💡 In simple words: Most decisions do not need a vote at all: people just try things, undo mistakes and talk. But when a big question really needs a vote, there are many ways to count votes. Sometimes the biggest pile of hands wins. Sometimes you rank your favorites, like first, second and third place. This page explains the main ways of voting, with a small example, so you can see why the counting method can change who wins.

🎯 In 20 seconds (expert summary): None of these methods is meant for everyday decisions: formal voting is reserved for questions whose stakes, expressed by the number of people who raise them, call for it, after lighter tools (direct action, revert, discussion) have been tried. Condorcet voting in particular is considered relevant where a case is complex and there is a clear, shared perception that it must be put to a vote. Survey of the voting methods WikiDeal is studying: simple majority, approval voting, ranked methods (Borda, instant-runoff), Condorcet methods with the Schulze completion, and single transferable vote for multi-seat elections. A worked 9-voter example shows a Condorcet winner losing under plurality. Each method is assessed against known criteria (clone independence, monotonicity) and known limits (Condorcet paradox, Arrow's impossibility theorem, strategic voting). The mapping of methods to decision types is a first hypothesis, consistent with Wikimedia and Debian practice.

Voting methods under study

Status: first hypothesis (draft). WikiDeal has not adopted any voting method yet. This page documents the options being explored. When to vote at all is treated on the portal main page: most decisions are intended to be settled by lighter tools (direct action, revert, discussion), and a formal vote is a last resort. Rules of procedure are on the Voting rules page; evidence and sources are on the research page.

Simple majority

The most familiar method: an option wins if more than half of the votes support it (majority rule). It is simple, fast and easy to verify, which makes it a good candidate for binary questions (adopt or reject a proposal, confirm or not a mandate).

Its main limit appears as soon as there are more than two options: with three or more candidates, the option with the most votes (plurality) can win while a majority of voters actually prefer another option. The worked example below shows this.

Approval voting

In approval voting, each voter marks all the options they find acceptable, and the option approved by the most voters wins. It is barely more complex than simple majority, works with any number of options, and tends to favor broadly acceptable options. Its limits: it does not capture the strength of preferences, and voters face a strategic question about where to draw their approval line (tactical voting).

Ranked voting

In ranked voting, voters order the options: first choice, second choice, and so on. A ranking carries much more information than a single cross, which matters when the decision is complex. Several counting rules exist for the same ranked ballots, and they can produce different winners:

The Borda count gives points by rank position. It is simple but very sensitive to the addition of similar options.
Instant-runoff voting eliminates the weakest option round by round. It is widely used but fails some fairness criteria, including monotonicity (ranking an option higher can, in rare cases, make it lose).

Condorcet methods and the Schulze method

A Condorcet method compares every pair of options: if one option beats every other option in head-to-head comparisons, it is the Condorcet winner. The idea goes back to the Marquis de Condorcet in the 18th century.

A worked example

Nine voters rank three options A, B and C:

Voters	Ranking
4 voters	A, then B, then C
3 voters	B, then C, then A
2 voters	C, then B, then A

Counting only first choices, A wins with 4 votes out of 9. But the head-to-head comparisons tell another story:

B against A: 5 voters prefer B (3 + 2), 4 prefer A. B beats A.
B against C: 7 voters prefer B (4 + 3), 2 prefer C. B beats C.
C against A: 5 voters prefer C (3 + 2), 4 prefer A. C beats A.

B beats every other option one-on-one: B is the Condorcet winner, even though A had the most first choices. A majority of voters (5 out of 9) ranked A last, and a Condorcet count respects that.

The Condorcet paradox and the Schulze completion

Sometimes there is no Condorcet winner: A can beat B, B beat C, and C beat A, like rock-paper-scissors. This is the Condorcet paradox, and it can be observed in real ballot data: the published pairwise table of the 2008 Wikimedia Board election shows such a circle among three mid-ranked candidates (the exact figures are on the research page and the history and results page).

The Schulze method resolves these cycles by comparing chains of victories (beatpaths): a chain is only as strong as its weakest link, and the option with the strongest chains wins. The Schulze method always elects the Condorcet winner when one exists, satisfies monotonicity and independence of clones (adding a similar option does not change the outcome), and is used by free software communities such as Debian. This combination of properties is why the initial hypothesis considers it relevant where a decision is complex and multi-option, and where the need to vote at all is clearly and widely perceived. Even then, a Schulze ballot is intended as a complement to the simpler tools, not a replacement: a contribution that can simply be reverted needs no vote, and a disagreement that discussion can settle needs none either.

Single transferable vote

When several seats have to be filled at once (a committee, a board), electing each seat separately can leave large minorities without any representation. The single transferable vote (STV) uses ranked ballots to reach proportional representation: options that pass a quota are elected and their surplus votes are transferred to the next preferences. Several counting variants exist; the Wikimedia Foundation adopted Meek's variant for its Board elections in 2021.

What theory says: there is no perfect method

Social choice theory proves that the search for a flawless voting method is hopeless in the strict sense:

Arrow's impossibility theorem (Kenneth Arrow, 1951) shows that no ranked voting method can satisfy a small set of natural fairness conditions at the same time.
The Gibbard-Satterthwaite theorem shows that every reasonable voting method can, in some situations, be manipulated by strategic voting.

The practical conclusion is not that voting is pointless, but that methods must be chosen for the properties that matter most in each context, and that the choice of method is itself a governance decision. This is the basis of the principle stated on the portal main page: match the method to the complexity of the decision.

Which method for which decision

First hypothesis, to be validated. This table applies only once a decision actually needs a formal vote; most decisions are intended to be settled earlier by direct action, revert and discussion (see the ladder of decision tools):

Decision	Envisaged method	Why
Binary questions	Simple majority	Simple, fast, verifiable
Complex multi-option choices	Schulze (Condorcet)	Respects head-to-head majorities, clone-independent, monotonic
Multi-seat elections	Single transferable vote	Proportional representation of the community
Everyday editorial decisions	Consensus, votes as last resort	Keeps discussion and improvement first
Structural changes	Majority with supermajority threshold	Stability of foundations