How to write an essay marking rubric (a teacher's step-by-step guide)

Every teacher inherits rubrics — from the syllabus, from the faculty drive, from whoever taught the unit last. But sooner or later you have to write one yourself, and it turns out that knowing how to write an essay marking rubricis a different skill from knowing how to mark with one. A good rubric makes marking faster, makes feedback clearer, and makes your judgements defensible when a student or parent asks why the essay got the mark it did. A bad one does the opposite: it slows you down, hides what you actually value, and forces you to mark on instinct while pretending you didn't.

Here's a step-by-step way to build one that works — drawn from the same logic exam bodies use, sized for a classroom.

Step 1: Decide what the task is actually assessing

A rubric is not a list of everything good writing can do. It's a record of what this taskwas set to find out. So start upstream of the rubric: what were students meant to learn in this unit, and what would visible evidence of that learning look like in an essay? If the unit was about how composers represent an idea, then "analyses the composer's choices" belongs on the rubric and "engaging opening" probably doesn't — however much you enjoy an engaging opening.

This is the same discipline exam bodies apply: syllabus outcomes come first, and every criterion traces back to one. Your classroom rubric can do the same in miniature. If you can't say which outcome a criterion serves, it's decoration — cut it.

Step 2: Choose analytic over holistic (for essays, almost always)

A holisticrubric describes whole essays at each grade — "an A essay does this, a C essay does that" — and asks you to best-fit the essay to one description. An analytic rubric splits the task into criteria and marks each one separately. Holistic marking is quick, and it has its place in moderation and quick sorting. But for anything a student will learn from, analytic wins, for one reason: it tells the student which part to fix. "C+, solid effort" is a verdict. "Argument 4/5, evidence 2/5" is a to-do list.

Step 3: Keep it to three to five criteria

Ten-criterion rubrics feel rigorous and mark terribly. Nobody can hold ten separate judgements in mind while reading one paragraph, so the marks blur into an impression anyway — you've just done extra paperwork to disguise it. Three to five criteria is the workable range. For most essay tasks the spine is some version of:

• Thesis and argument — is there a position, and is it sustained rather than restated?
• Evidence and analysis — is evidence selected for the point, and analysed rather than retold?
• Structure and organisation — do paragraphs advance the argument in a deliberate order?
• Expression and control — does the language carry the ideas cleanly, in the register the form demands?

Add a task-specific criterion where the task demands one — engagement with a stimulus, use of a required text, a reflection component — and stop there.

Step 4: Write descriptors you could point to in an essay

This is where most rubrics fail. A descriptor earns its place only if two markers reading the same essay could agree on whether it's present. "Shows deep understanding" fails that test — deep according to whom? "Explains how the composer's language choices shape the reader's response, with reference to specific lines" passes it: either the essay does that or it doesn't.

The practical trick is to write the top level first, as a description of observable behaviour, then step the verbs down through the levels the way exam-style criteria do:

• analyses → explains → describes → identifies
• sustained → mostly maintained → inconsistent → absent
• selected and integrated → relevant → present but undigested

A descriptor is working when you can quote the line of the essay that satisfies it. If you can't imagine the quote, rewrite the descriptor.

Four levels per criterion is usually enough. Five if the cohort is large; three if the task is short. More levels than that and the distinctions stop being real.

Step 5: Make the marks say what you value

Weighting is the honest part of rubric design. If analysis matters most, it should carry the most marks — because students read the numbers, not your intentions. A 5/5/5/5 rubric announces that expression matters exactly as much as argument; if that's not what you believe, don't publish it. A simple rule: allocate marks in proportion to the teaching time each skill received. If you spent four weeks on analysis and one lesson on formal register, 8/4/4/2 tells the truth.

Step 6: Test it on real essays before you publish it

Before the rubric goes on the assessment notification, run three past essays through it — a strong one, a middling one, a weak one. You're checking for three failure modes: criteria that produce the same mark for essays you know are different (the descriptors aren't discriminating), essays that score well but you know are poor (a criterion is missing), and places where you hesitate over which level applies (the verbs are too close together). Ten minutes of testing saves a term of arguing with yourself mid-stack — and gives you annotated samples to teach the rubric to students before they write, which is where a rubric earns most of its value.

Where AI fits (and where it doesn't)

A well-written rubric has a second payoff: it's exactly the instrument an AI marker needs. Paste your rubric and a class set into Markand he applies it the way you wrote it — a score per criterion for each essay, with the quoted lines that earned it, in about a minute per essay. That's a first pass, not a verdict: you stay the marker of record, confirming or overruling every score, and the per-criterion results show you what to reteach — the same way a rubric-first marking workflow does by hand. Mark never writes or rewrites for students, and if students use him on drafts, they should acknowledge that honestly. Teachers get 25 free credits to test it on a real stack; plans start from $12 a month.

But the thinking — deciding what the task assesses, what the descriptors name, what the weighting says — that's teaching, and it stays yours. Write the rubric well and everything downstream gets easier: the marking, the feedback, and the conversation with the student who wants to know what a 12/20 means. With a good rubric, the answer is never "because I said so." It's "look — criterion three. Here's the line. Here's what the next level up asks for."