You're reading a governance proposal. The vote closes in six hours, quorum is 8% of circulating supply, and the treasury transfer being approved would move $40 million to a multisig controlled by three addresses. You check the top holders. Two of them are voting yes. You do the arithmetic. It's already over.

That scenario isn't exotic. It's what minimum viable decentralization, or the lack of it, actually looks like at ground level.

The geometry of capture

Capture doesn't require a villain. It only requires concentration. When governance power pools tightly enough, rational self-interest does the rest, no conspiracy needed.

The standard model governance researchers converge on is the 51% attack threshold, borrowed from blockchain consensus. For DAOs, that number misleads in a specific and important way. A DAO proposal doesn't need 51% of all tokens to pass. It needs 51% of tokens actually cast in a vote. If turnout is 8%, then 4.1% of total supply controls the outcome. That's not a hypothetical. Several major DeFi protocols have seen quorum thresholds so low that a coordinated bloc of under five addresses carried consequential treasury votes.

The minimum viable decentralization threshold, then, is not a fixed number. It's a ratio: the concentration of the top N holders divided by realistic participation rates. When that ratio tips above 0.5, the DAO is capturable by a small coalition. Call it the Capture Ratio. You want it well below 0.5, and you want it stable across market conditions, not just during calm periods when everyone's paying attention.

Consider two hypothetical protocols, both with 100 million governance tokens in circulation. Protocol A: top ten wallets hold 38% of supply, and average turnout on contested votes runs around 18%. Those ten wallets can credibly muster maybe 25% of votes cast, assuming some abstain. Influential, but not dominant. Protocol B carries identical top-ten concentration. Turnout collapses to 6% during a contentious upgrade vote. Now those same ten wallets represent over 60% of participating votes. Protocol B just became capturable, and nothing changed except apathy.

Apathy is the real attack vector. Governance token designs consistently underestimate this.

Participation is not a constant. It's a function of token price, market conditions, proposal complexity, and how exhausted holders are from voting on routine housekeeping. A DAO that looks decentralized in a bull market, when holders are engaged and paying attention, can look entirely different during a quiet stretch when only motivated insiders bother to show up. The governance structure that reads as healthy in one season is, under identical rules, a rubber stamp in another.

What actually moves the threshold

Four factors determine whether a DAO sits above or below the capture line.

Token distribution at genesis. This is close to destiny. If a founding team and their investors receive upward of 40% of supply at launch, with vesting schedules that align their lockups to the DAO's formative governance period, the distribution math will fight you for years. Uniswap's initial UNI allocation gave 40% to team and investors with four-year vesting. Compound's COMP distribution leaned heavily toward early protocol users and liquidity miners. Neither is inherently corrupt, but both created structural concentration that took years of secondary-market diffusion to dilute. You can't vote your way out of a bad genesis.

Delegation mechanics. Liquid delegation, where passive holders assign their voting power to active delegates (think of it as a governance power of attorney, revocable at will), is a singularly practical lever available to DAO designers. Compound pioneered it. When it works, it converts apathetic token holders into meaningful participants by proxy. When it fails, it creates a class of professional delegates who accumulate power quietly and become the new concentrated bloc. The design detail that matters: whether delegation is revocable instantly, and whether delegate voting records are legibly public. Opacity in delegation is almost as bad as no delegation at all.

Quorum floors with teeth. A quorum requirement demanding 10% of circulating supply to validate a vote sets a real floor on capture. Below that floor, the vote simply fails, which is preferable to a low-turnout result being executed as legitimate consensus. The catch is calibration. Set quorum too high and the DAO gridlocks on everything. Set it too low and you've built a theater prop.

Time delays and veto windows. Compound's 48-hour timelock, and similar mechanisms in Aave and Maker, give token holders a window to mobilize against a proposal that passed on thin turnout. It's a circuit breaker, not a solution. But it's an honest acknowledgment that capture can happen fast and reversal needs breathing room.

Here's a practical benchmark worth internalizing: if the top-ten holder concentration sits below 30%, turnout on contested votes consistently clears 15%, delegation is active and transparent, and a timelock exists, you're looking at something with genuine structural resistance. All four, not just one. A timelock on a captured vote is just a polite countdown.

The uncomfortable truth that governance forums rarely say plainly: a large share of DAOs are not at minimum viable decentralization. They're at minimum viable appearance of decentralization. Governance theater is a real genre, and it is considerably more widespread, a dominant genre, really, alongside actual governance, which remains a minority practice.

The threshold isn't a line you cross once at launch and celebrate. It's a ratio you have to maintain actively, across market cycles, across team turnover, across the slow gravitational pull of tokens back toward whoever has the greatest time and capital to accumulate them. Decentralization decays like an untended network, quietly and then all at once. Governance is maintenance work, not a founding-day achievement.