Question Design: Matrix Questions

A matrix asks several items on the same scale. There is no grid widget: you write the matrix as text, and Elaiia recognises it from the wording. At launch, each row becomes its own question with its own report, asked one at a time in the order you wrote them. So a matrix saves you writing time and counts as only one question towards your question limit — but every row is still a full question for the twins.

When to use a matrix

Use a matrix when two or more items share one framing and one answer format. Otherwise:

Items need different scales or answer types → separate questions
Items ranked once on one criterion → a ranking question
You also want the reasons behind the ratings → the matrix, followed by a separate open question

How to write one

A framing sentence, one statement per line starting with a dash, then the scale written once with every point numbered and labelled:

Here is a new slogan: "Fresh ideas, every day."
How strongly do you agree with the following statements about this slogan?

- The slogan is easy to understand.
- The slogan appeals to me personally.
- The slogan fits the brand.
- The slogan is believable.

Please rate each statement on the following scale:
1 = Not at all
2 = Slightly
3 = Somewhat
4 = Moderately
5 = Quite a lot
6 = Strongly
7 = Very strongly

When Elaiia recognises the matrix, the question shows “This question will be split into N separate questions”. Click “Show the questions” to read exactly what the twins will get. No split message means no split: an unrecognised matrix stays a single question and is usually reported as an open question.

Rows

Every row is asked on its own, so write each one as if it will be read alone.

Avoid

The slogan fits the brand and is believable.

Better

Two rows: "The slogan fits the brand." and "The slogan is believable." One rating can't cover two ideas.

Avoid

The slogan improves my view of the brand.

Better

Neutral statements. With a built-in direction, a low score could mean "no effect" or "worse" — ask that as its own question on a two-sided scale.

Put shared framing (product, brand, slogan, timeframe) in the question — it is repeated in every row.
Don’t let rows lean on each other (“same for the website”, “compared with the above”).
Name your stimuli (“Design A”, “Option 1”), or they become “the first image”.
Keep it to about 10 rows. Skip attention-check rows — twins don’t lose attention.
Avoid negations and double negatives (“not unattractive”).

Scale

The scale is copied word for word into every row.

Number the points if you want averages. “1 = Not at all … 7 = Very strongly” gives a mean per row; a word-only scale (“Never / Sometimes / Often”) gives percentages per option instead.
Label every point, not just the ends, with even steps that mirror around the midpoint.
Keep one direction and one length (5 or 7 points) across the whole study.
Put opt-outs on their own line with no number — “8 = Don’t know” on a 1–7 scale reads as a rating.
One scale per matrix. Importance and satisfaction in one grid turns every row into two questions.