How it works

How2Vote compares your views with what parties have actually done in federal parliament — the votes their members cast, not what they said in a campaign. The method is a fixed set of arithmetic steps, published in full below with nothing withheld: the same answers always produce the same result, and every step below is exactly what the scoring engine computes. Which issues to include and how to describe a division still involve editorial judgement, so this is one reasonable reading of the public record, not the only one.

In short

Find your electorate. Choose your state and federal electorate so the card is built for the candidates on your House and Senate ballot.
Answer the propositions. Answer real propositions parliament has voted on, on a five-point scale, and star the ones that matter most to you.
Compare, then build your plan. See how your answers align with each party's recorded votes, shown in official ballot order — then choose your own preferences to build your voting plan.

Where the data comes from

Every division Division A formal, recorded vote in parliament, where members are counted for and against. They Vote For You publishes these votes, and How2Vote's scores are built from them. Full glossary (formal vote) in the House and Senate is recorded in Hansard Hansard The official written record of what is said and done in parliament. Full glossary and published, in a machine-readable form, by the OpenAustralia Foundation's They Vote For You(opens in a new tab). For each issue, They Vote For You gives an agreement figure from 0 to 100 for every member — how often that member voted the way the issue describes. We use those figures under the Open Database Licence (ODbL), and we credit They Vote For You on every screen and every printed card.

From members to parties

A party's score reflects the historical parliamentary voting record of its members. It describes the party, not any individual candidate's personal views, and does not predict how a candidate would vote in future. For each issue we take the plain average of the agreement figures of that party's members (each member counted once), then place the party on a five-point scale by dividing 0–100 into five equal bands:

0–20 → 1, strongly disagree
20–40 → 2, disagree
40–60 → 3, equal merits
60–80 → 4, agree
80–100 → 5, strongly agree

Independents who have sat in parliament are scored as their own one-person “party”. A party or candidate with no member on the record for an issue simply has no position on it — it is never guessed, and it earns and risks nothing on that question.

Your answers

You answer each proposition on the same five-point scale (1 strongly disagree … 5 strongly agree), or skip it. When you review your answers at the end, you can star the issues that matter most to you as extremely important, which multiplies that one question's weight by exactly ten. Only your two strongest answers — strongly disagree and strongly agree — can be starred. Skipped questions count for nothing, for any party.

Scoring one question

For each question you answered, we compare your position with the party's using the plain gap between the two points on the scale — the distance d = |party − you|, a whole number from 0 (identical) to 4 (opposite ends). A party earns the most points when the gap is zero and fewer as it grows, down to none once the gap is wide enough. How much is on the table depends on how strongly you answered:

a strong answer (strongly disagree/agree) is worth up to 4 points: max(0, 4 − d);
a moderate answer (disagree/agree) is worth up to 3: max(0, 3 − d);
equal merits is worth up to 2: max(0, 2 − d).

Marking a strong answer extremely important multiplies both the points earned and the points at stake by ten — up to 40 for that question: max(0, 40 − 10d). The “points at stake” (the most a party could have earned) is tracked alongside the points it did earn, so a bigger question counts for proportionally more in the final total.

For example, if you answer strongly agree and mark it extremely important, a party recorded at agree (a gap of 1) earns 40 − 10 = 30 of a possible 40, while a party recorded at strongly disagree (a gap of 4) earns 0.

Your overall match

Across every question you answered, we add up the points each party earned and the points it could have earned. Parties that have merged or been renamed between parliaments are combined at this stage — we add their raw earned and possible points together first, then compute a single percentage, so a merged party is treated as one continuous entity rather than an average of two rounded scores. The match is then

match % = round( points earned ÷ points possible × 100 ).

A party you never had a scorable question in common with (it had no position on everything you answered) has no denominator, so it shows as no data rather than 0%.

Your comparison, then your plan

We compute a percentage for every party this way, then place those scores beside the candidates actually printed on your ballot — your House electorate and your Senate state — in official ballot order. This is shown as evidence only: nothing is ranked, nothing is crowned, and How2Vote never suggests who to put first. Every party's score links to the divisions behind it, so you can check any number.

When you're ready, you build your own voting plan from a blank ballot — you choose every preference number yourself, by typing it or using the move-up / move-down controls. The order is entirely yours; we only offer a mechanical check that flags boxes you haven't numbered yet. Copy your numbers onto the real ballot paper and follow the AEC's instructions.

For the Senate you can work above the line (the party groups; number at least six) or below the line (every individual candidate; number at least twelve), both in ballot order. Because our scoring is party-level, an above-the-line group takes the score of the party that heads it, and every candidate in a group shares that party's score. Number one way or the other — not both on the same paper.

The scoring engine is open source and covered by tests that pin these numbers exactly; any change to the method is versioned and shown here and in the footer.

What the method does and doesn't capture

The comparison rests on editorial and methodological choices, and it is one reasonable reading of the public record — not the only one:

Absences and pairs. A party's agreement figure comes from They Vote For You, whose figures already account for how members attended and voted; we do not add our own treatment of absences or paired votes, and we take each figure as published.
Which issues appear. The propositions are selected from They Vote For You policies to span the parliamentary term; choosing and wording them is an editorial judgement. Each proposition states a concrete policy action alongside the genuine competing interest it trades off against, so agreeing and disagreeing are equally reasonable rather than a foregone conclusion. Each proposition links to the divisions behind it so you can read the record yourself. The set of propositions and the reasons for choosing them are recorded as a versioned proposition selection v3, bound to the published dataset, so any change to which issues appear is dated and explained rather than silent.
Party-level records beside new candidates. A candidate with no personal voting record is shown against their party's historical record — that describes the party, not the candidate's own views, promises or future conduct.
Ties. Equal alignment figures need no tie-break: candidates are never ranked by score, only ever shown in official ballot order.

Provenance and changes

The scoring method described on this page is versioned. It is currently methodology 2026.1. This number is bumped only when the method itself changes — the banding, the distance and points formula, the weighting, or how party records are combined — not when the data is simply refreshed.

Each election's card records its data vintage (the latest division the positions were compiled from), the data and app versions, and the parliamentary voting source, so a saved or printed card carries its own provenance. The dataset also records the official source of its candidate ballots (AEC candidate nominations) and the data version they arrived in. Ballots are ingested through a manual, human-reviewed change process, and once an election's snapshot is locked, automated checks refuse any further change to it.

We keep a published correction and methodology-change history: every correction we make and every change to the scoring method, dated and versioned. Suspected errors — a wrong candidate, party mapping, ballot order or figure — can be reported through the contact page, and the channel is monitored during a campaign.

Aggregate insights

From the research contributions people choose to make after building a comparison, we publish aggregate figures on how views line up with these voting records. The analysis starts on the contributor's own device — their individual answers and weights never leave it; only the derived result travels — and what we store is group counts, not individual records. A figure appears only once a group is large enough that a single contribution cannot readily be singled out — see insights and the privacy page for how that works. On election day we don't publish this analysis while people are still voting: insights are closed from midnight until the last polls close nationally (8 pm AEST), and return after that.

People may also answer an optional survey about themselves. Its questions follow Australian Bureau of Statistics categories and Australian Election Study conventions, so the responses can be compared against Census and other public data; the exact wordings are recorded in our research codebook. We hold no name, contact detail, IP address or device identifier, and no per-person research record exists at all — though no technique eliminates every possible risk: while counts are small, a rare category could still say something about a person known to have contributed. That is why counts pairing results with sensitive answers are kept national-only, stored counts are kept apart from operational systems, and only privacy-protected aggregates are published — see the privacy page.