Anti-inference: thresholds and cohorts
Heads up — This page is the deep dive for 3+ (poll) and 5+ (cohort) rules. For everyday copy, start with Viewing results.
Anti-inference: thresholds and what admins see
InviziPoll blocks small-sample disclosure in two layers: poll-wide results and demographic cohort slices.
Poll-wide unlock threshold
Until there are at least three responses, aggregate results remain locked. The admin experience shows a fixed locked message and does not show aggregate charts or per-option breakdowns for that poll—avoiding premature small-group disclosure.
In the admin results view, locked state shows guidance such as "Results unlock after 3+ responses."
Cohort filtering
After decryption, the browser can filter aggregates by demographic dimensions. If fewer than five decrypted responses match the filter, the view stays locked for that slice — no exact counts or percentages for the filtered cohort.
The admin UI shows: "Insufficient responses for this cohort. No breakdown is shown."
Example: protecting a small team member
Suppose five people answer a survey, including one person in the Legal department. Overall results may unlock (n ≥ 3). But filtering by "Department = Legal" yields only one matching response — well below the cohort threshold of 5 — so the cohort breakdown stays locked with no per-cohort count revealed. Combining filters (e.g., Legal + Chicago) usually reduces the count further.
This means an admin cannot use demographic filters to isolate an individual respondent's answers.