Matrix (grid) questions

A matrix question asks the same scale of several statements at once. Instead of six separate “rate your agreement” questions, one grid does it in a single view.

It is the right tool whenever a set of items shares one scale — a Likert battery is the classic case. It is the wrong tool when the items do not share a scale, or when the grid would run too long for a phone screen.

Building one

1. Open your survey, go to Design, and select the Elements tab in the left sidebar.
2. Drag Matrix onto the canvas, or double-click it to add it to the end of the current page. Then click the edit (pencil) button on the element to open its settings.
3. Add your rows — the individual statements you want rated, for example “The support agent was knowledgeable”.
4. Choose your scale. You can use one of ten built-in standard scales — Agreement, Frequency, Importance, Likelihood, Quality, Performance, Satisfaction, Ease of Use, Helpfulness, and Recommendation — at 3, 4, 5, 6, 7, or 10 points. The default is 5. You can also write your own scale labels instead.
5. Choose the Choice Type, which decides how respondents answer each row:
   • Allow one answer — one selection per row. The standard Likert setup.
   • Allow multiple answers — respondents can select more than one option in a row.
   • Text entry — a free-text box per row, for short written answers instead of a scale.
   • Dropdown — a select list instead of visible buttons, which saves horizontal space.
   • Rating — a rating control in each cell.
   • Bipolar scale — opposing labels at each end, for semantic-differential style questions.
6. Set Show to control which scale labels appear: All, or First/Last to label only the two endpoints on a long scale.
7. Adjust the layout. Swap rows and columns transposes the grid, Reverse order flips the scale direction, and Extra space widens the gaps between columns.
8. Preview at both desktop and mobile widths. Grids are the question type most likely to need adjusting for small screens.

Repeating the headers on a long grid

Repeat headers shows the column headers again further down the grid, so respondents working through a long list do not lose track of which column means what.

It is not a sticky header that follows the screen — it inserts the headers again at a position you choose:

• Middle — repeat at the middle row
• Bottom — repeat at the bottom row
• Both — repeat at the middle and the bottom
• Each — repeat above every row

The setting panel notes that this gives minimal benefit on matrices of three rows or fewer, which is worth heeding: on a short grid the original header is still on screen.

Example

A market researcher running a brand-perception study wants to rate six brand attributes — “trustworthy”, “innovative”, “good value” and three others — all on the same 5-point Agreement scale, expecting most respondents to be on mobile.

They add the six attributes as rows, select the standard Agreement scale at 5 points, set Choice Type to Allow one answer, and turn on Repeat headers so people scrolling six rows on a phone do not lose track of which column means “strongly agree”.

Troubleshooting

The grid looks squeezed or cut off on mobile. Turn on Swap rows and columns, or reduce the number of scale points. A 7- or 10-point scale next to long row labels rarely fits comfortably on a phone.

Respondents lose track of which column is which on a long grid. Turn on Repeat headers and pick a position — Middle, Bottom, Both, or Each.

Only some of my scale labels are showing. Check the Show setting. First/Last is a deliberate space-saving option that labels only the two endpoints. It is not a bug, but it is easy to set by accident.

I want respondents to pick more than one option per row. That is a Choice Type setting, not a different question type — switch from Allow one answer to Allow multiple answers rather than rebuilding the question.

This wave’s results don’t line up with last quarter’s. Avoid changing scale length between waves of a tracking survey. A 5-point answer and a 7-point answer are not directly comparable, and switching breaks comparability with everything measured before it.