Shape, Shard, and Batch: Practical Heuristics for More Effective DAO Voting
*We are not presenting a template for all DAOs but generally applicable heuristics.
Myth: Onchain voting will solve the problem of the internet being in the hands of a few stakeholders.
Today’s internet has diverged from its original vision where instead of an internet that’s controlled by the community, powerful controllers in the forms of big tech and governments dominate.
"The good news is that billions of people got access to amazing technologies, many of which were free to use. The bad news is that it became much harder for startups, creators, and other groups to grow their internet presence without worrying about centralized platforms changing the rules on them, taking away their audiences and profits. This in turn stifled innovation, making the internet less interesting and dynamic." — Chris Dixon
Giant organizations controlled by a small group of founders and early shareholders can change the rules of the game at their whim. Who is affected by these dynamics matters. Changes by founders and early shareholders influence other stakeholders such as developers, users, and retail investors in several ways.
- Developers and third party complements building on platforms risk a bait-and-switch. This has happened before. Twitter first attracted third party developers and switched to close its network to third party developers. Facebook initially attracted third party developers to build apps increasing the utility of Facebook then shut down its API.
- Users fare no better than developers. Platforms enforce censorship that appears arbitrary. From a controversial deplatforming of a former president and shadowbanning of models to censoring the press.
- The internet made investing more accessible for retail investors. The Gamestop short squeeze highlighted that the stock market is less fair than what the small investor thinks and that most have zero understanding of market structures.
If all stakeholders cannot make their voice heard when it matters to them, what is the solution? How can we prevent a small group of stakeholders from taking advantage of others?
Different solutions to the problem of the internet being in the hands of a few exist and one of the solutions is stakeholders controlling internet platforms through onchain voting.
When a decision is made in the physical world, the decision maker(s) tasks an individual with seeing that the decision is executed. Most of the time, execution is flawless. There are a few cases in which the decision is not executed to satisfaction. For one, factors such as miscommunication can cause delays. In another, the person tasked chooses not to execute by adding/omitting parts of the decision or refusing to execute. Laws are what hold individuals responsible and guarantee that execution will happen most of the time. You are punished according to how you failed to oblige by the law. The greatest of these punishments is bodily harm. In the physical world, it is martial law that enforces execution.
Onchain voting simplifies this. Onchain voting as a control mechanism does not suffer from problems in the physical world. Onchain voting makes execution instant where there are no delays between decision making and execution. It also guarantees that a decision will execute all the time. In the digital world, laws written in smart contracts enforce execution. The need to trust a person for execution disappears as we are disintermediated.
Onchain voting can represent ownership and control of whatever is on the internet—including platforms and protocols. It empowers voices in a community with the ability to take meaningful action. Voting signals inclusivity when it increases the number of stakeholders, enables wisdom of the crowds, and promises to foster stronger connections through the democratic process.
What can be more open, transparent, and fair?
Myth-Buster: Every stakeholder having the right to vote is more important than every stakeholder exercising that right to vote.
Fueled by this enthusiasm the crypto community has applied onchain voting everywhere the past couple years. And we’ve seen varying results.
1. Onchain voting decreases risk for developers and third party complements building on platforms.
Miners vote to determine whether Bitcoin upgrades and changes its core rules. This may sound like the perfect setup for a small, closed, non-transparent group of miners to collude.
It doesn’t happen in practice because of several reasons. New entrants join and incumbents leave with changing market conditions. Miners must gain support from users, developers, wallets, and exchanges. They otherwise risk becoming the newest addition to many "hard forks". No single group of stakeholders can determine how Bitcoin should upgrade.
2. Onchain voting enables users to have a voice and enforce execution of key decisions in platforms—with controversial results.
Brantly wore several hats for a protocol called Ethereum Name Service (ENS). He was director of operations at True Names Limited (TNL), a Singapore nonprofit that organizes and funds ENS. He was the director of the ENS Foundation, a Cayman Islands legal entity governed by the ENS DAO. Brantly was a popular ENS delegates for ENS DAO. When Twitter surfaced a series of controversial tweets Brantly made, TNL fired Brantly right away.
In turn, ENS DAO moved slower and put a vote to remove Brantly as the director of the ENS Foundation. Although ENS holders undelegated tokens from Brantly, the vote did not pass. The ENS Foundation honored the result of the DAO vote and Brantly remained as the director and a top delegate.
A single entity in Juno unwittingly received a disproportionate amount of tokens in their stakedrop that was intended to be fair. The community voted on a total of four proposals to solve this problem. The community first sought to correct this error by reducing 90% of tokens the entity controlled. This first proposal did not pass at that time. Later the community put forth a second proposal that signaled intent to make the same correction as before, citing the growing risk to the network. A third proposal followed and with an onchain code that would execute to fork the ledger and send the token balance of the entity to a smart contract controlled by the community. The third proposal passed with a typo in the smart contract address which resulted in the tokens being burnt. The final proposal corrected the address typo from the third proposal.
3. Onchain voting enables retail investors to voice and enforce execution of their interests on a platform—with mixed results.
In 2016, The DAO raised around 14% of all circulating ETH at the time. The founders envisioned The DAO to empower small investors by giving them control of their funds through formalized onchain governance.
“In the case a single token holder or a group of token holders does not agree with decisions by the curator and following proposals to a certain address, he can split off the original DAO to a new DAO. This mechanism is primarily to avoid the ‘Majority robbing the minority attack.’” — The DAO Wiki
It included a “split function” that could safeguard an investor against a 51% majority-robs-the-minority attack (a precursor to “ragequit”). The irony was that the splitting feature introduced a vulnerability that enabled a hacker to drain funds from The DAO. Ethereum developers suggested a range of solutions right away. Discussions sprawled across SNS and forums like wildfire.
The Ethereum community reached a rough consensus to attempt a “hard fork” to save Ethereum. Stakeholders understood the hard fork was contentious. Community members, not wanting a split, devised various ways of measuring sentiment. A community-created site tallied ballots on whether the hard fork option should be the default, giving miners the ability to opt out, or the other way, where miners opt-in. The result was 87% for the hard fork option being the default with 5.5% of all ETH voting. The hard fork took place and the majority of the community stayed with Ethereum. The minority forked off to form Ethereum Classic.
TRIBE token holders control Tribe DAO, “a DeFi megaDAO made up of several sub-daos” such as Fei Protocol, Rari Capital Fuse platform, etc. Fei and Rari made headlines with the biggest DAO to DAO merger to date in December 2021. It took place through an onchain governance process resulting in both operating under the TRIBE token.
Tribe DAO chose to enhance DAO governance by upgrading its version of “optimistic approval” (OA). OA is a mechanism that grants a multisig address power to modify a narrowly defined set of protocol parameters where any actions carried out by the multisig can be vetoed by the DAO. This structure was immediately put to the test with the second Fuse hack. Knee jerk reactions put the issue under the industry limelight.
Discussions circled around a poor understanding of how the new governance structure works, the extent to which insiders can vote to exert control over the community, and arguments that the refund could negatively impact PCV collateral.
A couple of months passed and Fei Labs, the main entity behind Tribe DAO, proposed to pay Fuse victims and withdraw their participation from the DAO. They would, in effect, wind down Tribe DAO. Another set of activities and discussions ensued: arguments for full repayment especially from other DAOs such as Frax and Olympus DAO who’ve contributed to Fuse, a bribe to align incentives between TRIBE holders and Fuse victims to swing votes, opaqueness and silence from team and VCs on how much assets each stakeholder could redeem when the DAO liquidates, and a surprise last minute appearance from the team and VCs to swing a vote in their favor.
Merit Circle DAO (MC) is a play-to-earn gaming guild with an incorporated company, Merit Circle Limited (MCL), as its core team. MC had caught the attention of several investors, one of which was Yield Guild Games (YGG), and raised funding through a SAFT agreement. Two points were notable: first, MC tokens were to vest over a predefined schedule regardless of whether an investor added value to MC; second, Merit Circle Limited signed the agreement that stated all authority over tokens were to be transferred from the incorporated company, MCL, to Merit Circle DAO. This transferred responsibility over distributing tokens to a DAO that did not have legal personhood.
After the raise, MC started a voluntary transparency and accountability initiative. It brought up concerns over early investors' value-add and their token release schedules. Each early investor wrote a report to communicate how they’ve added value to MC. A MC contributor created a proposal that MC renege on YGG’s SAFT, refund YGG’s initial investment, and remove YGG’s token allocation—highlighting that YGG did not add value per their accountability report.
YGG was not bound to any legal obligation. Add to this that the MC token was up around 30x from YGG’s initial investment price—MC was unwilling to fairly compensate YGG for taking an early bet on MC. Thankfully, no one pursued the path of legal action recognizing the danger of setting a premature precedent. MCL stepped in to mediate between YGG and MC, requesting to add the possibility of a counterproposal from YGG and delay the original proposal. The MC community accepted this request. YGG and MCL renegotiated and produced a counterproposal where MC would buy out all of YGG’s share of tokens for a 10x return. Both parties formally agreed to this compromise and terminated the relationship.
While the above examples are not exhaustive, they illustrate two key points.
One, onchain voting made what was before inconceivable on the Internet possible. Now, any stakeholder can vote on cases where they could not before. Voting supercharges their smaller voices with direct control over key decisions. The community is taking back control of the Internet.
Two, the fact that anyone in the community can vote gave us a new set of concerns and possibilities.
“On-chain governance’s underlying assumption is that because token holders and ‘owners’ of a protocol are one and the same, they are economically incentivized to vote in the protocol’s best interest. If tokenholders vote for a proposal that has a negative effect on the protocol, the token’s price will reflect that decision and they will lose money.” — Roy Learner
Onchain voting, similar to its offchain counterpart, proved just as vulnerable to authority, politics, and mob rule. The biggest problem, though, was the lack of engagement. Decision making felt bureaucratic for most and frustrated individuals vented that the community was unable to reach quick decisions.
It seems reasonable to assume that once given the right to influence the direction of a platform where you’ve judged it important enough to opt-in and become a stakeholder, that you would actively participate in organizational decision making. This couldn’t be further from the truth. Though having the ability to vote is better than no vote at all and higher participation is better than zero participation, having given everyone the right to vote, everyone decides is not what’s happening on the ground. Voter turnouts tend to be far lower than political and corporate votes.
We learned an important lesson: the percentage of actual voters isn’t as important as the fact that everyone can vote. Our expectation shouldn’t start with the fact that everyone will vote given a timeframe for a proposal. Our expectation should be that while every stakeholder has the right to vote, not everyone who has the right will vote. This is because:
There’s more to engagement, participation, and governance than the existence of voting. Voting participation is dependent on external factors such as time costly participation in deliberation. This is an example of what gives people a sense of ownership—leading them to vote.
It's natural for voter participation to fall as a DAO grows. Attention is THE scarce resource. Even with the best intentions, all DAOs face phenomena such as governance fatigue, difficulty meeting quorum requirements, and/or uneven distribution of information and knowledge as it grows.
All DAOs should avoid 0% voter participation. But the right amount of participation is dependent on the purpose of the DAO and the goal of the proposal.
PizzaDAO sells digital NFTs and uses proceeds to buy and give away pizzas in pizzerias around the globe. PizzaDAO votes using sesh, a Discord bot, and a 4-of-7 multisig to veto improper use of funds. The quorum per proposal is in the lower two digit range with around 20-40 people showing up for votes. PizzaDAO has about 800 NFT holders at the time of this writing. If that's the number of members, PizzaDAO has at max a 5% voter turnout. The number of votes does not obstruct PizzaDAO’s mission of giving pizza away for free.
Compare this with MakerDAO’s three contentious proposals LOVE-001, Makershire Hathaway, and MIP75c3-SP1 at the end of June 2022. These proposals aimed to change the operation of MakerDAO. MakerDAO would grant a group authority and a sizable budget to make autonomous decisions.
It marked record highs in governance participation as this vote would have an outsized effect on the purpose of MakerDAO regardless of results.
While the possibilities enabled by voting are exciting, there’s ambiguity around what it means to improve voting. Voter participation is a quantitative way to improve voting for everyone. What are qualitative ways to improve voting for stakeholders?
A New Myth: For effective DAO voting, stakeholders should a) shape high-level decisions, b) shard votes, c) and batch proposals.
We offer three practical heuristics to make voting more effective for everyone:
Previously, we’ve stated that it is fundamental to a DAO that “every member can directly influence the execution of key organizational decisions.” While there are plenty of decisions, in DAOs key organizational decisions refer to high-level decisions.
High-level decisions include those about the allocation of decision rights, an incentive system, and a system of information collection and flow. High-level decisions are strategic, typically occurring at times when a decision leads to actions that are difficult to reverse. Examples are investment decisions for investment DAOs like Metacartel Ventures DAO, fee-flipping decisions for DeFi DAOs like Uniswap, establishing new or breaking existing partnerships decisions for DAOs of DAOs like Tribe DAO, and raising the threshold at which members are eligible to create proposals like in Lido DAO. It goes without saying that sharding votes and batching proposals are high-level decisions as they address who should vote and when to vote respectively.
Low-level decisions are tactical, often encompassing most day-to-day operational decisions. Examples are choosing merchandise delivery providers, party venues, and community call time.
A DAO focuses on high-level over low-level decisions (Siggelkow, 2009). Don’t put every decision to an onchain vote. A DAO reserves onchain voting only for high-level decisions that have the greatest impact for the organization.
A DAO shards its votes by delegating decision-making power or creating smaller internal organizational structures such as workstreams, subDAOs, or pods. This organizes members according to space to decrease coordination costs.
Delegation systems for DAOs are more powerful than traditional models. In a traditional organization, the ability to make decisions doesn't necessarily translate to holding executive power. An individual DAO member is different in that she has several tools to keep direct control over key decisions. She can exercise her executive power by delegating to subject matter experts—individuals or organizations whom she believes know better—trusting that the delegates will execute on her behalf. Since undelegating and redelegating votes is easy, instantaneous, and directly accessible this keeps delegates accountable. She holds a nuclear option to exit i.e. “ragequit” the system faster than any traditional model and independent of any governance structure.
DAOs may go down another path of creating smaller internal organizations that function as committees. While everyone holds executive power, not everyone has similar levels of information, expertise, and available time. These internal organizations then make decisions on the DAO's behalf. Sharding sorts members according to their strengths and the whole organization benefits.
A DAO and its internal organizations can choose to vote less frequently by batching proposals together. The motivation behind batching proposals goes beyond saving gas fees to reducing coordination costs by organizing members according to time. A more regular voting cycle that models after best practices in representative democracies can help find a sweet spot between attention and action. The two guiding principles would be that a) a DAO should help its voting members pace themselves to make the best decision and b) a DAO should not nudge its voting members constantly unless the situation is urgent.
Contingencies which require swift action are unavoidable in any organization. The sensible way forward is not a shorter voting window. Rather, a DAO can preemptively establish an emergency organization with limited powers and veto rights e.g. Security Guardians to take appropriate measures or veto passed proposals during a grace period before execution.
Conviction voting would be an alternative implementation of batching that maximally reduces voting frequency. A member would not vote in a x vs y fashion but consider all proposals at once. Each proposal the community entertains is like a bucket; token-weighted opinions are like taps. A member pours preferences into selected buckets in whatever proportion she chooses. She continuously expresses her preferences “rather than casting votes in a single time-boxed session.” The longer she keeps her preference for the same proposal, conviction ‘builds’ and grants her consistent preferences with more weight than another who tries to influence a vote in the short term.
Putting these heuristics together, imagine a DAO that has three modes of voting: regular, irregular, and emergency. The DAO sends regular emails to its members. It summarizes upcoming high-level decisions and their ramifications. Delegates hold responsibility to follow up, inform, ruminate, and discuss among themselves over decisions that affect the DAO and/or their respective pods. During the DAO's voting cycle, the community expects the delegates to be well informed and consistent when they show up to vote on proposals outlined in the email. A small number of proposals, due to their nature, may have to be put to a vote outside the regular cycle. The expectation towards delegates for an irregular vote is that while they’re not given as much time to reach a decision, they show up and decide, understanding the tradeoff between precision and speed. When faced with a contingency, the emergency team in the DAO kicks into effect to protect the DAO until the situation is resolved. Members hold the power to redelegate their vote from those delegates who do not act in the interest of the DAO. Members can hold the delegates accountable for how meaningful their participation was across all three voting modes. In the extreme case that the DAO passes a controversial decision going against an individual member’s interest, the individual has the choice to “ragequit” and leave with her contributions.
We started by highlighting a persistent problem in many internet organizations today: a small group of stakeholders in an organization can take advantage of other stakeholders on a platform including developers, users, and investors. A mechanism to solve this problem is to enable every stakeholder to vote on key decisions on the platform. We then suggested ways of implementing voting that avoids and improves upon the current problem of vote-by-all. These ways range from emphasizing voting on high over low-level decisions, sharding, and batching votes.
A question remains: where and when will these mechanisms be more or less likely to be successful? Answer: Internet native organizations that value decisions by collectives rather than individuals. The early part of the internet was about individuals on platforms. The later part of the internet will be about creating collectives, organizations, and societies symbiotic with platforms living and breathing on the internet. We’ll flush out what this means in future writings.
Siggelkow, N., & Rivkin, J. W. (2009). Hiding the evidence of valid theories: How coupled search processes obscure performance differences among organizations. Administrative Science Quarterly, 54(4), 602-634. ↩︎
Comments powered by Talkyard.