Tokenomics: How to Design a Token Economy That Holds Up

Tokenomics is the economic design of a crypto token: how many exist, who holds them, when those holders can sell, what the token does, and how value gets back to it. That design is fixed in contracts and schedules long before most users arrive, and it decides more about whether a project survives than the code does. A protocol with mediocre engineering and honest tokenomics outlives a beautifully engineered one whose insiders hold 60 percent of supply on a six month cliff.

One disambiguation first, because search results mix two subjects. In machine learning a token is a fragment of text, and model providers bill per token: OpenAI's pricing page quotes rates "per 1M tokens." Managing that spend is sometimes called AI tokenomics. It's a real discipline that shares nothing with blockchain token design except the noun. Everything below is about crypto tokens.

This is a tokenomics design guide, not a glossary: the parts of a token economy, the allocation numbers real launches used, dated examples from Ethereum through Bitcoin, two failures with the arithmetic that caused them, and a method you can run against your own project.

The short version

• Tokenomics has five moving parts: supply, distribution, vesting, utility and value accrual. The first four get written down. The fifth usually doesn't, and that's where most designs fail.
• Across 150+ launches, the 2023 average allocation was 28 percent community treasury, 24 percent core team, 20 percent private investors, 20 percent airdrop and 8 percent ecosystem incentives, per Lauren Stephanian and Cooper Turley.
• Keyrock analysed more than 16,000 unlock events. Ninety percent create negative price pressure, and the effect starts around 30 days before the unlock date.
• Ethereum's burn only offsets issuance when blockspace is contested. On 10 September 2026 the network issued about 2,940 ETH a day and burned about 38.
• Emissions without sinks is the most common design error there is. Read Terra and Axie Infinity before you write an emission schedule.
• Your tokenomics document is a legal document. Regulators read the allocation table and the marketing copy together.

What tokenomics means

Tokenomics means the rules governing a token's supply and demand, written into smart contract development and vesting agreements. Four questions: how many tokens will ever exist, who receives them and on what schedule, why anyone needs to hold one, and where the money comes from.

Why does this decide survival? Because a token is a claim on nothing unless you make it a claim on something. Equity has a legal claim on residual cash flow, a bond has a contractual one, and a token has whatever the design gives it. If that's only the hope of resale, the price becomes a pure function of net inflows. Fine during a bull market. Ugly when inflows stop.

There's a second reason, less discussed. Tokenomics is the only part of a crypto product that's genuinely irreversible. Rewrite the front end, migrate the database, upgrade the contracts, fine. You can't un-give 20 percent of supply to a fund that already sold it.

Every other part of the product can be shipped, measured and changed. The token only ships once.

Token supply, distribution and vesting schedules

Supply, distribution and vesting are the three tokenomics decisions founders keep collapsing into one slide. They answer different questions: how much, to whom, and when.

Token supply: fixed, inflationary and deflationary

Token supply is the total units the protocol will ever create, plus the rule governing new creation. Three models cover almost everything in production.

Fixed supply caps issuance at a number written into the protocol. Bitcoin is the canonical case: the block subsidy halves every 210,000 blocks. The fourth halving landed at block 840,000, mined on 20 April 2024, taking the subsidy to 3.125 BTC. Block 966,350 on 10 September 2026 paid a total reward of 3.1296 BTC, subsidy plus fees, so the schedule does exactly what it says. Next halving is block 1,050,000, roughly 83,600 blocks out, around 580 days at ten minutes a block. Anyone can check that against mempool.space, which is the whole point of a fixed supply.

Inflationary supply mints new tokens on a schedule, usually to pay for security. Solana's documented schedule starts at 8 percent, disinflates 15 percent a year, and settles at a long term 1.5 percent. Query the chain and you can watch it happen: at epoch 1032 on 10 September 2026 the mainnet RPC method getInflationRate returned 3.654 percent, against a total supply of 633.8 million SOL and 586.3 million circulating. In November 2025 SIMD-0411 proposed doubling the disinflation rate to 30 percent, reaching the floor in about 3.1 years instead of 6.2. That a published schedule can be renegotiated by governance is itself a property worth knowing.

Deflationary supply burns faster than it mints, or at least tries to. Ethereum has the best documented burn in the industry and the best cautionary tale attached to it, which we get to below.

Most real designs are hybrids. A capped supply with a burn on top is still capped, and an inflationary chain whose fee burn exceeds issuance is deflationary while that condition holds, not a minute longer.

Token distribution and allocation benchmarks

Distribution is who receives the supply. Sophisticated buyers check it first, because a token where insiders hold most of the float has its price set by insiders' patience.

Lauren Stephanian and Cooper Turley aggregated distributions across more than 150 projects spanning 2014 to 2023 and published the updated analysis in March 2024. Their 2023 averages are the closest thing tokenomics has to a published benchmark.

Two caveats before anyone pastes that into a deck. Variance by project type is enormous, since an L1 paying validators needs a different shape from a DAO. And the authors say it themselves: average isn't optimal.

Named launches show how wide the range gets. Uniswap minted 1 billion UNI in September 2020 and split it 60.00 percent to community, 21.266 percent to team and future employees, 18.044 percent to investors and 0.69 percent to advisors, the last three on four year vesting. Arbitrum issued 10 billion ARB: 35.28 percent to the DAO treasury, 26.94 percent to team, contributors and advisors, 17.53 percent to investors, 11.62 percent airdropped to users, 7.5 percent to the Arbitrum Foundation and 1.13 percent airdropped to DAOs building on the chain. Optimism created 4,294,967,296 OP and reserved 20 percent of it for retroactive public goods funding, ahead of its 19 percent to core contributors and 17 percent to investors.

Uniswap's community share reads generously until you notice Arbitrum's insider block, team plus investors, is 44.47 percent. Neither is wrong. Different bets about who needs paying to make the network work.

Token vesting, cliffs and unlock schedules

Vesting is when allocations become sellable, and it's the part of tokenomics with the biggest measurable price impact.

Keyrock analysed more than 16,000 unlock events for From Locked to Liquidity, and the findings are blunt. Over 600 million dollars of tokens unlock every week. Ninety percent of unlocks create negative price pressure regardless of size or type, and the effect typically begins about 30 days before the unlock date, not on it. Larger unlocks produce roughly 2.4 times sharper drops. Team unlocks are the worst category, with drawdowns reaching 25 percent, while investor unlocks land softer because funds use OTC desks and TWAP execution instead of hitting the book. Ecosystem development unlocks are the one category averaging positive, at plus 1.18 percent.

Their worked example is ApeCoin. From 1 March 2023 a linear team unlock released 0.7 percent of total supply per month, about 11 million dollars against a 1.6 billion dollar market cap. Over the next seven months APE fell 77 percent while ETH fell 9 percent, and on-chain data showed the team depositing into market maker OTC accounts throughout.

Three structural choices follow from that data:

• Cliff length. Stephanian and Turley's vesting analysis, built from 150+ data points contributed by Arca, Delphi Digital and Pantera Capital, found a six month cliff beats a one year cliff or no cliff for specific date vests, and that lockups have lengthened over time.
• Linear over batch. The same research found linear vesting produces lower volatility than specific date vesting. Keyrock agrees, though large cliffs often recover better after 30 days.
• Trigger design. Around 70 percent of schedules are trigger based, 65 percent of those starting at the token generation event. Tying vesting to a launch date you control is a decision, not a default.

Practical rule: assume every cliff in your calendar is a day the market priced in and started front running a month earlier. Design so no single day matters much.

Token utility, demand sinks and value accrual

Utility is what the token is for. A demand sink forces tokens out of circulation to get it. The second holds a price. The first, on its own, doesn't, and confusing the two is the most common tokenomics mistake in a pitch deck.

That loop has to close. Utility creates demand, demand produces fee revenue, fee revenue feeds a sink that removes supply or pays holders, and that value funds the incentives bringing in the next cohort. Break any arrow and you've got an emissions programme, not an economy.

Utility that creates a demand sink

Four utility patterns actually produce buying pressure, and they differ in how easy they are to fake.

• Fees paid in the token. The strongest sink, because usage forces purchase. XRP takes the mechanic to its limit: XRP Ledger documentation states each transaction destroys a small amount of XRP, minimum 10 drops, and that "the XRP is irrevocably destroyed" rather than paid to any party.
• Collateral. Posting the token to borrow, mint or underwrite locks it while the position stays open. Useful, but reversible in one transaction.
• Staking with real unbonding. Locks supply and pays for security. Watch where the yield comes from, because issuance-funded yield is a transfer from non stakers, not income.
• Access. Gated features, allocations, tiers. Cheap to build, easy to route around, and the pattern most often used as a fig leaf.

Governance rights are a weak sink on their own. Voting doesn't remove a token from the market, and most participation concentrates in a handful of delegates anyway.

Incentives and emissions

Emissions are tokens paid out to attract behaviour: liquidity mining, points programmes, quest rewards. They work, in the narrow sense that they buy the metric they target. The tokenomics question is what happens the day they stop.

Model emissions as a customer acquisition cost with a payback period, not as a growth strategy. Emit 8 percent of supply over two years to attract liquidity and you've spent 8 percent of the company renting a balance sheet. Sometimes that's the right trade, but only if users stay after the subsidy ends. Model the post subsidy cohort before you commit.

Value accrual: fee switches and buybacks

Value accrual is how protocol revenue reaches token holders. Most tokens launch without one and bolt it on later under pressure, which is why the Uniswap sequence repays a close read.

UNI existed for five years with governance rights and no claim on protocol fees. In November 2025 Uniswap Labs and the Uniswap Foundation proposed UNIfication: cut v2 LP fees from 0.30 to 0.25 percent and route the difference to the protocol, take one quarter of LP fees on v3's 0.01 and 0.05 percent tiers and one sixth on the 0.30 and 1 percent tiers, burn 100 million UNI from the treasury, and fund a 20 million UNI annual growth budget from 1 January 2026. The on-chain vote executed on 28 December 2025 with 125,342,017 votes in favour and 742 against.

It happened. Querying Ethereum mainnet on 10 September 2026, the UNI contract reports a total supply of exactly 1,000,000,000, while the burn address at 0x...dEaD holds 110,817,581 UNI, about 11.08 percent of supply. Note the mechanic: tokens go to an address nobody controls rather than being destroyed, so total supply is unchanged and circulating supply is what moved. Worth remembering next time you read a burn announcement.

Tokenomics examples worth studying

Useful tokenomics examples are the ones you can verify yourself. Everything here was measured on 10 September 2026 from public APIs and on-chain calls, and every query repeats.

Ethereum has the best documented supply mechanism in crypto, and the best warning about reading a mechanism without its conditions. The Merge cut issuance by about 88 percent: ethereum.org puts pre-Merge annual issuance at roughly 4,930,000 ETH from mining plus 620,500 ETH from staking, a combined 4.61 percent, dropping to about 0.52 percent after. EIP-1559 added a fee burn on 5 August 2021, and ultrasound.money records 4,638,309 ETH burned since, 2,015,000 of it post-Merge.

Now look at today. Ethereum's 30 day average base fee on 10 September 2026 was 0.177 gwei against an all time average of 19.35 gwei, because most activity settles on L2s. At that level it burned 1,149 ETH over 30 days, roughly 38 a day, while issuing about 2,940 a day.

Net supply growth works out to about 0.87 percent a year. The deflation is real but conditional, appearing only when L1 blockspace is contested. A burn is a mechanism, not a promise.

Solana is a disinflation curve in flight, publicly queryable and subject to governance. Uniswap proves value accrual can be retrofitted, at the cost of five years of holders wondering whether it ever would be.

Arbitrum runs a DAO treasury at scale: 3.528 billion ARB under governance control, an airdrop claim window running 23 March to 24 September 2023, and inflation capped at 2 percent a year.

Optimism built its allocation around a funding thesis rather than a cap table. Its community documentation records the early retroactive funding rounds: 1 million dollars across 58 projects at the end of 2021, 10 million OP across 195 projects in Q1 2023, and 30 million OP across 501 projects in Q4 2023. Keyrock picks Optimism out as its unlock success story: the June 2022 governance fund unlock of roughly 3 percent of market cap caused an initial selloff, then drove growth, because the tokens went into incentives rather than to sellers.

Bitcoin demonstrates what seventeen years of an unchanged schedule buys: a supply curve nobody argues about. It's also the design least available to you, since it works only with no team allocation and no investors.

When tokenomics fails: Terra, Axie and airdrop farming

Two failures explain most of what goes wrong, and both were visible in the arithmetic before the price.

Terra and LUNA: tokenomics as a reflexive loop

Terra's UST was an algorithmic stablecoin backed by its own governance token. Burn one dollar of LUNA, mint one UST, and back again. The peg held as long as LUNA's market capitalisation could absorb redemptions, so the stablecoin was collateralised by the price of the thing whose value depended on the stablecoin working.

The Federal Reserve Bank of Richmond's post mortem lays out the numbers. At the peak, UST circulating supply was about 18 billion dollars, with 16 billion sitting in Anchor, a protocol paying a 19.5 percent deposit rate. UST fell as low as 0.985 dollars on 8 May 2022, 0.60 on 9 May and 0.22 on Binance. Luna Foundation Guard deployed 1.5 billion dollars of reserves against an 18 billion dollar liability. LUNA supply hyperinflated roughly 80 times, from 0.4 billion on 10 May to 32.0 billion on 12 May, and the price went from 31 dollars to one cent.

Collateral and liability were the same bet. Not an exotic error: a stress test shocking LUNA's price by 60 percent would have shown the loop failing. Nobody was required to run one.

Axie Infinity: emissions without a drain

Axie Infinity paid players in Smooth Love Potion for winning battles, and SLP was spent on breeding new Axies. Faucet and sink, in principle. In practice the faucet ran faster.

Sky Mavis said so themselves. Their Dev Journal on economic balancing, published 4 January 2022, reported SLP burn growing over 500 times during 2021 while creation grew over 160 times, and stated plainly that the data showed "a growing chasm between SLP minted vs. SLP burned" and that "this inflation in the token is not sustainable." Average SLP per PvP battle had risen about 40 percent across a single season, which is what happens when a reward curve gets tuned for engagement instead of supply.

General form: a play to earn economy where earning beats spending converts new player deposits into old player withdrawals. Designing a game economy? Our breakdown of how blockchain games like Pixels structure their loops goes deeper.

The airdrop farming problem

Airdrops are supposed to buy distribution and community, which is why so much tokenomics attention goes into them. Increasingly they buy neither, because the recipients are professional farmers.

Keyrock's study of 2024 airdrops covered 62 airdrops across six chains. Only 8 produced a positive return after 90 days, an 11 percent success rate, and 88 percent of tokens declined within months despite initial price spikes. Solana managed a 25 percent hit rate and Ethereum 14.8 percent, while BNB, Starknet, Arbitrum, Merlin, Blast, Mode and zkSync had no winners in that sample.

Patterns separating the survivors are usable, though. Airdrops distributing more than 10 percent of total supply saw stronger retention, while those under 5 percent tended to get sold quickly. Inflated fully diluted valuations did the most damage, particularly where liquidity was too thin to absorb the sell pressure a high FDV invites. Size the airdrop to matter, price the launch so there's room above it, and fund the order book before the tokens land.

How to design tokenomics step by step

Here's the tokenomics design sequence we run. It's deliberately front loaded, because the expensive mistakes happen in steps one and two, and no amount of simulation rescues a token with no job.

1. Define the job of the token. Write one sentence: this token exists so that X. If the sentence is "so we can raise money" or "so users feel ownership," stop. Plenty of good blockchain products don't need a token, and shipping one anyway adds a securities question to a product problem.
2. Model the demand sinks before the supply. List every mechanism that forces someone to acquire and hold the token, then size each at realistic usage. If total demand at your year two forecast is smaller than your year two emissions, the design is already broken and no schedule fixes it.
3. Set supply and emission. Fixed, inflationary or hybrid. Write the rule so it can be verified on-chain rather than trusted, and define the terminal state as well as the opening rate.
4. Allocate and vest. Split supply between team, investors, treasury, ecosystem and community, then attach a schedule to each. Benchmark against the 2023 averages above and defend every deviation. Keep the combined insider share defensible, stagger the cliffs.
5. Simulate it. cadCAD, the open source Python framework maintained by BlockScience, supports agent based modelling, system dynamics and Monte Carlo runs, which is what you need to test behavioural assumptions rather than one spreadsheet path. Machinations covers similar ground visually and lists tokenomics among its use cases. Run distributions, not point estimates. Common mistake here: simulating the happy path and calling it validation.
6. Stress test the unlock cliffs. For each unlock, compute released value against 30 day average traded volume. Anything above a few percent is a day the market trades against you a month in advance. Restructure until no single date dominates.
7. Plan liquidity. Initial float, venues, market making, funded depth. Keyrock's airdrop data points at thin liquidity paired with a high FDV as the common failure mode, and that's a treasury decision made weeks before launch.
8. Write it down. A token paper stating supply rules, the full allocation table, every vesting schedule with dates, the sinks, the value accrual mechanism and the governance process. Fair warning: if you can't write it clearly, you haven't finished designing it.

That chart models one schedule so you can see what step six looks for. Assumptions: 1 billion tokens, no inflation, split on the 2023 benchmark of 20 percent airdrop, 8 percent ecosystem incentives, 28 percent treasury, 24 percent core team and 20 percent private investors. Airdrop unlocks fully at launch. Ecosystem incentives release linearly over 24 months, the treasury linearly from month 6 to month 48. Team tokens run a 48 month schedule with a 12 month cliff, investors a 36 month schedule with a 12 month cliff, and both cliffs release the accrued quarter and third respectively in one transaction.

Read month 12. Circulating supply jumps from about 27 percent to about 41 percent, so 13 percentage points of supply becomes sellable in a single day. That's the number to design away, and you only see it by drawing the curve.

Tokenomics and the law: utility versus security

Regulators read tokenomics documents, and the allocation table is evidence.

In the United States, whether a token sale is an investment contract runs through the Howey test: an investment of money in a common enterprise with a reasonable expectation of profits from the efforts of others. The SEC published a staff framework applying that test to digital assets in 2019, then withdrew and superseded it with a new interpretive release dated 17 March 2026. The withdrawal is itself the lesson: US crypto guidance has moved materially twice in seven years, so treat any published framework as a snapshot and get current counsel.

What hasn't changed is which facts the analysis turns on, and they're tokenomics facts. Who does the essential managerial work. Whether the token can be used today or only after a roadmap ships. Whether purchase quantities look like consumption or investment. Whether marketing sells the team's ability to increase the price. A design where the team holds most of supply, the utility arrives next year and the deck shows a price chart answers all four the wrong way.

Europe took a categorical approach instead. MiCA, as ESMA describes it, sorts crypto assets into three buckets: asset referenced tokens under Title III, e money tokens under Title IV, and other crypto assets including utility tokens under Title II, each carrying different authorisation, white paper and disclosure obligations. The regulation entered into force in June 2023 and applied fully from December 2024, with white papers required in iXBRL format from 23 December 2025 and a grandfathering transitional period running to 1 July 2026.

Two consequences you'll feel. Your white paper is a regulated disclosure document in the EU, not a marketing artefact, and its numbers have to match the contracts. And categorisation follows what your token references and does, so pegging to a basket can drop you into a heavier regime. Choose deliberately.

Common tokenomics mistakes

Four errors show up in most reviews.

A pre-mine that's too large. When team plus investors beats the community allocation, every later decision reads as insiders managing their exit. Arbitrum's 44.47 percent insider block worked because the DAO treasury was larger still and visible on-chain. The same split without a credible treasury reads differently.

Cliffs that are too short and too synchronised. Keyrock's 16,000 unlocks make the argument: 90 percent negative, team unlocks worst at 25 percent, price impact starting 30 days early. Stagger the dates and lengthen the tails.

Incentives without sinks. Sky Mavis named this in their own dev journal before the collapse, and Terra is the same error with leverage on top. If you emit tokens to buy a metric, write down the mechanism that removes them again, and size it.

Governance theatre. Voting rights that can't change anything material are a utility claim that dies the moment a holder reads the contracts. Either give governance real control over the treasury and the parameters, or don't sell the token on governance.

A tokenomics design checklist

Run this tokenomics checklist before the token paper goes out. Every no is a design task, not a disclosure task.

Where to get help designing token economics

Tokenomics sits where economic design, contract engineering and compliance meet, and the teams that get it right have all three in the room before launch. Once you're past the whiteboard, our guide to creating a cryptocurrency covers the build side, and Web3 development sets out the wider stack a token lives in. If your token touches lending, staking or trading, the mechanics in DeFi and where the trend is heading shape much of the sink design, and for anything backed by off-chain value, real world asset tokenization covers the custody and legal wrapping.

We build the systems these designs turn into. On SeedBox, a crypto venture investing product, we shipped token holder governance, client tiers, loyalty points and a referral system alongside individual investor wallets, KYC checked onboarding, fiat and crypto rails and dynamic investment pools, with the Solidity contracts written against a product requirement document built at the idea stage. On Glue, we built a mobile crypto trading app anchored on gold as the base asset, with a real time dashboard, exchange pair switching, a supply and borrow lending side and a no hidden fee policy in the interface.

Neither was a token launch advisory engagement, and we'd rather say so than imply otherwise. What they show is the engineering half: contracts, wallets, compliance flow and the product a token economy runs on.

Good tokenomics isn't a clever supply curve. It's a set of decisions, each defensible in writing, that give someone a reason to hold the token which doesn't depend on someone else buying it later. Write the reason down first. Everything else here checks whether that reason survives.