Investment DAO Votes and How to Sway Them

Most investment DAO votes are lost to a threshold, not to an opponent. Aave counts no result unless 320,000 AAVE, stkAAVE or aAAVE takes part. The episode that produced its rules on vote timing began when two uncontested votes failed in a single weekend for lack of turnout.

This list covers seven levers that move outcomes in token-weighted DeFi governance, drawn from protocol documentation and governance portals. They are ordered from the cheapest to use to the most expensive. The conclusion follows from that order. For anyone without a large balance, the useful levers sit near the top, and the influence they offer comes from attention and timing rather than from tokens.

Aave's Governance Framework v2, published by TokenLogic, sends a proposal down one of three routes:

• The Standard Process runs from an ARFC forum thread to a Snapshot vote to an on-chain AIP.
• Direct-to-AIP skips Snapshot. An existing asset expanding to another Aave instance can clear it in as little as five days.
• The Steward Process handles routine updates within limits governance has already set, and needs no full vote.

The framework also removed the mandatory TEMP CHECK stage. Live Bitcoin News reported that the standard cycle fell from 19 days to 13.

The practical result is that the forum thread is often the only place where a proposal's wording can still change. A Snapshot vote offers yes or no on text that is already settled.

Sentora's proposal to run a curated Aave V4 market shows what is open at that earlier stage. It was posted on September 28 to 29, 2026, and is still at ARFC. Its terms include a 50/50 revenue split, a 100 million RLUSD deposit cap and the exclusion of USDC and USDT as borrowable assets, and all of them are still open to comment.

Sky follows the same sequence. Forum discussion feeds a weekly poll. The poll result becomes a markdown write-up in a public GitHub repository, and that write-up becomes spell code.

An Aave proposal needs two things to pass. At least 320,000 tokens must participate, and the winning option must itself receive at least 320,000 votes.

After the weekend failures, the quorum-failure ARFC barred AIP votes from starting or ending on a weekend, except in emergencies. It named Monday and Tuesday as the standard start days.

Sky's weekly Atlas Edit polls have a similar test. They run for three days, and 'Yes' must both beat 'No' and reach a Minimum Positive Participation of 480,000,000 SKY.

For a holder, the useful question about an open vote is whether it is close to quorum as the deadline nears. Tokens that carry a vote over the threshold do more than tokens added to a vote that has already cleared it.

Snapshot votes are gasless, so they cost a voter roughly nothing. An on-chain Governor vote on Ethereum mainnet can cost several dollars per voter during congestion, according to Startupik. In Aave's Standard Process, Snapshot records sentiment and the AIP is what executes.

Sky's binding layer works differently. It uses continuous approval voting: SKY locked in the chief contract backs an executive spell, and a new spell becomes active once it holds more weighted support than the one before it. The September 10, 2026 spell passed on September 11 at 12:22 UTC with 6,663,579,099 SKY behind it, and was executed two days later.

The September 24 spell came with a 48-hour pause delay, an execution window of 14:00 to 21:00 UTC on weekdays, and a 30-day expiry. In Sky, a vote is a position that stays in place, not a ballot cast once. Timelocks give holders on the losing side a fixed period to act before the change takes effect.

Delegation lets holders pool voting weight without giving up custody of their tokens. ENS, Uniswap, Sky and Optimism all use it. Even so, DAO Times puts typical turnout in healthy DAOs at about 10%.

LedgerMind, a less rigorous source, gives figures for Optimism's Token House. Of 487,000 OP holders, 1,247 voted directly and 86,431 delegated, which gave delegates 73% of voting power. In ENS, 142 active delegates held voting power, and the top ten held 64% of it between them.

Delegates are also a target. LedgerMind reports that a $2.3 million bribe offered to ENS delegates in August 2025 failed after it was made public.

For a small holder, delegation is the main route from a balance too small to matter to how voting power is counted at the threshold. That makes a delegate's past voting record the thing to check before delegating to them.

Vote-escrow, the model Curve introduced, gives more voting weight to tokens locked for longer. veCRV locks run from one week to four years, and the weight decays linearly as the unlock date approaches. The cost of this lever is illiquidity.

Lock-ups also tend to produce blocs. In the 2024 to 2025 'Curve Wars', LedgerMind counted:

• Convex Finance: 51.2 million veCRV, or 51% of supply
• Yearn: 7.8 million
• StakeDAO: 4.3 million

Research cited alongside the ACM study links vote-locking models with TVL growth 5.6 to 7.2 percentage points higher than in DAOs without them.

The ACM study found a median Gini coefficient of 0.98 for token-based protocol DAOs, compared with 0.75 for DAOhaus DAOs and 0.46 for DAOstack DAOs. A score of 1.0 would mean one holder has everything.

The scale of the votes shows what that means for a buyer. Sky's September 21, 2026 poll passed with 7,055,924,296 SKY voting yes. A hypothetical holder of 1,000,000 SKY would have made up about 0.014% of that vote, or about 0.21% of the 480 million threshold. Turnout of that size keeps concentration at the top intact.

Buying votes for a short time is what timelocks are built to stop. In 2022, an attacker borrowed $1 billion through a flash loan, took 79% of Beanstalk's governance tokens and drained $182 million in 13 seconds.

The strongest objection to this list is that concentration makes its earlier items irrelevant. LedgerMind counted the top 100 MakerDAO addresses holding 89% of voting power in March 2026, and at that level a median holder's attention seems to count for little. For contested votes, that is largely true. It does not apply to the votes that actually failed on Aave, which were uncontested and fell short on turnout. It also does not apply to forum threads, where comments are not weighted by tokens.

Quadratic voting prices each vote at the square of the number cast: two votes cost four tokens, and ten votes cost a hundred. DAO Times reports that its adoption grew about 30% from 2024 to 2025.

The weak point is identity. Startupik puts it bluntly: "Wallet splitting is trivial — QV collapses back toward linear power" unless voters face identity costs such as deposits, attestations or proof of personhood.

For social and product-community DAOs, Startupik recommends token or reputation voting on gasless Snapshot, with a Safe council able to pause. Among the common mistakes it lists are a Snapshot vote with no execution path behind it, and quorum failure caused by voters never being taught how delegation works.

The order of the list is also a guide to where a small holder's effort counts. Items one and two cost nothing and decide more outcomes than their price suggests, because Aave's documented failures were failures of turnout, not of opposition. For anyone without a large balance, the time to sway an investment DAO vote is the week before it opens and the hours before it closes. In between, the outcome mostly follows the tokens.