A grading rubric is judged on the twenty-fifth paper, not the first. When it is vague, you grade the end of the stack differently from the start, you hesitate over every half point and end up rereading the first papers to even things out. When it is precise, every decision has already been made: all you have to do is note what the student did. This guide offers a method to create a solid grading rubric (or marking scheme), complete examples in four subjects and a few pointers for linking it to a competency-based assessment grid.
Why a written rubric changes grading
Many teachers keep a rubric “in their head”. Writing it down before opening the stack brings four concrete benefits.
- Fairness: two papers with the same mistake lose the same points, whether they are graded at 6 p.m. or 11 p.m.
- Consistency across the whole batch: fatigue and order effects (a strong paper makes the next one look worse) weigh less when every point is tied to a criterion.
- Speed: you no longer deliberate over every partial answer, you apply a rule set once and for all.
- Feedback the student can read: “1/3 on question 2, right method but wrong result” tells them what to work on; an overall grade tells them nothing.
The principles of a good rubric
Start from what was taught and asked
The rubric assesses the task, not what you would have liked to read. If the question asks students to “justify”, the justification must earn points. If it only asks for a result, don't penalize the lack of explanation. Reread your exam and underline the command words: they are what dictate the criteria.
Allocate by question first, then by criterion
Start with the weight of each question in the grade, based on its difficulty and the time it takes. Then break each question down into observable criteria: a method, a calculation, an argument, an example, a conclusion. A criterion must be something you can see on the paper without interpretation.
Separate method from result
A student who sets up the calculation correctly and slips on the arithmetic is not at the same level as one who understood nothing. Award points for the approach, for the result and, if the task requires it, for the justification. This is also what lets you award partial credit predictably.
Avoid double penalties
If a mistake in question 1 throws off question 2, decide the rule in advance: usually, question 2 is graded on its consistency with the result obtained in 1 (“error carried forward”). Without that rule, a single mistake can cost a third of the grade.
Decide what happens to missing answers
A question left blank scores 0, even if the rest of the paper is excellent. It sounds obvious, but a good overall impression easily tempts you to “round up”. Write it into the rubric, along with how an off-topic answer is handled.
Choose the granularity and the total
Half points are enough for most tests; quarter points are useful for short exercises with many criteria. Nothing forces the rubric to add up to a round number such as 20 or 100: an exam can naturally total 24 or 30 points. Just convert the score to your grading scale at the end (score / total × 100 for a percentage, for example) and announce the rounding rule.
The step-by-step method
- Write your own full answer key, as a very strong student would. That is when the exam's difficulties show up.
- List what is expected for each question: what must appear on the paper to earn full marks.
- Weight the questions, then split each question's points between its criteria.
- Write a descriptor for each criterion: what earns full points, what earns half, what earns zero.
- Note the cross-cutting rules: error carried forward, blank answers, presentation, spelling if it is assessed.
- Test the rubric on three or four papers of different levels before grading the batch (see below).
- Adjust, then freeze it: once grading has started, you no longer change the rules along the way.
Grading rubric examples by subject
The four examples below can be used as templates. Above all, they show the level of detail that is useful in the “What is expected” column.
Maths: linear function exercise (6 points)
| Criterion | What is expected | Points |
|---|---|---|
| 1a. Evaluating the function | x replaced by the given value (0.5) and correct result (0.5) | 1 |
| 1b. Finding an input from an output | Equation set up (0.5), solved correctly (0.5), conclusion written out (0.5) | 1.5 |
| 2. Slope | Formula (y₂ − y₁) / (x₂ − x₁) applied (0.5), result (0.5) | 1 |
| 3. Equation of the function | y-intercept found (0.5), full expression f(x) = ax + b (0.5); error carried forward from question 2 accepted | 1 |
| 4. Graph | Two correct points plotted (0.5), line drawn across the whole interval (0.5), labeled axes (0.5) | 1.5 |
The detail in brackets is not fussiness: it is what lets you grade a partial answer quickly. For more examples and for grading science papers, see the page on grading maths papers.
English: essay writing (20 points)
| Criterion | What is expected | Points |
|---|---|---|
| Following the task | Required form (story, letter, dialogue), point of view and constraints of the prompt respected | 4 |
| Organization | Introduction, distinct paragraphs with one idea each, conclusion or ending | 3 |
| Content and arguments | Relevant, developed ideas, at least two precise examples | 5 |
| Style | Varied and precise vocabulary, well-built sentences, appropriate register | 4 |
| Language | Spelling and grammar: 2 points for fewer than 5 errors, 1 point for fewer than 10, 0 beyond; same logic for punctuation | 4 |
For language accuracy, an error threshold is faster and fairer than taking off a point per mistake, which mostly penalizes long papers. More pointers on grading English essays.
History and geography: extended response (10 points)
| Criterion | What is expected | Points |
|---|---|---|
| Introduction | Topic introduced, time period or area defined, key question or outline stated | 1.5 |
| Knowledge | At least three precise, accurate elements: dates, people, places, concepts from the course | 3.5 |
| Organization | Two or three logical parts, linking words, one idea per paragraph | 2 |
| Developed examples | At least one example placed in context and explained, not just mentioned | 2 |
| Conclusion and expression | Clear answer to the question asked, accurate historical or geographical vocabulary | 1 |
For a source analysis, replace “Knowledge” with two criteria: extracting information from the source, then linking it to the course. See also grading history and geography papers.
World languages: written production (10 points)
| Criterion | What is expected | Points |
|---|---|---|
| Task achievement | Required text type, every point of the task covered, length respected | 3 |
| Coherence | Ideas linked together, varied connectors, paragraphs | 2 |
| Range of vocabulary | Vocabulary from the topic studied reused, few repetitions | 2 |
| Grammatical accuracy | Tenses and structures from the course mastered; errors do not get in the way of understanding | 3 |
These criteria follow the logic of most official language exam marking grids: first assess what the student manages to communicate, then the quality of the language. More examples on grading world language tests.
Common mistakes
- Overlapping criteria: “expression” and “clarity” often grade the same thing twice.
- Too many points for presentation, which ends up counting as much as a whole question.
- A rubric built after reading the first papers, which then adapts to the students' answers rather than to what was expected.
- Totals that don't add up: check that the criteria add up to the stated total, question by question.
- No room for a correct but unexpected approach: add a line saying “any correct method is accepted”.
Points-based rubrics and competency grids
A points-based rubric assigns points to answers; a competency-based assessment grid places the student on levels of mastery (not met, partly met, met, exceeded) for skills such as “reasoning”, “communicating” or “applying knowledge”. The two are not opposed. A well-broken-down rubric is already a grid: you just need to link each criterion to a competency. In the maths example above, “equation set up” belongs to “modeling”, “conclusion written out” to “communicating”.
In practice, many teachers grade with a points rubric, then group the criteria by competency to fill in the report. The reverse is also possible: convert each mastery level into points (for example 0, 1, 2, 3). What matters is that the descriptors are the same in both cases.
Test the rubric on a few papers
Before grading the batch, take three or four papers of different levels: a strong one, an average one, a weak one and, if possible, an unusual one. Grade them with the rubric and ask yourself three questions.
- Do the resulting grades match your overall judgment of these papers?
- Did you come across an answer the rubric did not anticipate?
- Did you hesitate over a criterion? If so, make its descriptor more precise.
This test takes a quarter of an hour and saves you from regrading twenty papers. It is especially useful for a common assessment or a mock exam, where several graders apply the same rubric.
Using your rubric with Correxio
Correxio grades papers, handwritten or digital, against your own rubric. You upload the exam statement (PDF, photo or Word document); while you import the papers, the AI drafts a rubric criterion by criterion. You edit it like a list: each criterion has its points, in steps of 0.25, and a description of what is expected.
The AI then grades each paper strictly against that rubric: a score per criterion with its justification, annotations placed on the scan and overall feedback. A criterion left unanswered scores 0. You review and adjust everything before exporting the grades as CSV and the annotated papers as PDF. The details are on the features page.
The principles in this guide apply directly: the more explicit a criterion's description, the more accurate the suggested score. “1 point for the method, 0.5 for the result” gives better results than “question 2: 1.5 points”.