Bloom's Taxonomy for exam questions.

• Remember — Recall facts and terms, dates, definitions, labelled diagrams.
• Understand — Explain an idea in your own words, summarise, give an example.
• Apply — Use what you know in a new, unfamiliar situation or problem.
• Analyse — Break something down and compare its parts, spot patterns or causes.
• Evaluate — Justify a position, weigh evidence, critique an argument or method.
• Create — Design, plan, or produce something new using what was learnt.

Left unchecked, a test naturally drifts toward Remember and Understand. They're the fastest questions to write, and the easiest to mark. The problem is that a paper built mostly from those two levels tells you who memorised the work, not who actually understands it well enough to use it.

A balanced paper mixes in Apply, Analyse, Evaluate, and Create questions too, so learners who've genuinely grasped the material have a chance to show it, and so the mark distribution reflects more than recall.

Go through each question and tag it with the level it's actually testing, not the level you intended when you wrote it. Add up the marks per level. If one or two levels make up most of the total, that's the gap to fix, usually by rewriting a couple of the easier questions to ask learners to apply, compare, or justify something instead of just stating it.

Setwise does this check automatically. When you generate a paper, it flags any level that's under-represented against the mix you asked for, and estimates how many marks would close the gap, before you finalise anything.