What we count as a public good
We count contributions that flow to entities producing non-rival, non-excludable goods for the Ethereum commons. In practice, that means Protocol Guild, Gitcoin matching rounds, Octant epoch allocations, gardens.fund grants, direct EF grants to open-source Ethereum infrastructure, and verifiable splits to non-profit research entities. The throughline is that the recipient is producing something that benefits the network whether or not the funder ever uses it directly.
We do not count grants to commercial subsidiaries, marketing programs, hackathon prize pools without onchain delivery, or donations to causes unrelated to Ethereum's open-source commons. The line is fuzzy. A grant to a venture-backed startup building a public RPC endpoint is one thing; a sponsorship of an industry conference is another. We are open to dispute on every entry — that is the point of publishing the methodology rather than only the rankings.
How we count ETH staked
Stake counted is the sum of validators an entity directly operates (it controls the keys) plus delegated stake where the entity has formal governance over reward routing. Lido and Rocket Pool are the obvious cases here: stakers delegate, but the DAO decides what happens with a slice of the rewards. For staking-as-a-service providers — Kiln, Figment, P2P.org, Blockdaemon — we count only the share they discretionarily redirect, not the full delegated stake they merely operate on behalf of customers. Treating SaaS-operated stake as if the operator chose to redirect it would overstate their commitment by an order of magnitude.
Time window
We use a rolling 12-month window. Lifetime totals reward entities that funded heavily three years ago and stopped, which is not what we want this dataset to track. Shorter windows — quarterly, epoch-level — get drowned in noise from the lumpy timing of grant rounds. Twelve months smooths out epoch cadence without flattering historical commitment that no longer reflects current behavior.
Grade derivation
Grades are derived directly from the redistribution percentage and absolute ETH amount in the 12-month window:
A → redistribution % ≥ 2% OR ETH ≥ 1,500
B+ → redistribution % ≥ 1% OR ETH ≥ 500
B → redistribution % ≥ 0.5% OR ETH ≥ 100
C → any non-zero redistribution
F → no public program
These thresholds are arbitrary. They are calibrated to current network conditions and will likely shift as the network matures and more entities adopt programs. We welcome arguments for different cutoffs, especially from entities currently graded C or F. If the floor is too low, A+ becomes meaningless. If it is too high, the grade collapses into a binary "did anything / did nothing" — which is what we already have without grades at all.
Data sources
- Octant — epoch reports
- Protocol Guild — splits contract data
- Lido Snapshot — DAO votes
- Rocket Pool — GMC grants charter
- beaconcha.in — stake data API
- Public Twitter announcements where formal records are absent
Limitations
Data lag is real. We update monthly, not in realtime. An entity that runs a grant round on the 28th will not show up here until the following month. For now this is a deliberate trade — realtime indexing of off-chain governance is expensive and not particularly useful for a dataset whose unit of analysis is "what did this entity do over a year."
Attribution is harder than it looks. When Coinbase Earn delegates customer stake across multiple operators, the allocation across those operators is often opaque. When a grant is co-funded by an entity and a third party, splitting credit is judgmental. We err toward visible, traceable contributions, which probably understates the true number for entities who keep their funding informal.
Subjective category boundaries are unavoidable. A grant to a venture-backed startup building public infrastructure — is that public goods or private capture with a thin commons veneer? We lean toward counting it if the artifact produced is open-source and operates without rent extraction, but reasonable people disagree. If you find a specific error, email us. If you find a structural critique of the methodology, write us a note — we will publish it inline as commentary.
How to dispute a grade
Open an an email with (1) the validator slug, (2) the disputed claim, (3) primary-source evidence — links to a DAO vote, a Snapshot proposal, an onchain transfer, a Protocol Guild split address, an Octant epoch report — and (4) your suggested correction. We respond within a week and update the dataset publicly with attribution. We do not arbitrate disputes privately.