How to write a marking rubric that students can actually use

• Analytic rubrics have several criteria down the side and performance levels across the top. Each cell describes one criterion at one level. Use them for tasks with distinct components (an essay's argument, evidence, structure and expression) and whenever feedback needs to be specific.
• Holistic rubrics describe the whole piece at each level in one paragraph. They are quick to mark with but give thin feedback. Use them for low-stakes work or where the parts genuinely cannot be separated.
• Single-point rubrics list the criteria with the "proficient" descriptor only, leaving space either side to note where work fell short or went beyond. Students read them more carefully than a full grid, and they are the fastest to write.

Senior secondary authorities generally expect analytic rubrics: the QCAA's instrument-specific marking guides, the VCAA's performance descriptors and NESA's marking guidelines are all criteria-by-level grids.

Criteria are the qualities you are actually assessing. Take them from the task and the curriculum, not from the format of the product. "Uses evidence to support a contention" is a criterion; "Body paragraphs" is a section of an essay.

• Keep to three to six criteria. Beyond that, marking slows down and students stop reading.
• Each criterion should be independent. If two criteria always rise and fall together, merge them.
• Name the criterion with a verb phrase that names the skill: "analyses the causes", "applies the formula correctly", "sustains a personal voice".
• Where a curriculum code exists (a VCE key skill, a QCAA syllabus objective, an NSW outcome), record it against the criterion. It makes moderation and reporting far easier later.

Four levels is the sweet spot for most tasks. Three cannot separate "solid" from "excellent"; five or more invite hair-splitting nobody can defend in a moderation meeting. Name levels for what the work does, not for how you feel about it: "Developing, Sound, Strong, Exemplary" beats "Poor, OK, Good, Great".

Decide the points per level up front and keep them consistent across criteria unless you have a deliberate reason to weight one criterion more heavily. If a criterion matters twice as much, double its points rather than adding levels.

The descriptor in each cell is the whole rubric. Weak descriptors count things ("uses three sources") or rely on adverbs ("clearly explains", "somewhat explains", "poorly explains"). Strong descriptors name observable features of the work at that level.

• Write the top level first, in full. Then write the bottom level. Fill the middle by describing what is present and what is missing at each step, keeping the sentence structure parallel across the row.
• Prefer qualities to quantities. "Selects evidence that directly supports each claim" says more than "uses at least two quotes per paragraph".
• Avoid words that only make sense relative to the next cell. If a student reads one cell in isolation, it should still tell them what to do.
• Test each descriptor against a real piece of work. If two markers place the same piece in different cells, the descriptor is not doing its job.

• Hand the rubric out with the task, not with the results. Better still, have students mark an anonymised sample against it first.
• Mark by criterion, not by student: score every student's argument, then every student's evidence. It keeps your standard steady and halves the time.
• Return the marked rubric with one sentence per criterion about the next step. The grid is the mark; the sentence is the feedback.
• Moderate with a colleague on a handful of scripts before either of you marks the set.