Matrix and grid questions
How a grid works in InviziPoll: rows stay separate questions, arrive with an import, render as one table, and still export and filter per row.
On this page
A grid is a way of showing several questions, not a question type of its own. Each row is a complete question underneath, which is what keeps exports, cohorts, branching and wave-over-wave trends working per row instead of per grid.
A matrix, or grid, asks several statements against one shared rating: "How much do you agree with each statement?" with a row per statement and the same columns across all of them. It is the compact way to ask five related things without five separate cards.
Where a grid comes from
Grids arrive with an import. A rating matrix in a Qualtrics or SurveyMonkey export becomes a grid: one question per row, all carrying a shared group so the builder, the respondent page and the results view put them back together as the table you authored in the other tool.
That is a change from earlier behaviour, where the same export arrived as a run of near-identical questions and read like an importer artefact. See importing questions from another tool for what else changes on the way in.
There is no control in the poll wizard for building a grid from scratch yet. If you need one today, the route is an import.
What the rows are
A row is a scale or a single-select question, and every row in one grid is the same kind with the same columns. Each row keeps three things: its own full question text, the short row label shown in the table, and the group it belongs to.
The full text matters more than it looks. The CSV, the snapshot and a wave trend all read the row's own stem, so a row that later leaves the grid is still a complete, sensible question rather than an orphaned fragment reading "Deadlines".
Editing one
Each row appears as an ordinary question card with a note reading Part of the grid, followed by the grid's title. You can edit the row label there, and you can edit the question text, the options and the scale ends the same way you would for any other question.
Rows of a grid must stay next to each other. Move one away from its neighbours and it stops being part of the grid and becomes an ordinary standalone question, keeping its own wording. Nothing is lost when that happens, but it is a one-way move: putting it back beside the others does not re-form the group.
What a respondent sees
On a normal screen, one table. The statements are row headers down the left, the rating runs across the top as columns, and each row is a set of radio buttons sharing a group, so a row is one tab stop with arrow keys between the options. The row and column headings are real table headers, which is what lets a screen reader announce "Deadlines: Agree" rather than an unlabelled button in a sea of them.
On a phone the same rows are stacked, with each row's options named beside the control, because a five-column grid on a narrow screen is unusable rather than compact.
How the results read
The results view draws the grid as a grid: one panel, one row per statement, compared against each other on the shared scale. That comparison is the thing a grid says that its rows do not each say alone.
Everything else still works per row, because a row is still a question:
| Surface | What it sees |
|---|---|
| Aggregate CSV | one row per member question |
| PDF report | the grid panel and its questions |
| Shared results slice | member questions and aggregates |
| Cohort filters | each row, like any other question |
| Wave-over-wave trends | each row's movement across waves |
An on-device poll digest describes a grid as one result rather than as a run of separate ones, for the same reason the results view draws it as one.
AI and grids
The AI poll builder does not create grids, and an Improve pass on a row cannot dissolve one. Improve rewrites the wording of a single question, and the grid it belonged to is put back afterward, so a rewrite cannot quietly turn a five-row table into five loose questions. Layout is yours, not the model's. See AI poll building.