Valuing a token economy — practical case

23 min readMay 11, 2018

Much ink has already been spilled on token valuation with the majority of the work going into developing (and criticizing) different token valuation models. This valuation exercise analyzes the basic dynamics of a token economy in order to explore its many different aspects. I believe one should use whatever is needed for the respective case based on its unique design. Trying to fit a token into an existing model seems limiting. Yes, bad news for the the lazy among us who want to plug in a few numbers and the model to spit out a token value. In an ideal world, ICO teams will have something like the below as part of their design and investor deck.


  • Summary how the project works
  • Stakeholders and incentives
  • Bottom-up valuation
  • Valuation recap — Sum Of Parts
  • Size of the opportunity
  • Trade-offs
  • Side effect

The example used is a recent ICO project called Ethearnal which uses both ETH and a native utility/work token called ERT.

UPDATE: The ICO managed to only raise 552 ETH of their ultimate goal of 30,000 ETH. Maybe the DAICO 2.0 model they chose for the ICO decided their fate, making it very hard for institutional crypto funds to participate — no discount, step-wise purchase which increases over time to allow more participants etc.

DISCLAIMER: I have neither participated in the ICO nor have any relation to the company or the team. This is not investment advice. Invest at your own risk and do your own research.


A freelancing marketplace token economy called Ethearnal, conceptually similar to Upwork or Fiverr.

What is Ethearnal? From their website:

Ethearnal is a peer-to-peer (P2P) freelance system, in which employers and freelancers meet, enter into trustless smart contracts with reputation and money in escrow, and take advantage of a decentralized system of moderators if needed. We collide reputation and economic initiatives into one by tokenizing reputation and giving it value. All parties, moderators included, have strong and aligned initiatives to act honestly, since everyone has something of value at stake and something to gain if the desired outcome is achieved.

In order to fully understand the system, please read the whitepaper.

Let’s start by identifying the stakeholders:

  • Freelancers: offer their skills for money
  • Employers: post jobs and look for talent
  • Moderators (workers): people who help settle disputes on the system and earn tokens for their work
  • Holders: hold (invest in) the token to receive fees and be able to vote during the deployment of the ICO funds

The system uses two tokens: ETH (for payments) and the native ERT (reputation token).

Why do you need ETH? When an employer posts a job for say $1000, they need to lock the ETH equivalent in escrow for the duration of the job. When the job is done, the freelancer receives the unlocked ETH minus a very small fee.

What is the ERT token? In their own words:

Short answer: Tokenized reputation

Longer one: It’s a token that has actual utility, outside of being used as payment method needlessly replacing ether. Actually, all payments in the network are done with ether. The token represents reputation, which has monetary value since the token is traded on the free market. It’s used for staking, aligning initiatives, moderating, taking job offers and just holding to get proportionally some of the fees the system collects.

First, we should address the two most pressing questions for every project:

  1. Do you need a blockchain solution for that? With the ultimate goal being to build an open, censorship resistant platform and eliminate rent-seeking, probably yes.
  2. Do you need your own token? The ERT token is used for governance, staking, moderating and fee collection so let’s assume yes.

Having ticked the two boxes, let’s move on.


One way to approach valuation is by comparing the incentives of the various stakeholders and what they optimize for. We can get an idea how their actions affect the supply/demand for tokens.

Freelancers and employers

In order to use the platform, an employer needs ETH and ERT and the freelancer needs ERT.

Step 1: posting a job

  1. Employer locks ETH equal to the job value in escrow.
  2. Both the employer and the freelancer need to stake ERT for the job duration (calculated as % of the job).

Step 2: job finished

  1. Freelancer receives 99% of the locked ETH (most likely) sells it on the market to cash out.
  2. The remaining 1% ETH is a fee, which gets split evenly between the employer and the freelancer in the form of ERT.
  3. The staked ERT are released to both.

Freelancers will naturally want to optimize for annual earnings based on variables like average hourly rate, job size and duration, # of jobs worked on, probability of success etc, but these variables do not concern ERT. What drives the ERT hold or sell decision will be an analysis around the optimal ERT holding to allow for working on multiple jobs and/or moderator working capital (if also acting as one). Ultimately, the decision also depends on the ERT price. There could be a mathematical model which optimizes for the above and spits out an optimal ERT holding. A similar exercise was done by Alex Evans in his VOLT model where he optimizes for token holdings based on transaction costs.

In short, the effects:

  1. ETH staking for the duration of the job
  2. Demand increase for ERT from both freelancers and employers
  3. ERT staking reduces its velocity (and potentially caps it)
  4. Further incremental demand for ERT with every successful job based on the 1% fee
  5. ERT kept by freelancers has its velocity reduced further until the optimum holding amount is reached at which point the excess is sold on the market

To sum up, ERT demand is positively correlated with the total number of jobs and the average holding period. The holding period drives velocity and is at least equal to the average job duration.


Here is where things get more interesting. As with any freelancing platform, there will be disputes. Since Ethearnal is decentralized, they propose to solve them via moderators. These are users with the necessary qualification to decide who is right and wrong in a dispute. There is no clear moderator attestation included in the system so anybody can be a moderator as long as they qualify. How? You stake ERT and if you are attributed a case, voila. From the whitepaper:

Mods can stake at minimum 5% reputation tokens of the predefined stake value in the job listing, and no more than 33.4% (implying a range between 3–20 mods per gig). So, effectively the employer, freelancer, and pool of moderators all have the same total reputation at stake. Once enough moderators have entered the pool, so that their collectively staked ERT tokens are equal to that of the other parties, the moderation process starts. Each moderator has a vote weight of
one, regardless of how many rep tokens he stakes. However, what they earn from solving the dispute is proportional to their rep at stake. When they decide the winning side in the dispute by simple majority (51%), the rep at stake of the losing side is distributed proportionally to the moderators based on THEIR rep at stake. ONLY THE MODERATORS WHO VOTED WITH THE MAJORITY (the winning decision) get rep tokens. The mods that voted with the minority lose their stake to the winning side of the dispute. The logic is that, since they tried to rule unfairly against him, but he turned out to be right, he deserves some rep. The mods who failed to vote within the determined timeframe lose their rep at stake. This will eliminate non-active mods automatically. In case the moderator votes are at odds, only then the system looks at the tokens behind the votes to decide the majority.

Anyone, who wants to work as a mod needs to stake ERT for the duration of the dispute, leading to incremental demand for ERT and reduction in velocity.

In short, the incentives when a dispute is settled:

The winning side (employer or freelancer) gets:

  1. The ETH from the escrow.
  2. Its staked ERT
  3. The ERT of the minority voting mods

The incentives and effects on price and velocity are similar to these with employers and freelancers.

The losing side loses only its staked ERT if the freelancer and the staked ERT AND the money in escrow if the employer. No effect on price or velocity.

Voting mods:

  1. The majority voting mods get the ERT of the losing side proportionally to their stakes. They have to optimize for available staking capital, i.e. keep the newly earned ERT or sell it. Mods who sell the earned ERT push the price down and velocity up. Mods who decide to accumulate ERT produce the opposite effect. Speculators or investors could participate in this part of the platform only and treat it as a pure financial trade.
  2. The minority vote mods lose their staked ERT to the winning side. No impact on the price, but heavy incentive not to be a losing mod.

Many questions arise including: How to manage the incentive for mods to cooperate by using outside channels to find out who the other mods are on a given case and increase their chances of winning (even if the system is designed to award mod jobs randomly and anonymously)? What is the optimal ERT holding for a mod? Can you come up with an optimal ERT holding as a participant? What is the optimal behavior when voting given the effect that the own vote and size of the stake have on the outcome? All of these would require another post. Below is a short visual example using the data from the whitepaper.

Here is a short summary of the incentives and actors.

Platform rent-seeking

The system has the option to charge the 1% fee of the job amount also in the case of a dispute. Instead of being split between the employer and the freelancer, this dispute resolution fee is used to buy ERT on the market and distribute it to all ERT holders (the reasoning is on pg. 12 of the whitepaper). It is an interesting idea as a price support vector for ERT and equalize the 1% fee among the potential outcomes (job completed with or without a dispute). The flip side being that it is a rent seeking burden for one stakeholder (the employer) and a benefit for another (the token holders). In my opinion, this will not be enough of a reason for a fork if the other aspects of the platform all work well. Furthermore, if this is put up for a vote by token holders, everyone will be incentivized to accept as it is effectively a passive “dividend”.


Now that we’ve established what and who drives the token price, let’s put a value on it.

Ethearnal planned to issue maximum 40,000,000 ERT tokens, based on a hard cap of 30,000 ETH. 30m ERT in the public ICO, 8m for the founders and 2m allocated for bounties and business development (held by the founders). The ICO price was to be determined based on the ETH/USD price after the ICO period ended (31 March 2018). For each ETH, the ICO would generate 1,000 ERT.

The plan was to use the ICO money for development of the platform and to market to employers and freelancers.

We need a few assumptions to be able to run the numbers:

  • Customer adoption according to the S-curve model (more here) to get to 770k freelancers (1% global market share)
  • Average annual earnings per freelancer of $19,480 based on Freelance Union estimate of the global market size of $1.5trn and 77m
  • 1 job done per freelancer with an average job duration of 2 weeks leading to $750 per job
  • Employers to freelancers ratio of 0.5x
  • 20% staking by each party (freelancers, employers and moderators in dispute cases). For simplicity, I have assumed all mods are external, hence the full incremental token demand is added. In reality, this number will be slightly lower
  • 10% disputes of all jobs done
  • Founder tokens are released over 3 years, rest is available for trading from day 1 of the network launch
  • ICO price of $0.50 — the ICO did not happen so this is an arbitrarily chosen price corresponding to an ETH price of $500

Valuation of ERT by the team

The way Ethearnal approaches the valuation of the ERT token is by doing simplified calculations about a) 1% fee on every completed project driving marginal demand and hence ERT price appreciation and b) 1% fee from disputes being distributed among all ERT holders. These are described in a document on the website called “Token profitability”.

The logic behind a) is simply incorrect. First, the team uses numbers for the mature state of the network when it has successfully captured 1% global market share. If you only applied a massive probability and time discount to this number, it would reduce their estimate dramatically. EVEN if we assume that they succeed, the team claims that the 1% fee will drive up the demand and increase the token price. However, they miss the point that at least 0.5% of this (the employer’s share) will be immediately sold back on the market (we discussed the incentives and price action earlier). For the other 0.5%, only part of it will be held by freelancers. So these numbers will have a very small effect.

On b) they again make the mistake of assuming a fully mature network numbers from day one. The logic of the “dividend” is that if the token holders decide to enact the 1% fee on disputes, then the money is used by the platform to buy ERT and distribute proportionally to all token holders. For the sake of the analysis, I have assumed this option is ON. Here is the summary from their document — looks quite promising :)

Funnily, the team seem to have missed the main driver for token price appreciation — the demand increase with the growth of the network in a fixed token supply world. Economics 101.

What they should have done instead: bottom-up approach to establish the token’s value and comapre it to the ICO price

  1. Use the assumptions for # of freelancers and employers and average job value to estimate the total value of jobs per year.
  2. Calculate the staking demand in USD using the staking % assumption.
  3. Using the ERT release schedule assumptions, translate the amount of staked USD chasing ERT to arrive at an implied ERT/USD price. For the avoidance of doubt, this would be the equilibrium price where demand and supply meet. The market price will (almost) always be different and impossible to predict. The fundamental equilibrium price is something we can at least use as a reference for the value.
  4. Discount the equilibrium values to the present day
  5. Compare to the ICO price

If we forecast 10 years out (network maturity), we can see the network growth effect on the price.

Here we hit an obstacle! In the first 3 years, the ERT/USD price is “fundamentally” below the ICO price, given the assumptions for token supply and demand (the blue cells). The initial demand will naturally be low as the network gains traction, which leads to the implied ERT/USD price being much lower than the ICO price. No rational market participant would sell at the equilibrium price early on. It will need to be at least equal to the ICO price, most likely a premium for the assumed early risk.

How can this be solved? I see two possibilities:

  1. Liquidity by the founders: another way would be for the founders to have a token working capital pool which is used to provide supply when needed at “some price”. This is tricky — at what price? Many challenges arise. Maybe the ICO price should be the floor? That would be fair to ICO investors. A lower price could reflect the state of the network but could be problematic for marking down of investor books. Such a liquidity provider of last resort would be akin to a state intervention in a free market and could trigger a sell-off spiral. Founders need to make sure that they provide liquidity to the network users and not to speculators who don’t contribute to the network growth and know that there is a backstop if the network is not developing as planned (for which there is a non trivial probability). The usage and growth of the platform should be leading here because this is the only factor leading to higher prices in future. Solving this will be key. ICO investors simply need to wait for the natural token demand to build up. It becomes a matter of timing and cost of capital.
  2. Via the token price: users of the platform price the service in USD so they don’t care so much about the token price itself. In other words, if 1 job is $750 and both sides need to stake 20%, they need to buy $144 in ERT tokens each. Whether that is 200 ERT or 100 ERT, they would not care. Those who care are the ERT holders who decide at what price to part with their ERT. Obviously, people can’t mark up the price indefinitely, as, in a free market, available supply will emerge at different prices for different reasons. For example, the ICO investors paid X ETH for the ERTs and hence and bear the opportunity cost of Ether. If the platform is not developing as quickly as desired, some people may decide to sell their ERT at a lower than the ICO price. Investors may have a required rate of return, which, once reached, will trigger a partial or full sale of their position. Fundamentally, if the platform is developing well, the token mechanics should allow an increase in price due to the supply cap and demand growth. Herein lies the challenge with this driver: if people just hold the tokens and wait for ever higher prices, the network will die. It absolutely needs the early liquidity.

In the case of Ethearnal, if 33% of founder tokens are attributed to a liquidity pool (2.67m tokens — shown as the only available supply in year 1 in the table above), the platform will need to be able to process ca. $7mn in total value of jobs before the equilibrium ERT/USD price is higher than the ICO price. Then, market forces come into play and sellers should emerge at the higher prices, increasing supply to match the coming demand.

One tricky point is how an internal working capital pool would interplay with exchange volumes in order to avoid arbitrage. A way to explore this is to link the internal pool price with exchange prices and take into consideration volumes, floor price, etc. Surely, much easier said than done.

Discount rates

The last column shows the discounted token price at an arbitrary discount rate of 25%. This is where investors will focus. As expected (and shown by other analysts before), the discounted prices don’t directionally follow the fundamental price. Time takes its tall, meaning there is an optimal holding period, all else being equal. These are mechanics which traditional investors are quite familiar with. With this, we have one piece of the value puzzle.

By the way, cost of capital is a big deal. To fully appreciate the effect of the discount rate, the graph below shows the undiscounted utility value and two options: 10% and 25% cost of capital.

How likely is this development?

Comparing the planned ICO size to the team’s ambitions and business plan should give us an insight into the potential price action. Ethearnal’s team aims to attract ca. 800,000 freelancers on its platform (1% global market share) with the funds raised plus the corresponding number of employers. For comparison, the global leader — has a total of 27m registered users (freelancers + employers). Ethearnal’s plan is basically to spend $23m on online ads to get everyone on the platform and the party starts. The way they get to $23m is by using cost per click statistics and an increasing conversion rate. IF such an approach proves successful, they should have enough funds to grow the network to a stage where the ERT equilibrium price is 3.02 USD (in 7 years) or $0.63 discounted. Assuming an ICO price of $0.50, this would imply an annual return of 29%. Whether this is enough or not would be different for everyone. Some investors use discount rates as high as 40%. To them, this will not be that attractive. However, the equilibrium price is an indication at best and the actual market price should give plenty of opportunity for investors to realize their projected return expectations at various levels.


As a crosscheck, we can also calculate the same with the MV=PQ framework in which we would only use an ERT based economy and then come up with a Purchasing Power Parity price of the token in USD (as shown below). The P here is the ERT price of a job. It decreases substantially over time with the growth of the network because the average price of a job in USD stays stable and network participants need to stake less tokens per job. What enables us to use this framework is the fact that a freelancer job is a service which can be globally priced in USD. Hence, in our mini ERT economy, we could set the initial price of the same service in ERT tokens to equal the global price in USD by using the ICO price. From there, supply and demand take over. The PPP framework to value currencies comes from economics and is best known for the Big Mac Index. For more on PPP and token pricing, read here. Please note that V is stable at 25x from year 4. This is an output, resulting from the token economy mechanics and the fact that we assumed an average job duration to be 2 weeks, which leads to a turnover of all tokens 25 times per year.


There is a thesis out there in crypto space that even if you stake tokens which limits their velocity, the rest that are not staked (and are only held for a very short time) may have such high velocity that the overall velocity will still be so high as to push the value of the token close to zero. Sure, there may and very likely will be plenty of cases where this holds true and the market cap needed to support even very large networks will be relatively small.

Let’s take Ethearnal: assume that there is such high demand for ERT that all 40,000,000 tokens in existence are staked by network participants all the time. How? Simply, the network grows so that enough USD is chasing ERT and every time a job ends and a participant sells ERT, there is a new buyer. Being an ERC-20 token, ERT is divisible up to 18 decimals (if the default option is kept). These 40m tokens are in fact more than enough to sustain the growth. When the demand starts to push the supply limit, the price will reflect this and you will need less and less ERT for staking. All this means is that the token holders can sell parts their holdings as/if the network grows and still keep enough for staking, moderation etc. There will always be sellers at various price levels.

An interesting alternative would also be a market for token lending so that holders lend out their tokens to employers or freelancers for staking (work) purposes, thus keeping the upside. In this case, a return is earned as opposed to pushing the price up. This could be a better solution when the buyer does not want the token, but rather needs it temporarily. Depending on the mechanics and the implied rates, this could be another reason to hold.

Cash flows

A common issue that traditional investors have with tokens is the lack of cashflows. ERT is a work token and there are two ways to earn it:

  1. Participate in dispute resolution — I touched upon this in the Moderator section using an example from the whitepaper. One could do very simple probability weighted calculations to arrive at an expected return. Using the example in the moderator section and a 50/50 chance of either losing everything or winning 143% return, the expected return is 21% for the duration of the dispute. Things are, of course, more complex than that, but if someone approaches it with multiple bets over the year, the annual return when the network is mature could be in this range. One way to mentally crosscheck whether people would participate in the so designed moderation system would be to use Prospect Theory.
  2. Hold ERT if the option to distribute the dispute fee of 1% as a reward among all ERT holders is implemented (like a dividend). Having cashflows means we can use traditional NPV valuation models from corporate finance to value this part of the network. For example, buying ERT at the start of each year at the respective equilibrium price, you earn 6.5% dividend yield for a year. However, what investors will do is monitor the implied dividend yield based on the current market price and decide when to enter. A similar idea is the yield on running a masternode and receiving crypto (for example, Neptune Dash does this with Dash). As an ICO investor, if you invest at $0.50 per token, hold for 10 years, receive dividends and sell in year 10 at the equilibrium price of $4.20, the IRR would be 29%. Using 25% discount rate, the NPV comes to $0.67 per ERT (above the ICO price of $0.50). Needless to say, the discount rate is a huge driver. I have written more about its effect on valuations here.


Overall, every project should be evaluated from the ground up in terms of its design and stakeholder incentives. Then it can be valued as a sum of the different parts. The issue is that there are so many variables to consider that, honestly, these should only be used as reference for sensitivity analysis.

Nevertheless, in my opinion, every project aiming to do an ICO should run a similar model in house in order to understand the drivers. It will allow them to better structure the mechanism design and be much better prepared when meeting investors. I have only used well known and understood frameworks for this analysis — supply/demand equilibrium and PPP prices, net present value and S-curve. Others are being developed and will surely emerge as being useful in particular situations.

Table 5


Having a value range for the token, the next challenge is the size of the investment opportunity, i.e. will this be attractive for larger players. By year 7 (if all goes according to plan!), there should be an annual total job transaction volume of ca. $7.5bn. This translates to an implied ERT demand for ca. $115m (job staking) every two weeks. This translates to $3bn in annual USD volume. If we use a trading multiplier, then this number could be much bigger. Such numbers for liquidity start to look very interesting for a variety of investors, from individuals to hedge funds, to family offices and even private equity (large and small).


Whether a decentralized new entrant can successfully challenge a dominating centralized entity is riddled with various trade-offs:

Capital upfront vs. step-wise raise based on traction

By far, my biggest problem with this project is that they literally rely on dumping most ICO funds on Google and Facebook ads and hope the users will flow in. The best and largest online marketplaces have gotten to where they are with a lot more than spending on ads. This is not to say that they don’t spend millions online, but I view this as the gasoline that you put on a burning fire and not the dry wood to start it. I fear the spending may be grossly underestimated both in terms of timing and volume. To get things right, you need to iterate the product and lose many users along the way which increases the cost 2–3–4x. In their budget, they don’t have much room to maneuver.

Focus on decentralization (censorship resistance) and elimination of rent-seeking vs. focus on UX, product and business development

One of the major difficulties that decentralized consumer marketplaces face is that they optimize for decentralization and removal of rent seeking, and not necessarily for user experience. AirBnB, Uber and other highly successful companies have spent immense resources on developing and testing the best product for the customer. A crypto alternative faces many UX problems and trade offs on its otherwise noble path to remove fees and to give more power in the hand of the customer. Specifically in the case of Ethearnal, many questions/issues need to be answered/addressed:

  1. Onboarding: how do you onboard employers and freelancers? How do you explain the concept of tokens and the need to stake which increases the effective cost and hampers UX? How do you enable seamless conversion from fiat to Ether and back. Where do you buy ERT and how is it stored securely?
  2. Staking: when staking ERT, how do you ensure price stability during the job — asan employer who wants to sell the ERT released back to them once the job is done, how do you explain or prevent price volatility?
  3. Fees: is the eliminated commission of 10–20% plus payment fees high enough for people to switch to a 0.5%-1% fee alternative while sacrificing UX and convenience?
  4. Network growth: centralized companies have unified product management, quality testing, growth hacking and scaling approach. They don’t have to optimize for trust, verifiable assets and price, rather for customer experience and adoption. Simply spending a ton of money on AdWords and similar online channels to get users on the platform will surely be insufficient in order to replace a well functioning centralized organization with energetic and capable leadership, product vision and dedicated team.
  5. Business development: centralized marketplaces are obsessed with product-market-fit and spend a ton of time on business development and product testing — they optimize for that. A crypto team needs to first develop the network and then start massive business development to attract partners, customers, suppliers etc. It takes a lot more money, time, skill and effort and it is not clear if an amount raised upfront (unless it is insanely high), will be enough to sustain a team and new hires for a long time. Centralized marketplaces have cashflows to aid them along with regular capital raising into the billions of dollars.

Fungibility and verifiability

A general observation (courtesy of Naval Ravikant) is that crypto networks where the suppliers do not offer a fungible product/service don’t really have a robust design. In a crypto AirBnB, who is verifying that a house is worth what is says and that it is really available? Not that people will always lie, but you then need to trust someone else to verify that (miners can’t do that). In a file storage network, this is objectively verifiable, but for availability and quality of freelancers — maybe not.

Open vs. closed

On the other hand, there is a counterargument to the above. Open networks always end up outgrowing closed ones. Currently, all centralized marketplaces have their databases behind closed doors. These databases grow and shrink with the development of their user base. Decentralized networks start to emerge and get populated with data supplied by the users who can participate in the value co-creation! This is the first major difference — the incentive structure. The larger the datasets grow, the more valuable they become to everyone, especially with the current development of data exchange protocols like Ocean, AI protocols like Fetch etc. These set the stage for dApps to address some of the shortcomings like lack of verification, business development, insurance, UX etc. All of these can be solved with the proper incentive structure of value co-creation.


By doing the above exercise, I noticed another interesting effect. Even though I only discuss and analize ERT, Ethearnal has a direct impact on ETH. An amount needs to be locked for the duration of the job. This leads to ETH’s velocity being inversely correlated with the average job duration and the size of Ethearnal’s network.

Using the assumptions above, if Ethearnal ever got to a mature network, a number of ETH will always be locked for 2 weeks. Obviously, it all depends on the USD price of ETH. For the sake of the example, if ETH hovers around $1,500 over the next 10 years, then by the 10th year 266,000 ETH will be constantly locked. If we have 150m ETH in circulation by then, Ethearnal would represent 0.2% of all outstanding ETH. All because they decided to design the platform so that ETH is locked instead of fiat. This shows that only one project on the Ethereum blockchain could have a non-trivial effect on the velocity of ETH (and hence, its value). This opens up cool possibilities for programmatic sum-of-parts analysis and valuation of ETH.