How to write a marking rubric

A rubric that looks tidy on the assessment page can still fall apart the moment three tutors start marking with it. Most problems trace back to a handful of decisions made early: which criteria to include, how many levels, what each descriptor actually says, and how the weights add up. This guide walks through those decisions in order and finishes with a complete rubric you can adapt.

Start from the learning outcomes, not the task

Before you write a single criterion, put the unit's learning outcomes next to the assessment brief. Every criterion should trace back to at least one outcome, and every outcome the task claims to assess should appear in at least one criterion.

This sounds obvious, but it catches two common faults. The first is criteria that assess something nobody taught, such as "professional presentation" in a unit that never discussed formatting. The second is outcomes that are listed in the brief but have no criterion, so they are never actually assessed.

A quick check: write the outcome codes in the margin next to each criterion. If a criterion has no code, either cut it or be honest that it is a hygiene requirement and give it a small weight.

Choose criteria

A criterion names one quality you will judge. Good criteria have three properties.

• Distinct. Two criteria should not reward the same thing. "Analysis" and "Critical thinking" in the same rubric almost always overlap, and markers will double count.
• Observable in the work. You should be able to point to the part of the submission that the criterion is about.
• Few enough to hold in your head. Four to six criteria suits most written tasks. Past seven, markers stop reading the descriptors and start marking on overall impression.

Name each criterion with a noun phrase that says what is being judged, for example "Use of evidence" rather than "Evidence" or "Research".

Choose the number of levels

Levels are the columns of the rubric. Most Australian universities report grades in bands such as HD, D, C, P and N, and it is tempting to make the rubric match. That works if your descriptors can genuinely separate five levels. Often they cannot, and markers end up guessing between D and HD.

Some practical guidance:

• Four or five levels suit most tasks. Four forces a decision about whether work is above or below the pass line. Five maps neatly to grade bands.
• Three levels suit short tasks or hurdle requirements where finer distinctions add noise.
• Avoid six or more unless each level has clearly different observable features.

Whatever you choose, make the fail level describe what is missing or wrong, not just "does not meet the standard". A student who fails needs to know what the gap was.

Write the descriptors

Descriptors are the cells. Each one should describe what work at that level looks like on that criterion, in terms a marker can check against the script.

Write the pass level first. It is the standard the unit exists to certify, and it anchors everything else. Then write the top level, then fill in between. Writing top down tends to produce a ladder of adjectives ("excellent", "very good", "good") that tells a marker nothing.

Each step up should add something observable, not just more of the same adjective. Compare:

The right hand column still requires judgement, but two markers reading it are far more likely to land on the same level.

Set weights

Weights say how much each criterion contributes. Decide them by asking what matters most in this task at this stage of the course, and write them before you mark anything, not after you see how the cohort performed.

Common patterns:

• The criterion most tied to the core outcome gets 30 to 40 percent.
• Communication and referencing criteria usually sit at 10 to 20 percent each. Weighting them higher often means a well written but thin answer outscores a rough but insightful one.
• No single criterion should be small enough that markers ignore it. Anything under 10 percent is usually better folded into another criterion or treated as a requirement.

Decide how a level converts to marks. The simplest scheme gives each level a range within the criterion's weight, for example on a 30 mark criterion: N 0 to 14, P 15 to 18, C 19 to 21, D 22 to 25, HD 26 to 30. Ranges let markers place work high or low within a level without pretending the rubric is more precise than it is.

A complete worked example

The task: a second year public health unit, 2000 word policy brief recommending an intervention to reduce vaping among secondary school students in a named Australian region. Total 100 marks.

Mark ranges per level for a 30 mark criterion: N 0 to 14, P 15 to 18, C 19 to 21, D 22 to 25, HD 26 to 30. For 15 marks: N 0 to 7, P 8 to 9, C 10 to 11, D 12 to 13, HD 14 to 15. For 10 marks: N 0 to 4, P 5 to 6, C 7, D 8, HD 9 to 10.

Notice a few things. The descriptors in each row build on each other: a D in "Recommendation and feasibility" assumes the recommendation is already clear and addresses delivery. The pass column describes a real, acceptable piece of work, not a near fail. And the referencing criterion is deliberately light, because the unit cares more about how evidence is used than about the citation style.

Test the rubric before release

Pick three past submissions, or write short sample answers if the task is new: one you expect to fail, one borderline pass, one strong. Mark them with the rubric, and if possible ask a colleague to do the same without seeing your marks.

Look for:

• Cells neither of you used. They may describe work that does not occur.
• Places where you both hesitated between two levels. The descriptors there are probably too similar.
• A total that feels wrong. If the strong script lands on 68, check the weights before you change the descriptors.

Fix the rubric, then publish it with the brief so students can see it from the start.

Share it with students

A rubric only helps students if they read it. Walk through one row in class, show an anonymised example at pass and at distinction, and point out the difference. Ten minutes spent here saves many emails in the week before submission and many appeals after it.