Method
How this is measured, and what it cannot tell you.
Published before the first figure, so that no rule here can be chosen after seeing a number it would flatter.
The panel
Panellists register with an email address and a mobile number, both verified. Each person records at most one vote per election per day, and the answer locks on save. A missed day is recorded as a gap; we never carry an answer forward.
Weighting
An app that asks people to vote on politics every day recruits the politically engaged and the well educated. That is not a small skew, and it is the failure mode the 2019 Australian polling inquiry identified.
Our figures are not weighted. We do not yet collect the variables a weighting scheme would need, such as education and how people voted last time, so every published figure is the panel as it stands. Read it as a measure of how this panel moves, not as an estimate of the electorate. If we add weighting, we will publish weighted and unweighted figures side by side, and the scheme here first.
Location
When someone records a vote, their browser may ask to share their location. It never blocks anyone from voting. If they allow it, the reading is checked in memory and the coordinates are discarded; all we keep is whether it fell in Australia. We do not keep a state or region, and we do not use location to filter or report results.
If we do start, readings will resolve to three states — in jurisdiction, out of jurisdiction, and unknown — with unknown reported as its own category, never folded into "out". Location is weak evidence either way: it is straightforward to spoof, so it can be a disclosure statistic and one signal among several, never proof of eligibility.
What we publish
Every reading appears twice — raw and matched — because the difference between them is the part no single-night poll can show you. A party with fewer responses than our suppression floor is not named: it is pooled into "All others", with more parties pooled if needed so that nothing small can be recovered by subtraction, since a small enough group describes an identifiable person. The same applies between the two readings: if fewer than that many people voted on the day without having voted a week earlier, subtracting one reading from the other would describe them, so neither is published. If either reading cannot be published safely, neither is.
Privacy, stated accurately
No one can see how an individual panellist voted in anything we publish. That is a different claim from "we do not know", and we will not make the second one: the daily record has to be linked to a person or none of the matched-panel analysis on this site would be possible.
Still open
Three questions are not yet decided: whether results will be filtered by location, how the unknown-location group would be treated, and whether we will publish a locked prediction before election day. Nothing published depends on them, and each will be settled here before it affects a figure.
Corrections
6 October 2026. Earlier versions of this page, our home page and the summary we publish for search and AI tools said that our figures were weighted, that we collected education and past-vote recall, and that location was recorded at registration and resolved to a region. None of that was true. The figures are unweighted, those questions are not asked, and only whether a reading fell in Australia is kept. This page now describes what the site actually does. Some posts on our blog made the same claims; where they differ from this page, this page is correct.
Published by Synap AI Pty Ltd (ABN 86 695 003 653).