Learning From Failures to Build Better DAOs
No silver bullets - just heuristics and hard-earned ideas for fixing DAO governance
There is no silver bullet for fixing DAO governance.
It is the realm of people problems, not technical problems. And people problems are really hard to solve for.
Still, there are promising directions worth exploring, based on heuristics and lessons from real-world experimentation. None of these ideas are guaranteed to work, but they are worth trying as we search for the next steps in evolving DAO governance into something more effective.
We’ll first explore a set of protocol goals that we may want to strive for in the design of a DAO. Then, the discussion will focus on how the goals can be applied to the ArbitrumDAO. Of course, the discussion is applicable to other ecosystems tackling similar issues.
Considerations for Operating a DAO
Running a DAO is not just about voting — it is about making decisions, executing them, and ensuring accountability.
This process starts with a clearly defined voting system, a collection of voters, and a governance surface that outlines what the DAO controls. The voting population should reflect those with a long-term interest in the DAO.
The output of the voting system is a proposal which includes a set of actions that need to be carried out once approved by the collective. All participants should embrace a culture of collective responsibility: disagree and commit once a proposal has passed.
Yet, governance is costly.
Governance demands time, attention, and coordination. It takes time for everyone to come together, read a proposal, discuss with each other, and eventually make a decision on whether it should be approved. As such, only proposals that can truly entice everyone’s attention should go through this process.
Put another way, governance must be minimal for it to be effective and the entire process should be reserved for matters of great significance.
The DAO can approve actions, but it needs a team to carry them out. This requires two things:
Trusted Teams. The DAO should have a set of dedicated teams who are capable of executing the work.
Decision-Making Body. An agile decision-making body should be empowered to handle post-approval proposals and oversee their implementation.
The team and decision-making body can work together to iterate on the operational details of a proposal as long as it remains in the spirit of what was approved by the DAO. This is in line with governance minimalism as the proposal can focus on desirable outcomes and trusted teams work out the finer details as more information becomes available.
Finally, DAOs and governance in general, must be resilient to capture. In the sense that no actor can push self-serving proposals through the process or block the process entirely. Governance only matters if decisions are executed as promised. To that end, DAOs need clear structure to ensure accountability and follow through.
Protocol Goals for DAOs
Here is an incomplete list of goals that are important for DAOs and effective governance in general:
Autonomous Execution. Token holders should be able to directly enforce their collective will over the systems and resources the DAO governs.
Governance surface. The DAO should define what it has control over and its authority should be clearly defined.
Multi-Stakeholder Participation. Relevant stakeholders should have both the incentive and means to participate in governance. The influence of a stakeholder on the decision-making process should be measurable and the governance system should reflect all stakeholders.
Optimistic Decision-Making. A decision-making process that allows decisions to be made swiftly on behalf of token holders while still upholding the DAO’s right to veto or override any such action.
Flexible, Accountable and Trackable Proposals. An organisational layer that makes approved proposals easy to review, track, and hold accountable. This structure should allow proposals to be flexible in execution as long as the outcomes remain aligned with what was promised to the DAO.
Governance Minimisation. The decision-making process should be invoked infrequently, or only when needed, to make decisions that may impact all stakeholders in the ecosystem.
Resistance to Capture. It should be difficult for a single party to capture the entire decision-making process in the DAO, and if it occurs, there should be a mechanism available to deter and veto it.
Let’s explore whether any of these goals have been met, and if not, consider how they might be achieved.
Autonomous Execution
To date, the ArbitrumDAO has focused on Autonomous Execution, and in my view, they have done an overall good job at achieving the goal.
The suite of governance smart contracts control the treasury and has the power to upgrade the smart contracts. Beyond emergency and non-emergency patches to the codebase, as entrusted to the security council, all upgrades are approved via the proposal process in the DAO.
There is still an argument on whether security councils undermine autonomous execution and I have written about this previously. My short answer is no. Security councils are absolutely necessary, and in some ways I hope they never disappear.
When it comes to governing legal entities, the ArbitrumDAO generally reigns supreme with ultimate control. Token holders can replace directors of the Arbitrum Foundation, they must elect the AGV Council on an annual basis, and similarly so for the OpCo as well. There is an interesting debate about the long-term viability of Foundations which you can read here.
Governance Surface
Autonomous execution is only meaningful if the DAO’s authority is clearly defined. Governance surface determines what the DAO actually controls and to what extent.
This is where programming and law diverge in how they define control. In software, control is explicitly and precisely defined with little room for ambiguity. In contrast, the legal realm operates through legalese, a language designed for broad coverage and crafted to anticipate future developments and address unforeseen circumstances.
DAOs must navigate the tension between precision and broad coverage by leveraging both approaches to define control.
Ideally, a proposal should clearly state what the DAO will control if the vote passes. Examples of control includes the authority to claw back funds, authority to replace the team, execute on-chain actions or change configuration details such as extending a proposal’s term.
In practice, it has been easier to express the intention of control in the text and the final details are handled after the proposal has passed. For example, writing the smart contract code or preparing relevant legal documents like service level agreements or bylaws.
Interpretation issues do arise from time to time as proposal authors are not well-versed in defining governance surface of control mechanisms. This is why there must be an empowered body on behalf of the DAO that can act as an authority to interpret what the DAO has voted on. If there is a dispute between this body and the authors, then the matter should be brought forth to the DAO to adjudicate on.
One issue that does arise is scope creep after a proposal has passed. Voters may demand changes or add new conditions not originally included in the approved scope. While these demands are sometimes well-intentioned, they can create uncertainty and shift power towards micromanagement. There is no right answer on how to deal with this situation and generally should be evaluated on a case-by-case basis.
Governance Minimalism to Fix DAO Participation
The ArbitrumDAO is already remarkable in one respect: there are numerous occasions where over 70% of explicitly delegated tokens consistently participate in votes. This demonstrates that engaged stakeholders are actively involved in its governance.
However, given the size and diversity of the ecosystem, many key contributors including application builders, DeFi protocols, gaming teams, Orbit chains, and others who are building on top of Arbitrum, are missing in action. This has led to some issues around quorum thresholds as the total supply of votable tokens continues to increase every month.
The issue of multi-stakeholder participation in the ArbitrumDAO is not related to token distribution. The problem is accessibility, awareness, and perceived value of actually participating. Fixing this requires resetting expectations in the DAO, especially around the type of proposals that are brought forward and the overall volume of governance activity.
There are three key metrics for evaluating participation in the DAO:
Voting participation. Growth in the quantity of delegated tokens and the total number of votes cast per proposal.
Stakeholder Engagement. Total stakeholders consistently kept informed about proposals alongside efforts to collect feedback when relevant.
Execution Alignment. Presence of aligned teams actively executing DAO-approved proposals on behalf of the team.
Let’s consider voting participation and stakeholder engagement. The first step is personal outreach. They should be invited to participate in governance. There should be one-to-one teach-ons to walk the stakeholder through how the DAO works, why they should care for it, and how to get involved. At a minimum, the goal should be to encourage delegation of tokens, and ideally, active voting.
We must also respect that not every stakeholder wants to engage in the public. Some prefer private discussions. That is perfectly fine. We should embrace different modes of communication as long as all relevant stakeholders can remain informed and have an opportunity to provide constructive feedback. The goal is to meet stakeholders where they are and not force them into a specific mode of engagement.
It may be the case that a trusted team, like the OpCo or Arbitrum Foundation, acts as a point of contact to gather feedback, facilitate private conversations, and integrate stakeholder input into a proposal. Additionally, they can produce executive summaries, host debriefing calls, and support stakeholders in understanding key details.
To avoid voter apathy and fatigue, the DAO should only process a handful of proposals at any given time. We should work towards absolute minimal governance activity and consider it exceedingly expensive to invoke. A proposal should only go through the process if it is indeed truly significant and the proposal’s impact can truly entice the interest of relevant stakeholders.
Additionally, toxicity is one of the fastest ways to drive away participation, and the DAO should adopt a zero tolerance approach towards bad-faith actors, hostile behaviour, or agents who actively seek to bring conflict into the public arena outside of the DAO. A maximum politeness policy should be the cultural norm, conflicts should be resolved amongst active participants in the DAO, and wealth creation games should always be prioritised over status games.
Finally, participation is not limited to voting, but executing the DAO’s will. Teams entrusted to execute on proposals should have proper alignment with the ecosystem. They should care more about the long-term success of the ecosystem over their own short-term interests.
If a team’s survival depends on a proposal moving forward, it creates a perverse incentive to keep a failing initiative alive rather than acknowledge when it should be shut down. This is primarily why we should no longer entertain vendor-specific proposals.
The Arbitrum Ecosystem is exceedingly lucky.
There are five teams, with 3 teams organically emerging from the DAO, who are willing to execute on proposals on behalf of the DAO. This includes Offchain Labs, Arbitrum Foundation, Entropy Advisors, Arbitrum Gaming Ventures (prev. Gaming Catalyst Program), and the Operations Company (OpCo).
If the ArbitrumDAO is successful, and the Arbitrum ecosystem is successful, then ultimately these companies will also be successful. We should take advantage of them to do more work on behalf of the DAO.
Optimistic Governance and Flexible Proposals
The solution sounds obvious once it is stated: complement the waterfall model with an agile methodology to enable iterating on the proposal as more information becomes available.
In practice, this requires a secondary governance mechanism to be implemented to quickly approve changes as long as they remain within the spirit of what the DAO approved.
The second governance mechanism can take many forms:
Legal entity for oversight alongside a service-level agreement,
DAO-elected committee empowered to approve deviations,
Optimistic model where the author publicly proposes changes and proceeds unless an explicit community-wide vote vetoes the decision.
In the ArbitrumDAO, the second governance body mechanism is expected to be fulfilled by the Oversight Committee (OAT) and the Operations Company (OpCo).
The OpCo should maintain a dedicated team to monitor the implementation of approved proposals, follow all workstreams, and hold accountable the proposal’s progress. If a proposal requires a pivot or adjustment, the OAT can review and approve those changes on behalf of the DAO.
With the above in mind, proposals should no longer be rigid or read like a cooking recipe. They should be goal-oriented with an initial plan on how to achieve the goal. The plan should be expected to change over time as the team iterates and the governing body can validate changes to ensure the proposal remains aligned with the DAO’s priorities.
Finally, proposals should not be tied to specific vendors. Instead, the outcome of a proposal should be owned by a trusted team accountable to the DAO. That team should manage procurement, select and onboard service providers, and terminate them as needed. It should not require a return to the DAO proposal process to remove service providers and the public dispute that entails.
Resistance to Capture
There are three key components that every DAO must focus on to prevent capture:
Defending the integrity of the DAO’s voting system,
Enforcing control and autonomy when executing the DAO’s will,
Mitigating operational risk by eliminating single points of failure.
In regards to defending the integrity of the DAO’s voting system, a core pillar of governance is enabling stakeholders to express their views while ensuring no single actor’s voice carries disproportionate weight relative to their expected influence.
The principle is not so different from how proof of stake or proof of work protocols operate. In those systems, a validator with a certain percentage of voting power is expected to have an equivalent percentage of influence over the network.
DAOs should strive for the same standard.
Voting power is intentionally measurable through the voting system and directly reflects a participant’s influence in the decision-making process. This is why it is critical to encourage token holders, especially large ones, to actively participate in governance. The governance platform should accurately represent the views of all stakeholders in the ecosystem.
Capture is not just about pushing proposals through the process, but the ability to oscillate the decision-making process and block any progress.
Governance only works when there are conflicting views that generally all agree on the same direction, forcing compromising, negotiation, and rethinking approaches that people will eventually approve of. If we end up in a world with extremely divergent views, then no progress is made, or worse, the community splits. Effectively the same impact as capturing the DAO.
Influential bodies may emerge who can exert soft power over governance, even without holding significant voting power. This is why onboarding token holders to actively participate in governance and improving the distribution of hard power through voting is essential. Broadening participation helps dilute influence of any single entity and reinforces the legitimacy of the process.
There is an academic paper that studies decentralization in voting blocks for a DAO and the ArbitrumDAO ranks fairly well on this front.
In regards to enforcing control and autonomy of the DAO, capture is often discussed in the context of decision-making. This includes attempts to weaken the DAO’s control to approve software upgrades, transfer treasury assets, or exert authority over legal entities acting on its behalf.
However, there is a second, less obvious form of capture that occurs at the execution stage of a proposal. Even if a proposal has passed through the legitimate governance process, the DAO may lack mechanisms to ensure that the work is carried out as promised. This can lead to de facto capture – not through overt influence, but through neglect, lack of follow up and insufficient accountability.
This tends to happen as DAOs are overly focused on a proposal before it passes, but then overlooks its execution due to more immediate and pressing priorities that arise later. Ensuring follow-through is not a problem that technology alone can solve. It requires optimistic governance and trusted teams to keep track of proposals, uphold promises and the rules, and if needed, seek recourse via the legal system on behalf of the DAO.
Finally, the whole premise of DAOs and blockchains in general, is to avoid single points of failure and build resilient systems. Put simply, the risk of operational capture. This is why I find the emergence of an ecosystem of firms and companies working on behalf of the DAO so promising.
The idea that a multitude of companies can work for the DAO first struck me around the time that Plurality Labs received a grant from the ArbitrumDAO (July 2023). It sparked a strange but exciting thought experiment:
What if a DAO doesn’t just fund isolated vendors, but instead supports a group of independent companies?
Each focused on a different domain with its own leadership and yet all aligned working towards the same shared mission?
Would that resemble the departments of a traditional organisation? Or is it something entirely new that is decentralised, adaptive and emergent?
In the ArbitrumDAO, it has taken a few years, but it has organically emerged. The DAO has bootstrapped three organisations including Entropy Advisors, Arbitrum Gaming Ventures, and Arbitrum’s Operation Company (OpCo). All teams, including the ideas for the organisations, came from participants in the DAO. So, the ArbitrumDAO has proven, this can be done.
I do believe there should be no playbook on how to establish an organisation that works for the DAO. After all, an organisation needs to be funded, which can cost hundreds of thousands of dollars, if not millions. The organisation should arise naturally within the compounds of the DAO and everyone should reach a rough consensus that it is indeed a viable thing to do.
Conclusion and takeaways
I’d like to take this opportunity to put together a list of heuristics that can help guide DAOs towards more effective governance based on everything we have discussed:
Governance Design and Boundaries
Decide the surface of governance and what the DAO should control,
Proposals should be goal-oriented, flexible and trackable, not rigid or recipe-like.
Optimistic governance enables agility and adaptation during execution,
DAO Governance is expensive – minimalism is a feature, not a flaw.
Organisational Execution
Trusted teams, not vendors, should pursue proposals,
Vendors should be hirable/removable without DAO-wide votes,
Focus on strategic outcomes and not vendor pitches,
DAO should not be deciding on operational details,
Councils are easy to create, but often slow to move – effectiveness depends on members and not the structure,
Traditional recruitment processes and active outreach work wonders,
Voter participation
Embrace large token holders and focus on complete representation of stakeholders,
Engage voters where they are, even if it is outside public forums.
Voters should participate due to self-interest, not incentives or compensation.
Participation should be lightweight, predictable, and worth a delegate’s time.
Respecting some stakeholders prefer private discussion and that is ok.
Produce executive summaries, regular calls, and keep voters up to date in a way that informs, but does not sway decision-making.
Culture and Resilience
Maximum politeness and resolve conflicts within the DAO,
Wealth creation games over status games,
Accountability is critical in all aspects of the DAO,
Foster an ecosystem of firms who care just as much, or more so, than the project’s founders.
Dedicated teams will emerge organically, support and fund them when the moment is right.
One final thought – an ecosystem is truly successful when it can outlive its original creators and founding organisations. When it can reach escape velocity and become self-sustaining. At that point, the genie is out of the bottle. The role of governance, and of DAOs more broadly, is not to stop that from happening, but to ensure that once the genie escapes, it can remain accountable, aligned, and hopefully manageable.