Best Incentive & Grant Program Operators

Best incentive and grant program operators by use case

No operator leads on every criterion. Match the job you need done to the marks that measure it.

• An outside evaluation of a past program: Castle Labs. A program evaluator with 10 for impact report: its retention study defines its metrics and is backed by a public Dune dashboard, and its May 2025 Arbitrum assessment covers 11 programs. It is one of two operators with a general independence statement (5). OpenBlock Labs, the other evaluator, adds a written-out sybil method (abuse protection 10) but discloses no ties to the grantees it evaluates.
• A matched-funding grants round with sybil defences: Gitcoin. A grants administrator with 10 for abuse protection: rounds document Passport score thresholds, no matching for contributions under $1 and cluster-aware matching (COCM). It has no conflict-of-interest policy (independence 0).
• Grants with every season's payouts accounted for: Questbook. The only operator with 10 for disclosed volume. Each season of the Arbitrum D.A.O. Grant Program reports disbursed totals, Season 3 at $3.4 million of $5.3 million allocated as of July 22, 2026. Approvals above 25,000 USDC need a second allocator, a control that helped stop a fraudulent $35,000 payout in March 2026.
• Grants released only against reviewed milestones: Thrive Protocol. Milestones are scored from -3 to +3 and paid when a reviewer threshold agrees, and its Polygon reports set each grantee's target against its result (impact report 10, abuse protection 10).
• Reward distributions anyone can check: Merkl or Metrom. These are the two operators with 10 for open data. Merkl posts one Merkle root per chain, keeps the reward file on a public status page and gives anyone one to two hours to contest each update. Metrom's keyless API lists 66 campaigns with chain, dates and amounts; it also has a measured impact study (10), where Merkl's outcome figures carry no stated method (5).
• Quest campaigns or airdrops that need a one-person gate: Galxe. The only reward distributor with 10 for abuse protection: Galxe Passport runs a Sumsub identity check and issues a non-transferable credential. Its campaign outcomes are reported without a method (impact report 5).

What does an incentive program operator do?

Operators are the firms a chain or DAO hires to run the money side of an incentive or grant program; the Role column shows which job each does. Two are program evaluators (Castle Labs, OpenBlock Labs). Four are grants administrators (Gitcoin, Karma, Questbook, Thrive Protocol). Six are reward distributors (Boost, Galxe, Layer3, Merkl, Metrom, Turtle).

How the ranking works

Operators are scored only on what they publish, on their own pages or the forum where a program ran. All six criteria use fixed mark ladders, and a criterion on which nothing is published scores 0:

• Named programs: programs named individually with network and period. 0 for none, 5 for one or two, 8 for three or four, 10 for five or more. A stated run length counts as the period; for an evaluator, the programs it assessed count.
• Disclosed volume: the amount distributed for the programs it names. 0 for none, 5 for some programs or an aggregate, 10 for every named program. Budgets, funding targets and capital committed are not distributed volume, so an operator whose only figure is a budget scores 0.
• Impact report: 0 for none, 5 for a report whose figures come without a stated method, 10 for metrics with a stated method of measurement.
• Abuse protection: 0 if sybil attacks and farming are not addressed, 5 if they are mentioned, 10 if the mechanism itself is described. Eligibility filters, allowlists and per-user caps earn 5; 10 needs a description of how a person is verified and de-duplicated, how sybil clusters are detected or how payouts are reviewed before release.
• Open data: 0 if program data cannot be inspected without an account, 5 if dashboards or exports are partly open, 10 if the data is fully open with no login or key.
• Independence: 0 for no disclosure of ties to fund recipients, 5 for a general statement, 10 for a written policy plus a list of the relationships.

The overall score is the plain average of the six marks at equal weight, rounded to one decimal: 40 points over six criteria gives 6.7. Ties are listed alphabetically, which puts Castle Labs, Gitcoin and Metrom in that order at the top. The marks come from desk research of public pages, all read on November 6, 2026; no account was opened and no program was joined. Marks are re-checked when sources change, and each change is logged in the operator's review. The full rules are on our ratings methodology page.

#	Operator	Role	Named programs	Disclosed volume	Impact report	Abuse protection	Open data	Independence	Overall
1	Castle Labs	Program evaluation	10	5	10	5	5	5	6.7
2	Gitcoin	Grants administration	10	5	10	10	5	0	6.7
3	Metrom	Reward distribution	10	5	10	5	10	0	6.7
4	OpenBlock Labs	Program evaluation	8	5	10	10	5	0	6.3
5	Questbook	Grants administration	8	10	5	10	5	0	6.3
6	Thrive Protocol	Grants administration	8	5	10	10	5	0	6.3
7	Galxe	Reward distribution	10	5	5	10	5	0	5.8
8	Merkl	Reward distribution	10	5	5	5	10	0	5.8
9	Karma	Grants administration	10	5	5	5	5	0	5.0
10	Layer3	Reward distribution	8	0	10	5	5	0	4.7
11	Turtle	Reward distribution	5	0	5	5	0	5	3.3
12	Boost	Reward distribution	0	0	0	5	5	0	1.7

What changed in this update

For November 2026 every source was re-read and each ladder was read the same way for all twelve.

• Budgets no longer pass as payouts. Castle Labs moved from 10 to 5 on disclosed volume because its assessment labels most per-program figures as budgets; its overall went from 7.5 to 6.7. Turtle moved from 5 to 0, since its one reward figure is the 50,000 AVAX Avalanche Awakening budget; overall from 4.2 to 3.3. Gitcoin also lost 5 points on volume, as its GG24 domain figures ($1,807,500) are funding targets, and gained 5 on impact for the Public Goods Tooling retrospective, so its overall held.
• Eligibility filters score 5. Merkl's World ID requirement and per-user caps are filters rather than a detection mechanism, so abuse protection moved from 10 to 5. With 14 case studies now lifting named programs to 10, the net move was from 6.3 to 5.8.
• A stated run length counts as a period. Galxe now has five programs with network and period, moving named programs from 8 to 10 and its overall from 5.5 to 5.8.
• New evidence. OpenBlock Labs' Solana RetroPGF Round 1 page adds a third dated program (overall from 5.8 to 6.3). Karma added Filecoin ProPGF Batches 2 and 3 and a stated $81,000 GoodDollar payout (from 3.8 to 5.0). Layer3 gained for defined new-user metrics and its own public Dune dashboards, but dropped to 0 on volume because its 400,000 ARB grant is a budget (from 3.8 to 4.7).
• Unchanged: Metrom, Questbook, Thrive Protocol and Boost.
• Withdrawn: Blockworks Advisory, whose advisory practice folded into Blockworks Intel; SeriousPeople, whose site is offline with no confirmed program role; and Royco, which pivoted to yield tranching and no longer runs incentive programs.

What the scores do not cover

• Disclosure, not quality. The marks record what an operator publishes. They do not test contracts or software, or judge whether a program was good value; fees and operators' own tokens are not scored.
• Self-reported figures. Most numbers are company statements, such as Merkl's homepage counter of $1,813,562,101 distributed. Metrom's impact study is self-published and covers one program.
• Data that disappears. OpenBlock Labs' public dashboards now return 404, Gitcoin's open dataset ends with Grants Stack in May 2025, and Metrom's API shows no campaign ending after September 18, 2026.
• The independence gap. Ten of the twelve score 0 on independence and none publishes a written policy with a list of relationships. Karma lists its own project among Arbitrum grantees without a conflict disclosure, and Metrom names no operating company.
• Participant risk. A reward campaign carries the risk of the protocol being rewarded, and rewards paid in a network's token change in dollar value. Our yield farming explainer covers those risks. Quest and airdrop links are also common phishing lures.

How to choose

1. Decide on the role first: an evaluator, a grants administrator and a distributor do different jobs and are not substitutes.
2. Ask what was actually paid out, program by program, not what was budgeted.
3. Ask how impact was measured, including retention after rewards stopped.
4. Ask how farming is detected, not only which wallets are filtered out.
5. Check whether you can inspect the data without an account or API key; Turtle's Earn API needs a key for every endpoint.
6. Ask for a written disclosure of ties to fund recipients, and put conflict terms in the contract, since only two operators publish even a general statement.

How can you tell whether an incentive program worked?

Look at what remains after the incentives end. A useful evaluation compares activity before, during and after the program, prices each retained user or unit of liquidity rather than the peak, and separates the program's effect from market conditions. OpenBlock Labs' Arbitrum reports, for example, normalize growth by ARB value and compare against control groups. A report that only quotes peak deposits while rewards were flowing measures the spend, not the outcome.

What is sybil farming, and how do operators defend against it?

Sybil farming is one participant splitting activity across many addresses to collect rewards meant for many people. Defences range from eligibility filters, such as minimum balances, allowlists and per-user caps, to mechanisms that find the farmers: identity checks that issue one credential per person, clustering of addresses funded from the same source, and review of payouts before release. Publishing the mechanism is what lets a program's funder check that it exists.