Matrix Questions: Where They Shine and Where They Fall Apart

A matrix (grid) question puts several items that share one scale into a single table. Rows are items; columns are the scale points.

Where it's the best choice

• When several items share exactly the same scale.
• When you want the respondent to compare the items against each other, not consider them one by one.
• When the items are short (under 6–7 words).

A good example: "Rate your satisfaction with each of the following: workload / colleagues / tools / physical workspace."

Where it falls apart

On mobile. A 5-column table on a 375-pixel screen either scrolls horizontally or shrinks so much that the user taps the wrong column. If most of your respondents are on phones, the matrix has to turn into separate cards on mobile: each item as a block with its options stacked underneath.

When the items are long. A row that wraps to two lines makes the table unreadable.

When the items don't really share a scale. If you're asking about "satisfaction" for one item and "importance" for another, you need two separate matrices.

Limit the number of rows

Above 8 rows, straight-lining begins: the user ticks one column from top to bottom. If you have 15 items, two matrices of 7–8 with a heading in between are better than one of 15.

The "no opinion" option

In a matrix, forcing an answer for every row causes problems: someone who hasn't experienced one of the items is forced to tick something. Add a "haven't experienced this" column that is excluded from the average.

What you see in the report

A matrix report should give you two things: the average of each row (to rank the items) and the distribution of each row (to see polarization). An average of 3 can mean everyone gave a 3, or that half gave 1 and half gave 5, and those two situations are completely different and call for different decisions.