On-Chain Governance: Models, Outcomes and the Participation Problem

The promise of on-chain governance — that stakeholders in a protocol can collectively determine its direction through transparent, verifiable, and censorship-resistant mechanisms — remains one of the most compelling ideas in the digital assets space. The reality of how on-chain governance actually functions in practice is considerably more complicated, and more instructive, than the idealised vision.

After several years of real-world operation across dozens of major protocols, a clearer picture is emerging of what on-chain governance can and cannot reliably deliver. The participation problem is structural, not incidental. Governance attack surfaces are real and have been exploited. And the trend toward constitutional frameworks suggests that the industry is beginning to recognise that raw token voting is insufficient as a complete governance system.

The Spectrum of Governance Models

On-chain governance is not a single mechanism but a family of distinct approaches, each with different properties around plutocracy resistance, participation efficiency, and attack surface.

Pure token voting is the simplest and most prevalent model. One token equals one vote. Token holders submit proposals, a voting period runs (typically three to seven days), a quorum threshold must be met, and if the proposal passes, it is queued for execution after a timelock delay. This model powers the major Compound, Aave, and early Uniswap governance systems. Its virtues are simplicity and transparency. Its vices are well-documented: wealthy token holders (whales) exert disproportionate influence, small holders have little individual incentive to vote, and the system’s openness to large token accumulations creates governance attack vectors.

Quadratic voting adjusts the cost of votes so that the marginal cost of each additional vote increases quadratically. In theory, this reduces whale dominance while preserving the ability of participants with strong preferences to amplify their voice. In practice, quadratic voting requires robust Sybil resistance — the assurance that one person controls one address — which remains an unsolved problem in pseudonymous blockchain environments. Gitcoin Grants has implemented quadratic funding (a closely related mechanism) for public goods allocation, using social graph verification as a partial Sybil mitigation.

Conviction voting takes a different approach entirely. Rather than discrete voting periods, conviction voting allows token holders to continuously signal preference for proposals. The longer a token holder maintains support for a proposal, the more “conviction” accumulates behind it. A proposal passes when it accumulates sufficient conviction. This model, implemented in systems like Gardens (developed by 1Hive), reduces the short-term gaming of discrete voting windows and rewards sustained community preference. It has not achieved mainstream adoption in the largest DeFi protocols.

Optimistic governance inverts the standard model. Rather than requiring affirmative approval for actions, optimistic governance allows a trusted executor (often a multisig or elected council) to propose and execute actions automatically unless challenged within a defined window. The Optimism Collective’s structure incorporates optimistic elements, as does the Council architecture used by several protocols to manage routine parameter adjustments. This approach improves execution speed at the cost of requiring vigilant participation to catch problematic proposals.

Security councils have emerged as an important complement to pure token governance. A security council is typically a small, trusted group of addresses (often between nine and twelve signatories in various M-of-N configurations) empowered to act rapidly in emergency situations — particularly security vulnerabilities requiring immediate contract pauses or upgrades — without requiring a full token governance vote that would take days. Arbitrum, Compound, and several other major protocols have implemented security councils. The tension between their efficiency and the decentralisation ideal is openly acknowledged and managed through transparency requirements and accountability mechanisms.

The Participation Problem

Perhaps the most persistent challenge in on-chain governance is participation. Data across the major protocols consistently shows that typically five to fifteen per cent of eligible tokens participate in a given governance vote. On contentious, high-profile proposals, participation can rise significantly — Uniswap’s fee switch votes have historically driven higher-than-average turnout. Routine parameter adjustments and grant distributions frequently attract less than five per cent of eligible token supply.

This is not a bug caused by poor user experience design (though UX friction has historically been a real barrier, particularly when on-chain voting required paying Ethereum gas fees for each vote cast). It is a structural consequence of rational ignorance: for the vast majority of governance token holders, the cost of developing an informed opinion on each governance proposal exceeds any individual benefit from doing so. A holder with ten UNI tokens has essentially no individual effect on a governance outcome. The time cost of reading, understanding, and voting on three proposals per week is real. The rational choice is not to vote.

The aggregate consequence of individually rational non-participation is collectively dysfunctional governance. Effective control of token voting systems is concentrated not among the broad token holder base but among a small number of large holders who do participate consistently — typically venture capital firms and early investors with significant positions, core contributor teams, and professional delegate operations.

Liquid Delegation

Liquid delegation was developed explicitly to address the participation problem. Under delegation systems, token holders can assign their voting power to a delegate address — an individual or organisation that actively participates in governance on the delegated token holder’s behalf. Delegates can be changed at any time, providing accountability without lock-up.

Compound pioneered on-chain delegation, building it directly into the COMP token’s governance module. ENS DAO has developed one of the most mature delegate ecosystems, with dozens of active delegates publishing governance rationales and voting records. Uniswap’s delegation system has similarly evolved, with specialised governance firms and university blockchain clubs operating as professional delegates.

The emergence of professional delegation services represents a genuine improvement in governance efficiency. An engaged, informed delegate representing millions of delegated tokens can participate meaningfully in governance in a way that individual holders with small positions cannot. The accountability mechanism — delegates can be fired by withdrawing delegation at any time — creates at least some incentive for delegates to represent their constituents’ interests faithfully.

The limitation of delegation as a solution is that it does not eliminate concentration — it rearranges it. A small number of delegate addresses accumulating large delegated positions may be more engaged than passive whale holders, but they are still a small group making decisions for a large community.

The Snapshot Layer and On-Chain Execution

A significant proportion of DAO governance does not occur fully on-chain. Snapshot, a widely used off-chain signalling platform, allows token holders to vote using signed messages without incurring gas costs or writing directly to a blockchain. Snapshot results are not natively executable — they are expressions of community preference that a multisig, security council, or other execution layer must then translate into on-chain action.

The Snapshot layer creates useful efficiencies: gas-free voting increases participation, and the ability to signal preferences without committing to on-chain execution allows more exploratory governance discussions. But it also creates a trust dependency on whichever entity operates the execution layer. The gap between a Snapshot vote and on-chain execution is where discretion — and potential manipulation — can enter.

Sophisticated governance frameworks attempt to define clear rules for when Snapshot signals are binding, when additional on-chain confirmation is required, and what circumstances permit the execution layer to decline to implement a Snapshot decision. These rules, where they exist, form an implicit constitution for the DAO’s governance process.

Governance Attacks and Attack Surface

The openness of token voting creates a well-understood attack surface that has been exploited in practice. Several distinct attack vectors deserve attention.

Flash loan attacks on snapshot exploit the point-in-time nature of voting power measurement. If a governance system uses a token balance at a single block height to determine voting power, an attacker can borrow an enormous token position via flash loan, vote, and repay within a single transaction. This attack vector led most governance systems to adopt a “checkpoint” model where voting power is measured at a block prior to the proposal’s creation — requiring the attacker to actually hold tokens through the checkpoint, which involves capital commitment and counterparty risk.

Whale dominance is a subtler but more persistent concern. Large token holders — particularly early investors and venture firms with positions acquired at low cost — can dominate governance outcomes on proposals where the broader community does not mobilise. This is not necessarily malicious, but it does mean that governance outcomes may reflect the interests of early capital rather than the broader user community.

Last-minute vote swings occur when large holders observe a vote trending toward an outcome they oppose and deploy tokens in the final hours of a voting period to reverse the result. Some governance systems have implemented “vote extension” mechanisms: if a significant vote occurs near the end of the voting window, the window automatically extends by a defined period to allow counter-mobilisation. This is a reasonable safeguard but adds complexity.

Consequential Decisions: Three Case Studies

MakerDAO’s RWA allocation programme stands as perhaps the most consequential governance decision in DeFi history. Over several years of governance votes, MKR token holders approved the progressive deployment of billions of Dai reserves into real-world assets: US Treasury instruments, institutional lending facilities, and tokenised funds. The decisions required deep technical and financial expertise to evaluate properly, yet they passed through the standard token voting process. The outcome has been financially successful, but the process revealed the limits of amateur participation in highly specialised financial decisions.

Uniswap’s fee switch has been the subject of repeated governance campaigns. The question — whether UNI token holders should receive a share of Uniswap’s substantial protocol revenues — touches directly on securities law concerns, token economics, and contributor incentive structures. After multiple votes across several years, the outcome remains contested, illustrating how high-stakes economic decisions can become captured in governance gridlock when different constituencies have irreconcilable interests.

Compound’s security council creation represented a pragmatic acknowledgement that pure token governance was inadequate for security-critical decisions. The establishment of a faster-response security council, empowered to act without full governance votes in emergency situations, marked a recognition that governance speed and security had to be weighed against the ideal of full decentralisation.

Constitutional Frameworks and Governance Minimisation

The response to participatory and security failures in governance has been a trend toward constitutional frameworks — explicit documents or on-chain rules that constrain what governance can decide, how decisions can be made, and what processes are required for different categories of action.

Optimism’s governance constitution, ENS’s governance process documents, and Arbitrum’s governance framework all represent attempts to build rule-of-law structures above raw token voting. These frameworks define supermajority requirements for high-stakes decisions, mandatory deliberation periods, and non-waivable protections for core protocol properties.

Simultaneously, a “governance minimisation” school of thought argues that the ideal DAO is one that requires as few governance decisions as possible. If protocol parameters are set at launch and rarely need adjustment, and if the only on-chain governance required is for major upgrades, the attack surface and participation burden are both minimised. Protocols designed for governance minimisation accept some loss of adaptability in exchange for reduced governance risk.

Swiss Governance Practices

Among Swiss-domiciled foundations, governance architecture varies considerably. The Ethereum Foundation famously does not use token voting — the Ethereum protocol’s development direction is determined through a rough-consensus process among core developers and the broader research community, not through ETH token governance. The Web3 Foundation’s Polkadot protocol employs one of the most sophisticated on-chain governance systems in the industry, having evolved through multiple iterations (v1 Council-and-Referendum, OpenGov with its multi-track referendum system) toward a model that attempts to balance broad participation with decision quality.

Conclusion

On-chain governance in 2026 is more mature, more battle-tested, and more clearly understood in its failure modes than it was in 2021. The participation problem is structural and will not be solved by UX improvements alone. The governance attack surface is real and requires active mitigation through timelocks, checkpointing, and security councils. Constitutional frameworks and governance minimisation represent the direction of travel for sophisticated protocols that have learned from experience.

The enduring question — whether decentralised governance can actually produce better outcomes than centralised alternatives — remains open. The evidence to date is mixed: some DAO governance decisions have been genuinely good, others have been slow, captured, or simply wrong. What is increasingly clear is that raw token voting, applied uniformly across all decision categories, is not the answer. The protocols that will govern successfully in the long run are those that match their governance mechanism to the decision at hand.

Donovan Vanderbilt is a contributing editor at ZUG DAO, a publication of The Vanderbilt Portfolio AG, Zurich. The information presented is for educational purposes and does not constitute investment advice.