Types of assessment in education, and when each one earns its place

Most of us learned the types of assessment as a list to memorise, somewhere in a teaching qualification, and then went back to running the course the way the department had always run it. The terms stayed; the thinking behind them didn't always come with.

Which is a shame, because the categories aren't academic hair-splitting. Each one answers a different question, and running the wrong type is how a course ends up with plenty of assessment and very little information.

The four types, by the question each one answers

Diagnostic assessment — "what do they already know?" Run before teaching, to find out what students arrive with. A short pre-test, a concept inventory, a first task you don't grade. Its whole value is that it changes what you do next: if two-thirds of the cohort can already do week three, week three should change.

Formative assessment — "how is it going?" Runs during learning. It's low-stakes or no-stakes by design, because its job is to surface the gap while there's still time to close it. A quiz nobody's grade depends on, a draft with comments, a five-minute problem in class.

Summative assessment — "did they get there?" The judgement at the end. It counts toward the grade, it goes on the transcript, and it's the evidence a student met the outcomes.

Ipsative assessment — "how far have they come?" Measures a student against their own previous work rather than against a standard or a cohort. Common in portfolios, language learning and skills development, where the distance travelled is the thing worth knowing.

The other axis: compared to what?

Cutting across those four is a second distinction that decides what a mark means.

Criterion-referenced assessment compares work to a fixed standard: this is what a pass looks like, this is what excellent looks like, and any number of students can reach either. Almost all university assessment should work this way — it's what makes a grade portable and defensible.

Norm-referenced assessment compares students to each other, ranking them and distributing grades across the cohort. It answers "who's strongest here?", not "did they meet the standard". It's the right tool for selection — competitive entry, prizes — and the wrong one for judging whether someone can practise safely or competently.

The trap is running a criterion-referenced course and then quietly norm-referencing at the end, because the marks "came out too high". If the standard was clear and they met it, that isn't grade inflation. It's the standard doing its job.

Where courses go wrong

Too much summative, too little formative. The most common shape: two graded assessments, both late, and no cheap way for a student to find out they've misunderstood something in week four. Everyone discovers the problem at the same time — when it's too late to fix.

Formative work students skip. If it doesn't count, some won't do it. That's not a reason to grade it; it's a reason to make it visibly useful — feedback they can act on, in time to act on it, ideally on something that looks like the assessment that does count.

Diagnostic assessment that changes nothing. A pre-test whose results you never look at again is just an extra task. If you're not prepared to alter the plan, don't run it.

No ipsative view on a skills course. If a student arrived unable to do the thing and can now do it adequately, a purely criterion-referenced mark records "adequate" and loses the most interesting fact about their year.

A workable shape for one course

Diagnostic once at the start, and act on it. Formative running throughout, cheap and frequent, aimed at the same skills the graded work will test. Summative at the points where a real judgement is needed — fewer of them than most courses have, and richer. Criterion-referenced throughout, with the standard written down before you start marking.

That last part is where it usually falls apart. Criterion-referenced only works if the criteria actually exist in a form specific enough to apply consistently — which means a rubric with real descriptions per level, not four labels and a percentage. Writing that is an evening's work per assessment, and it's the step that quietly gets skipped.

Your course. Your style. Always aligned.

Quindaria builds the assessments and the rubrics that grade them, from your own course outcomes and level. [See a real grading rubric →](/example/rubric)

Get early access →