Why two sites report different revenue for the same protocol

Look up one protocol's revenue on two research sites and the figures will not match. Sometimes they differ by a few per cent, and sometimes one is twice the other. Neither site has misread the chain. Each had to decide what a fee is, who it is paid to, and which payments come out before calling the rest revenue. They decided differently, and only some of them print the decision next to the number.

The fundamentals and research category describes the income-statement frame these sites share, "from fees through revenue to earnings". This page is about where each publisher draws those lines. TVL has its own entry, TVL, and the pipeline under any on-chain figure is in where an on-chain number comes from. The definitions below are quoted from each publisher's own documentation, read on 27 October 2026.

How it works

Money moves through a protocol in roughly the same order everywhere, and every published metric is a cut at one point on that path:

1. A user pays, or a deposit earns. Swap fees on a DEX, interest on a loan, gas on a chain, rewards on staked ETH. This is the top line, and all three publishers call it fees.
2. Suppliers get paid. Liquidity providers, lenders, stakers, node operators, integrators, and, for a rollup, Ethereum, which charges for posting data.
3. The protocol keeps the rest, in a treasury, with a team, or for token holders.
4. Token holders may receive part of that through a burn, a buyback or a direct distribution.
5. The protocol pays out its own token as incentives, and pays salaries and audits off-chain.

"Revenue" is a cut somewhere between steps 2 and 4, and the publishers put it in different places.

DefiLlama publishes the most explicit version, in its data definitions and in the guidelines its adapter authors work to. Fees "include everything the protocol could theoretically keep if it took 100% of the value flow" — for a lending market where depositors get 70% of interest, fees are 100% of the interest. Revenue is Fees - Supply-Side Revenue, "closer to Gross Profit in traditional accounting", and supply-side revenue covers "liquidity providers, lenders, stakers, node operators, as well as integrators, referrers and creators". Revenue then splits into Protocol Revenue, which goes to the treasury, and Token Holder Revenue. Earnings is "Revenue minus the value of token incentives the protocol emitted over the same period".

Token Terminal uses the same words in the same order, with one more line. Fees are "the total value users pay to use a project's application". Supply-side fees go to lenders or liquidity providers, and fees = supply-side fees + revenue. Revenue is "the portion of fees that a project retains, as determined by its take rate", measured "before cost of revenue, token incentives, and operating expenses". The extra line, cost of revenue, is what a project pays third parties "for a core part of its business after it collects revenue", and the definition's own examples are "node operator payments for liquid staking protocols or DA fees for L2 blockchains". Earnings is revenue minus all of that, and Token Terminal warns that the figure "may understate true costs" where operating expenses are not on-chain.

Artemis defines revenue from the token holder's side. In the framework it published with Pantera Capital for the S&P Pantera Digital Asset Index, revenue is "the value that accrues to tokenholders from a protocol's core operations, on a recurring basis". That includes treasury fees, burns, buybacks and direct distributions, and fees are "the total amount users pay". The same page says that "different data providers report materially different revenue and supply figures for the same token". Its warehouse tables split the cash into named buckets — service, treasury, token, burned, buyback and others — rather than leaving it as one revenue column.

Six places the definitions split

1. What counts as a fee. "Fees" does not always mean money a user paid. DefiLlama's Lido adapter puts all staking rewards in fees and sets user fees to zero, with the methodology note "Lido takes no fees from users". Token Terminal also counts Lido's gross staking rewards as fees, and its project page says it gets there by scaling the protocol's 10% revenue back up rather than measuring rewards directly. DefiLlama's guidelines also list what stays out: block rewards are "incentives, not fees", token taxes and referral payouts are not fees, and for a marketplace the fee is the charge on a sale, not the sale amount.

2. Who counts as a supplier. This is where the largest gaps come from. DefiLlama's supply side includes anyone the protocol pays for capital or service: node operators, rollup blob costs paid to Ethereum, integrators and referrers. Token Terminal pays suppliers before revenue but pays service providers after it, as cost of revenue. A node operator is a supplier on one site and a cost on the other, which moves the revenue line by the size of that payment.

3. Whether incentives are netted, and what an incentive is. Neither DefiLlama nor Token Terminal subtracts incentives from revenue. Both subtract them at earnings. What they subtract differs. DefiLlama's incentives are "liquidity mining programs, staking emissions, and, for chains, block rewards", and governance-token emission schedules go to its unlocks dashboard instead. Token Terminal's token incentives are tokens paid "to its supply/demand-side users, core contributors, and/or investors", described as "analogous to stock-based compensation". Earnings on the second site can take out team and investor tokens that the first never counts. The legacy Messari documentation goes the other way at the network level. Its asset "Revenue" is miner revenue, "fees plus newly issued native units", sourced from Coin Metrics. Issuance that DefiLlama treats as a cost is added to revenue there.

4. How a buyback is treated. Artemis puts buybacks, burns and buyback-and-burn programs inside revenue. DefiLlama puts them in Token Holder Revenue, and only when "funded by actual receipts", "measured at the fee source rather than at distribution". Treasury-funded buybacks do not count, and neither does a buyback already counted in a parent listing. DefiLlama also says Revenue = Protocol Revenue + Token Holder Revenue "is an attribution rule rather than a daily equality", because buybacks land on their own schedule. Its Lido adapter subtracts LDO bought back under the LDO Accumulation Program from protocol revenue, which is allowed to go negative on buyback days. Token Terminal's catalogue has buybacks as a separate measure next to revenue. None of the metric pages read for this guide says whether a buyback changes Token Terminal's revenue.

5. Chain or app. A chain's fees can mean gas alone, or gas plus the fees charged by apps built on it. DefiLlama keeps them apart: "for chains, Fees is only the transaction fees paid by users", app fees sit under the apps, and Hyperliquid's perp trading fees are tracked under the perp listing rather than the Hyperliquid L1. Priority tips paid to validators, such as Jito tips, have a listing of their own. Artemis publishes two families. CHAIN_FEES covers the base layer, while FEES on Ethereum "also folds in application-level fees from DeFi protocols built on top". Its Solana REVENUE includes "Jito tips accruing to SOL stakers", which Solana's FEES does not count, so its documentation shows Solana revenue above Solana fees on 1 August 2026. On DefiLlama, chain revenue is only "the burnt portion of transaction fees (or fees flowing to the chain's treasury)", and validator and staker payouts are supply-side.

6. The window. All three publishers work from daily rows, and each turns them into longer periods its own way. DefiLlama books on an accrual basis — "a year of borrow interest repaid today is not today's revenue" — and its free fees endpoint returns the same series as total24h, total7d, total30d, total1y and annualized1y. Token Terminal's screener windows run 1, 7, 30, 90, 180 and 365 days, all ending on an as_of date. Its P/S definition says "annualized revenue" but does not say which window is annualised. Artemis labels a quarterly bucket with the quarter's first day, and its summarize=true option returns the end date's single value rather than a total for the range. Even the arithmetic adds a gap: the last 30 days times 12 is about 1.4% lower than the same 30 days times 365/30, and either can be well away from a trailing 365-day sum.

Where the three agree, and two protocols where they provably split

Ethereum is the baseline. All three tie Ethereum's revenue to the base fee that is burned. Artemis says ETH's REVENUE "is the USD value of ETH burned via EIP-1559". Token Terminal's project page says ETH holders "take a cut of the total transaction fees (base fee)" and that validators' share is supply-side. DefiLlama's chain rule counts only the burnt portion. Where the money has one destination and no third party is paid, the definitions agree.

Lido: one 10% fee, two revenue lines. Lido takes 10% of staking rewards and splits it between node operators and the DAO treasury.

• Token Terminal's Lido revenue is "a protocol take rate applied to staking rewards, which is split between the protocol treasury and node operators" — the whole 10%. The operators' share goes below revenue, as cost of revenue, which Token Terminal derives from the stETH "minted and distributed to staking modules to pay node operators".
• DefiLlama's Lido adapter says "Revenue is only the DAO-treasury portion of that fee (net of the node-operator share, which is a cost of production booked as SupplySideRevenue)". After Lido V2 went live on 15 May 2023, it reads the split from Lido's StakingRouter contract for each day. Before that it uses the fixed 5%/5% of the V1 era.

So DefiLlama's Lido revenue corresponds to what Token Terminal calls gross profit. While the V1 split held, Token Terminal's revenue covered twice the share of rewards DefiLlama's did. After it, the ratio follows the on-chain split, which the adapter's comment records changing for staking module 1 from 5%/5% to 3.5%/6.5% on 24 December 2025. The adapter also records its own earlier fallback, which booked the operator share as revenue and "overstated protocol Revenue ~2x" for the V1 era. A Lido series downloaded before that fix and one downloaded after it disagree with each other, on the same site.

Base: the rollup's Ethereum bill, above or below the line. A rollup charges users gas and pays Ethereum to post its data.

• Token Terminal's Base revenue "treats all gas fees paid by users as protocol revenue". Its cost of revenue is the L1 settlement fees, tracked from Base's batch-sender transactions on Ethereum.
• DefiLlama computes rollup revenue as L2 gas fees minus the L1 calldata, blob and verification costs. In its shared rollup helper, those costs come from the rollup_economics_ethereum.l1_fees table on Dune, so part of DefiLlama's Base revenue is computed from a third party's table.
• Artemis's Base schema agrees with DefiLlama. treasury_fee_allocation is "fees - L1 costs", and revenue is listed as "legacy naming" for it. Its Arbitrum REVENUE is likewise sequencer fees minus the cost of posting data to Ethereum.

Two of the three subtract Ethereum's cost before calling the rest revenue, and one subtracts it after. Under Token Terminal's two Base definitions, fees and revenue both cover all the gas users paid. On the other two sites, Base revenue is fees minus whatever Ethereum charged that day.

What a multiple inherits

Token Terminal's P/S ratio (circulating) is "the circulating market cap relative to annualized revenue retained by the project". Each part is a definition:

• The denominator is revenue in Token Terminal's take-rate sense, before cost of revenue. A P/S rebuilt on DefiLlama's Lido revenue uses a smaller number for the same protocol and day. Under the V1 split, that means a ratio twice as high, with nothing about Lido different.
• The annualisation turns a window into a year by a method the metric page does not state.
• The numerator has its own definition: what counts as circulating. Artemis publishes a supply waterfall — outstanding supply excludes foundation, labs and DAO treasury holdings, and circulating supply further excludes unvested team and investor tokens — which is a different definition again from one built on total supply. See circulating supply.

Artemis's protocol-economics recipe makes the same point about fees against revenue: a multiple on one and the same multiple on the other "can rank the same set of protocols in a completely different order". A ratio is a definition divided by a definition. It is a valuation input, and it says nothing about whether a token is priced high or low.

What it costs

The definitions are free on all three sites. Getting the numbers by API is what costs money.

• DefiLlama serves fees and revenue on its free dashboards and through its public API with no key. The fees summary endpoint returns, beside the numbers, a methodology object, a breakdownMethodology per label, and a methodologyURL pointing at the adapter's source file. The card notes the trade-off: adapters are usually written by the protocol being listed and reviewed by DefiLlama, and the repositories carry no LICENSE file.
• Token Terminal publishes a general definition for every metric and a project-specific methodology for each project, readable on the free plan with full history. Pro is 350 USD a month, and the API is quoted on request. Coverage varies by metric: when the card was checked on 19 September 2026, fees covered 323 projects and earnings 233.
• Artemis documents its metrics, aggregation rules and warehouse buckets in public. API access starts with the 500-dollar-a-month spreadsheet seat. The card also records that Artemis changed what a stablecoin field measured without renaming it, and that its public changelog was empty between 30 July 2025 and 27 May 2026.
• Blockworks Intel, formerly Messari, has no API endpoints on its free plan. Pro is 4,500 USD a year, and its API covers the Assets endpoints only. The fees and revenue definitions read for this guide are in its legacy documentation, for an API its docs say is in maintenance mode.

What you can do about it

Pick the line that matches the question, then pick the publisher.

• Activity generated: fees, which all three define similarly.
• What the protocol keeps after paying suppliers and service providers: DefiLlama's or Artemis's revenue, or Token Terminal's gross profit.
• What the protocol keeps before paying service providers: Token Terminal's revenue.
• What reaches token holders: DefiLlama's Token Holder Revenue, or Artemis's revenue with its bucket columns.
• Whether the protocol is profitable after paying in its own token: earnings, after checking which incentives that publisher subtracts.

Never compare two publishers on the word "revenue". To compare Token Terminal and DefiLlama, compare like for like: DefiLlama revenue against Token Terminal gross profit for liquid staking and rollups, or fees against fees. Then check that the windows match. The same protocol, the same day boundary and the same line usually close most of the gap. What is left is coverage: a contract one publisher indexes and the other does not.

Read the definition for the one protocol you care about. The general definition is not enough, because the split happens in the project-specific rule. On DefiLlama, open the adapter file the methodologyURL names; its methodology and breakdownMethodology strings say what each label holds. On Token Terminal, read the Methodology tab of the metric for that project. Both are free, and on both the Lido and Base answers above took one page each.

For a multiple, build it yourself rather than reading one off. Take one publisher's revenue series, sum the trailing 365 days rather than annualising a month, divide by a market cap whose supply definition you know, and write all three choices in the column header. A P/S ratio taken from a screen hides the window and the supply definition. A P/S ratio you built yourself shows both.

Questions to put to a vendor before you build on its revenue series:

• Are node-operator, sequencer or data-availability costs above or below the revenue line?
• Is a buyback inside revenue, beside it, or counted only when funded from fees?
• Do token incentives include team and investor vesting, or only emissions to users?
• Which window does an annualised figure use, and is the latest day complete?
• For a chain, do fees include the fees of apps built on it, and where do priority tips go?
• When a definition changes, is history restated, and where is that announced?

The last question matters because restating is the default. Token Terminal says it keeps raw data so that "when new smart contracts are deployed or methodologies change, we can re-process historical data within minutes". DefiLlama's review checklist asks whether a change to a live adapter needs a refill "and from which date". If a report has to show the same figure next quarter, save the series with the date you pulled it.

Use the free tiers for what they cover. DefiLlama's public API, with its methodology fields, is enough to compare protocols on one consistent definition at no cost. Token Terminal's free plan is enough to read its definitions and history on screen. Artemis's documentation alone explains its chain-level choices. Pay for API access only when you need one of these publishers' definitions delivered as a series, and once you are paying, you are paying for that definition too.