Demographic cohorts and minimum thresholds

Demographic questions let you slice results by dimensions such as department or location. Scores and choices come back as totals; written answers are shown word for word, shuffled and undated; InviziPoll never shows a list of who answered, per-response timestamps or submission order. Where the poll allows it, anyone who'd rather not be read can send just the topics they pick instead.

#What a cohort is

When a poll asks optional demographic questions, each combination of answers (for example "Engineering × EMEA") is a cohort. Admins can filter results by those dimensions after they're decrypted on their device, once the poll has closed. Demographic answers travel inside the encrypted response; they aren't a separate identifiable roster.

Cohorts wait for close because comparing a group before and after new answers arrive could reveal one person's answers. While the poll is collecting, the results tab says when breakdowns unlock, with the close date. Closing early counts. Polls that never close, the Exit pool and the speak-up channel, never show cohorts. How batched release works.

At close, all cohorts are checked together, once. Besides the floor on each one, a cohort that could be worked out by subtracting others from the whole stays locked, with the same message as a small one and no reason given.

#Self-reported demographics versus roster attributes

The dimensions on this page are self-reported: a respondent answers a question, and their answer is what you can filter by. That's a claim the respondent made about themselves, and people pick the department they feel part of rather than the one on the org chart.

Roster attributes are the other route. Department, location, tenure band and manager are imported from a contact list CSV, collapsed at send time so no combination points at fewer than five people, and attached to each ballot on email magic-link distribution. Your browser filters by them behind the same floor and the same wait for close, and they're labelled distinctly, "Department (roster)", so a roster cut is never mistaken for a self-reported one. The two sets never merge, because they're different claims that disagree in practice. Roster attributes cost no questions, and they also feed the fixed cut catalog inside a published results document. See Roster attributes and cuts.

Use self-reported dimensions when there's no roster, when the poll goes out by access code or open link, or when how people see themselves is the thing you're measuring.

#Prerequisites

• A published poll with demographic dimensions you intend to report on
• A closed poll: cohort charts don't open while it's collecting
• At least five answers in the filtered group, or the poll's minimum group size if higher, before a cohort chart opens

1. Add demographics carefully. In the poll wizard, tag only dimensions you will report. Prefer one or two high-value fields.
2. Read the Anonymity Coach. Multiple dimensions multiply combinations. Many slices may stay locked; that is expected.
3. Collect responses. Share the poll and wait. Early responders do not appear as a named list.
4. Filter after close. When the poll closes, apply a demographic filter. You may see "Checking breakdowns..." for a moment first. Under the floor, the slice stays locked with no count.
5. Adjust next time. If too many slices lock, reduce dimensions or widen the audience.

#Sharing one cohort with its manager

The filter you apply here stays in your browser. To let a manager see their own team's numbers without giving them the poll, use Share a slice on the results tab. After close, it takes the group you have filtered to, refuses below the same floor, and produces a read-only link to a fixed, aggregate-only snapshot with an expiry you choose. The full mechanism, including why it is a snapshot rather than a key, is in Share a slice.

Open-text answers narrow with the filter too: below the floor the written answers lock along with the charts.

#What you will not see

Admins do not get identifiable respondent lists or metadata that reveals who answered first or last. Small combinations return a locked state rather than a thin chart. Slices that could be worked out by subtraction stay locked too, but that check only covers this poll, so keep dimensions coarse when a group would be small. Aggregate results and anti-inference covers the limits.

#FAQ

#Related

• Viewing results
• Roster attributes and cuts
• Aggregate results and anti-inference controls
• Batched release and the exit pool