Quorum

What a quorum counts, and in what unit

Compound states its rule in tokens rather than percentages: at least 4,000,000 votes have to be cast in favour for a Governor Bravo proposal to succeed, with more for than against.¹ That is a fixed token count, not a share of anything.

A fixed count has one property worth thinking about before you copy it. It does not move when everything around it moves. Supply grows, delegation concentrates, a treasury unlocks, an exchange starts voting custodied tokens, and the parameter sits exactly where it was. Its real strictness drifts continuously and silently, because nothing in the system flags a threshold that has become easy. A percentage moves with its denominator instead, which trades that problem for a different one and makes the choice of denominator the actual decision.

The denominator is the decision

There are three denominators in common use and they are not close to each other. Total supply includes locked allocations, unclaimed airdrops and tokens sitting in contracts that will never vote, so it is the hardest bar to clear. Circulating supply removes some of that and is still optimistic. Delegated supply, the tokens whose voting power has actually been assigned to an address that could vote, is the closest thing to what is reachable on a given day.

Whichever you choose, the rule is that it has to match the denominator you measure turnout in. Teams routinely set a quorum against total supply and then read turnout reports computed against delegated supply, which makes the two numbers non comparable and the comparison meaningless. That mismatch is the single most common defect we find in a governance parameter review, and it costs nothing to fix at design time.

Worked: what the measured turnout figures do to a ten percent floor

The only rigorously dated, method stated turnout figures we could confirm come from Falk, Pathan, Rigas and Tsoukalas's 2024 study, which measured onchain voting across 130 Aave proposals and 141 Compound proposals and reported the average share of total token supply cast at 3.2 percent for Aave and 7.7 percent for Compound.²

Now take a quorum set at ten percent of total supply. Aave's average proposal reaches under a third of that floor. Compound's reaches about three quarters of it. Both proposals fail, having won every vote cast. Both of those protocols are large, liquid and heavily delegated. If they cannot average ten percent of total supply, a new DAO adopting ten percent because a guide suggested it has set a floor its community has never cleared and has no evidence it can.

One caution on where that ten percent came from. It sits at the bottom of a governance tooling vendor's recommended bands by DAO maturity, which run 3 to 5 percent for new DAOs, 5 to 10 percent for growing ones and 10 to 20 percent or higher for mature or high stakes ones.³ Those are design recommendations. They are not measured outcomes, the denominator is not stated, and they circulate widely enough that people quote them as data. Treat them as a starting hypothesis to test against your own votes.

And say the harder thing out loud: comparable measured participation data across DAOs does not exist. Two protocols, one paper, one denominator is the whole of the confirmed evidence. Anyone quoting a general figure is quoting folklore.

The two failure modes, and why the loud one is not the expensive one

A quorum set too low is the failure everyone designs against. A small coalition ratifies a change while the rest of the supply is asleep, and the floor that was supposed to prevent exactly that turns out to be lower than what one holder can produce alone.

A quorum set too high fails more quietly and costs more. Proposals win their vote and lose their count. After that happens three or four times, people stop drafting proposals, and the decisions do not disappear: they migrate back to whatever multisig can still act. That is the real price. A quorum that is too high does not make a protocol safer. It moves decision making to a venue with no delay, no public debate and no record, and the governance page still says the token holders are in charge.

Dynamic quorum, and what it inherits

The alternative to guessing is a floor that adapts. The Nouns pattern computes quorum as a minimum quorum plus a turnout coefficient multiplied by recent turnout, so the bar rises when participation rises and relaxes when it does not.⁴ It resolves the calibration problem honestly, because it stops asking anyone to predict turnout in advance.

It also inherits two things. The denominator question does not go away, it just gets applied to a moving number. And a threshold derived from recent history can be suppressed by an attacker willing to wait: a stretch of quiet, low turnout votes drags the floor down, and the expensive proposal arrives after it. In our view that is an acceptable trade for most protocols, provided the minimum quorum term is set high enough to survive a deliberately quiet quarter.

What we set quorum against

Four rules. Calibrate against delegated supply, because it is the only denominator that describes tokens capable of voting. Instrument turnout from the first proposal onward, so that by proposal twenty you are tuning against evidence rather than a vendor band. Keep quorum adjustable by governance itself, with a review on a fixed schedule rather than a review after the first failure. And keep the quorum separate from the approval threshold, because conflating how many voted with how many agreed produces a parameter that cannot be reasoned about.

The reason any of this matters commercially is that a governance system nobody uses is not a neutral outcome. It is a signal to holders that the votes attached to their tokens do not reach the decisions, which is the same conclusion they would draw from a protocol with no governance at all, arrived at more slowly and with more disappointment along the way.