How we count

What we will never claim

We will never tell you that a percentage of a suburb supports something. People who respond to a community website chose to respond; people who did not are not represented by them. That is called a self-selected sample, and no amount of care turns one into a survey of the whole population.

Who gets counted

1. You register with an email address and a Wyndham suburb, and give express consent.
2. We email you a confirmation link. Until you use it, your response is invisible and uncounted. Not hidden pending review — not counted at all.
3. One response per member per project, enforced by the database rather than by good manners.
4. Only Wyndham suburbs appear in the list, so a response cannot come from outside the municipality.

Our thresholds

A suburb’s figures are published only once 5 or more people there have responded, and a suburb is only shaded on a map once 10 or more have. A colour on a map reads as a finding, so it should take more than a handful of people to produce one.

We chose five following the Australian Bureau of Statistics, which uses a minimum number of contributors before publishing a cell and notes openly that there is no strict statistical basis for preferring three, five or ten. Higher protects more and tells you less. We picked a middle and wrote it down here rather than leaving it implicit.

Suppression that can be undone with arithmetic is not suppression, so a threshold on its own is not enough. Before publishing anything we check whether any hidden suburb’s figure is forced by what we are about to show. If it is, we hide another suburb — including one that is above the threshold — and check again, until nothing is recoverable.

Two consequences worth stating, because they will look odd otherwise. A suburb can be hidden even though it has plenty of responses: it is covering for a smaller one. And if no amount of hiding is enough — which is normal when a campaign is new — we publish no figures at all rather than a total that gives one person away.

We also withhold the overall support, oppose and unsure split whenever any suburb is hidden. Published separately, those three numbers are three equations to subtract with, and three equations pin values that one does not.

What we could do better

An email address is free and infinite. Our controls stop automated abuse and casual duplication; they would not stop a determined person with ten addresses, and we are not going to pretend otherwise. Real one-person-one-vote needs a resident register, which governments have and community projects do not.

What we offer instead is auditability: counts rather than percentages, a stated method, and openly published grouped data. Treat our figures as evidence that people care enough to act, not as a measurement of the whole municipality.