutch auctions (650) is substantially higher than in curated LBPs (20), indicating that Dutch auctions generally attract more participants at their lowest levels.
Top 10% participation: the top 10% of Dutch auctions have more than 8,627 participants, nearly three times higher than the top 10% of curated LBPs, which have more than 2,974 participants. This shows that the most popular Dutch auctions attract a significantly larger crowd than the most popular curated LBPs.

Using the same $10M token launch table from before and focusing on Participants:

We note that the Dutch auctions of Celo and Flow had an average number of participants of 8,498, against an average of 2,214 participants for the seven LBPs, namely 3.83x as much for participants for Celo and Flow, which reached a wider range of participants.

Since Celo and Flow were both listed on the centralised platform Coinlist, which requires undergoing a KYC procedure to access it unlike Fjord LBP which is a decentralised platform, we assume that this participant count is more robust to sybils relative to Fjord’s. We expect that the distribution of tokens over a wider range of unique individuals leads to a fairer distribution, namely one that is more uniform versus concentrated by a few participants that can be sybils. As a result, since holdings are more equitably allocated, we anticipate better outcomes for the community with a long-term sustainable price action since a single or a handful of individuals have fewer chances of moving prices by liquidating their position.

The ongoing detailed analysis of LBP event log data aims to provide insights into the fair distribution of holdings. By examining each participant’s final holdings at the end of the auction, we will understand the token distribution among holders, highlighting whether the allocation is fair or highly concentrated.

Conclusion

The above Dutch auction vs. LBP comparative analysis revealed that the median Dutch auction performance is +97%, while the median LBP performance for both (1) Coingecko, DefiLlama, Dexscreener APIs and (2) Dune data is -89.49% and -87.63% respectively.

Namely, a buy-and-hold strategy of $100 invested into Dutch auctions would lead to a $197 valuation today, while an LBP position would be worth $12.37 today, a 15.92x difference. We verified that our results are consistent for various scales of fundraisers.

Overall, curated LBPs experienced losses exceeding half their value, with typical performances between -56.56% and -96.17%. Only the top 10% of LBPs managed to break even or achieve a positive return, with performances of +0.30% or higher.

The median number of participants in Dutch auctions is approximately 5.3x higher than in curated LBPs. Further, we found that Dutch auctions lead to on average 3.83x as many participants than Fjord LBPs for fundraisers of comparable sizes. With a broader distribution of token holders, we anticipate more stable and resilient price action over the long term, as the impact of any single individual’s actions on the market is diminished.

In a follow-up analysis, we will parse the LBP event logs to distil further the user distribution of holdings and refine our auction performance results, including using the average price weighted by tokens bought as a new benchmark.

From the above data analysis of Dutch auctions for token launches, we conclude that this auction format promotes better price discovery and long-term sustainable price action. The Dutch auction’s mechanisms have consistently demonstrated the ability to create a more equitable process for its participants, ensuring fairer outcomes and long-term stability following the Token Generation Event (TGE).

Our findings indicate that auctions remain robust independently of the scale of the token launch, offering advantages such as a wider distribution compared to other auction types like Fjord Liquidity Bootstrapping Pools (LBPs).

Enso is committed to adopting the most effective strategies for our token launch by meticulously analysing historical data and employing a data-driven approach. This ensures optimal outcomes for our community, fostering trust and supporting the long-term success and sustainability of our project.

Appendix

Google sheet data collection process

The raw data and aggregates are available on this public Google Sheet.

The curated Fjord LBPs, names, dates, funds raised, participant count, token contract address and auction ending price, namely the price corresponding to the last valid transaction during the LBP auction, were scrapped from Fjord https://app.fjordfoundry.com/pools using a Python script.

We then fetched the price as of today from Coingecko, DefiLlama, and Dexscreener APIs. If no price was matched in any of the three API calls, we manually matched it from the Coingecko website if the pair was found, else we left it blank.

Dune data process

We uploaded the raw data table containing each LBP pool auction data including the token contract address and price corresponding to the last transaction of each LBP. We uploaded the table of Start Date, Name, Symbol, Funds Raised ($), token Contract Address, and LBP auction-end Price ($) to Dune.

From there, we matched the most recent prices from these three tables prices.usd, dex.prices, and dune.dune.result_dex_prices_evm_daily.

Using the Dexscreener API, we fetched the pool, FDV, and liquidity available for the token contract address. We cleaned the data for some instances where the prices were abnormal (>=$1,000,000 per token unit).

We then compute metrics. We compute the Performance as Performance = Latest Price ($) / LBP auction-end Price ($) — 1. This leads to 150 curated LBPs, 25 of which have no Dune price data and no resolved pool by Dexscreener.

Fjord Foundry team as LBP Curator analysis

In the framework of our comparative analysis of Dutch auctions with Fjord LBPs, we also identified six LBPs that were curated by the Fjord Foundry team itself.

In this specific case, we manually filled the missing `LBP End Price ($)` for Tapioca $TAP and Weave Synthetic $WSN.

We fetched the Tapioca auction-end price by finding the $LTAP price of the last auction transaction in the pool, where $LTAP is redeemable as $TAP at TGE, we found a price of $0.47720 for $TAP on Coingecko and computed a Performance of +558.01% to date.

Weave Synthetic current price was not resolved despite three API calls to Coingecko, DeFiLlama, and Dexscreener, and a manual check on Coingecko, and the Twitter account provided on both the LBP pool and their website no longer exists (https://twitter.com/Weave_Financial).

We assign $WSN a Price Today of $0, yielding a performance of -100%.

Given the data in the table, the Fjord Foundry team-curated LBPs have a median performance of -76.72%.

Note on indexed pools on fjordfoundry.com

Please note that there has been at least one occurrence of a curated LBP being initially listed on the website, then being removed from both curated and non-curated pool pages, making it only available via direct link access.

For instance, STARSHIP and the $STAR token (blast:0x36d6b34D9B10b91b857920784eb62Cb95DbBBCde) that launched on June 27th 2024, raised $614.

Find the link to the pool here:

https://app.fjordfoundry.com/pools/0x2dc3f7dc25de79ad18fAE0118F5FF391502c797E.

Given the above STARSHIP occurrence, we think that the process of scrapping all curated LBPs on the landing page https://app.fjordfoundry.com/pools might not provide the full picture of curated LBPs that were launched on the platform.

A follow-up analysis, directly parsing the smart contracts’ event logs from on-chain data should yield all pool addresses, which we should be able to fetch on the website to determine if they were curated or not.

Sources

Links to the Google Sheet and Dune dashboard

Dune Dashboard: https://dune.com/enso_finance/lbps
Google Sheet: https://docs.google.com/spreadsheets/d/1HF0MNEH1sCRjyf-4Lg7YHJ7U2fsk1ft1cvPMSd5vpe0/edit?usp=sharing