How to Make a Rubric: A Step-by-Step Guide for Writing
To make a rubric, decide what the assignment should prove, pick three to five criteria that match those goals, choose three to five performance levels, and describe what observable work looks like at each level. Then pilot it on sample papers and share it with students before they start writing. This guide walks through each step with a worked essay rubric you can adapt.
Writinglabwritinglab.ioBuild Clearer Writing RubricsWriting Lab offers rubric generation tools to help educators define assessment criteria and give students more actionable writing feedback.Explore Writing Lab
TL;DR
- Write the rubric alongside the assignment, so it clarifies your goals before students begin.
- Keep it short: most guidance lands between three and eight criteria, and many say three to five or six for analytic rubrics.
- Describe what the work shows, not whether it is “good” or “poor”.
- Weight higher-order concerns such as argument and evidence more heavily than mechanics.
- Test the rubric on real samples, then revise it after use.
Table of Contents
- What a writing rubric is, and what it must contain
- Step 1: Start from the learning outcomes
- Step 2: Pick the rubric type
- Step 3: Choose three to five criteria
- Step 4: Set the performance levels
- Step 5: Write descriptors that describe the work
- A worked essay rubric example
- Step 6: Decide the weighting and scores
- Step 7: Test it before you rely on it
- Adapting a rubric for IELTS, TOEFL and professional writing
- Making rubrics faster with Writing Lab
- Our perspective
- FAQ
- Sources
What a writing rubric is, and what it must contain
A rubric is a scoring guide that makes four things explicit: the task, the criteria, the performance levels and the descriptors for each level. ANU’s guidance on rubric elements says an effective rubric aligns criteria with learning outcomes, describes observable evidence in student-friendly language, and uses levels to show quality rather than hiding vague judgments in the criterion names.
The University of Washington’s writing program adds a useful frame: a good writing rubric works as a heuristic for feedback, describing higher-order characteristics of successful writing rather than serving as a checklist of penalties (Writing@UW).
In practice, a rubric is a grid. Criteria run down the left side as rows. Performance levels run across the top as columns. Each cell holds a short description of what work looks like at that point.

Step 1: Start from the learning outcomes
Begin with what students should be able to do after the assignment, and how they will show it. MIT’s teaching lab suggests listing the knowledge, skills, behaviors, task steps and final-product characteristics the work requires. Then prioritize the key skills and remove or combine the rest (MIT Teaching + Learning Lab).
Stanford’s TeachingWriting site offers a practical way to find criteria for writing: look at what strong examples of the assignment have in common, such as a clear main claim or thesis, use and presentation of sources, and required formatting (Stanford TeachingWriting).
A quick exercise: take two or three past papers you would call strong and ask what they share. Those shared qualities are your candidate criteria.
Step 2: Pick the rubric type
There are four common formats, described by the UNC assessment office:
| Type | What it gives you | Best for |
|---|---|---|
| Holistic | One overall judgment of the work | Quick scoring, large batches, timed writing |
| Analytic | A detailed description for each level of each criterion | Feedback students can act on |
| Checklist | Yes/no judgments | Requirements such as citation format or word range |
| Rating scale | Levels without full descriptions | Fast formative feedback |
The University of Missouri’s writing program also points to claim-evidence-reasoning, research-paper and single-point formats, and advises choosing the one that suits the assignment, the students and the instructor (University of Missouri Campus Writing Program). A single-point rubric lists only the expected standard and leaves space on either side for notes on where the work falls short or goes beyond.
Stanford also separates task-specific rubrics, built for one assignment or genre, from generic rubrics meant to cover several assignments. Missouri notes that generic rubrics can justify a grade but give limited direction for revision, so an assignment-specific guide is usually more useful.
Step 3: Choose three to five criteria
More criteria do not make a rubric fairer. They make it harder to use. Sources vary slightly on the ceiling:
- MIT: most rubrics have three to eight criteria.
- ANU: three to six for analytic rubrics, because too many can overwhelm students and markers while too few make judgments less valid.
- TU Delft: ideally three to five performance criteria, with the non-negotiable ones identified first (TU Delft Teaching Support).
For a standard argumentative essay, five criteria cover most of what matters:
- Thesis and argument: the main claim and how it is developed.
- Evidence and sources: the support offered and how it is used.
- Organization: the order of ideas, paragraphing and transitions.
- Style and audience awareness: word choice, tone and sentence variety suited to the reader.
- Conventions: grammar, spelling, punctuation and formatting.
Writing@UW cautions against making word count or mechanical correctness the principal criteria, and recommends higher-order criteria that matter to the genre and learning goals. Keep conventions on the list, but give them a modest share.
The IIT writing guide makes a related point: specify criteria in terms of how they appear in this assignment, because audience needs differ between, say, a technical report and a historical narrative (IIT guide to rubrics for writing). “Organization” in a lab report means something different from “organization” in a personal essay.
Step 4: Set the performance levels
The University of South Carolina recommends three to five consistently labeled levels, with descriptions that show qualitative differences focused on evidence of learning (University of South Carolina CTE). MIT says you need at least three levels, for inadequate, adequate and exemplary work, and usually no more than five, since finer distinctions are hard to apply consistently.
Labels should be parallel and neutral. Four works well for most classrooms:
- Exceeds expectations
- Meets expectations
- Approaching expectations
- Not yet evident
“Not yet evident” reads as a next step, which suits a rubric that doubles as a learning tool.
Step 5: Write descriptors that describe the work
Descriptors are where most rubrics succeed or fail. ANU’s advice is to describe observed work, avoid evaluative words such as “good” or “poor” and unfamiliar jargon, write parallel descriptions along a quality continuum, and make levels distinct enough that a paper can be placed unambiguously. TU Delft adds: use one or two sentences, set them at the standard’s level, and avoid relative wording like “better than”.
ACUE recommends writing the expected-level descriptor first, then the lowest and highest levels, and filling in the middle only where there is a real distinction to describe (ACUE).
Weak: “Good evidence.”
Stronger: “Each claim is supported by at least one specific, relevant example or source, and the writer explains how it supports the claim.”
The stronger version tells a student what to do and tells a marker what to look for.

A worked essay rubric example
Here is a four-level analytic rubric for a five-paragraph argumentative essay. Adapt the wording to your grade level and genre.
| Criterion | Exceeds | Meets | Approaching | Not yet evident |
|---|---|---|---|---|
| Thesis and argument | The thesis takes a clear, arguable position, and every paragraph advances it. Counterarguments are addressed. | The thesis takes a clear position, and most paragraphs support it. | The thesis is present but broad or only partly supported by the body. | No identifiable thesis, or the body does not relate to it. |
| Evidence | Claims rest on specific, relevant evidence, and the explanation links each piece to the claim. | Most claims have relevant evidence, with some explanation. | Evidence is general, or is listed without explanation. | Few claims are supported. |
| Organization | Paragraphs follow a deliberate order, and transitions show how ideas connect. | Paragraphs follow a logical order with basic transitions. | The order is partly logical, and transitions are missing in places. | Ideas appear in no discernible order. |
| Style and audience | Word choice and tone suit the reader, and sentence structures vary. | Tone is mostly consistent, with some variety. | Tone shifts, or sentences are repetitive. | Language often obscures meaning. |
| Conventions | Errors are rare and do not distract. | A few errors appear, but meaning is clear. | Errors sometimes interrupt reading. | Errors frequently block understanding. |
Notice that every cell describes something you could point to in the paper.
Step 6: Decide the weighting and scores
IIT’s guidance is to assign weights in proportion to the assignment’s purpose, giving more weight to higher-order rhetorical concerns and avoiding heavy penalties for low-level errors unlikely to decide whether the writing succeeds (IIT). Stanford’s analytic example allocates 30 points to complex ideas, reasons and logical argument and 20 points to sentence construction and copy editing.
A simple 100-point split for the rubric above:
| Criterion | Points |
|---|---|
| Thesis and argument | 30 |
| Evidence | 25 |
| Organization | 20 |
| Style and audience | 15 |
| Conventions | 10 |
Make the weighting visible on the rubric itself. Students should be able to see why a paper with a brilliant argument and a few typos outscores a tidy but empty one.
Step 7: Test it before you rely on it
A rubric is not fair just because it has point values. UNC recommends piloting the rubric on sample work, sharing it with colleagues and, when several people grade, calibrating by scoring common samples and discussing the standards. The University of South Carolina adds that you should share the rubric with the assignment and revise it after use.
A practical routine:
- Score three papers, one strong, one middling, one weak. If the rubric does not separate them, rewrite the descriptors that blur.
- Ask a colleague to score the same papers. Discuss every gap of more than one level.
- Hand the rubric out with the assignment and discuss it in class. Writing@UW suggests applying it to an anonymized sample paper with criteria-driven feedback where possible.
- After grading, note which cells you struggled to apply and fix them for next time.
TU Delft treats the rubric as a learning tool: provide it early, use plain English, involve students in improving the wording, and practice self- and peer-assessment with it.
Adapting a rubric for IELTS, TOEFL and professional writing
The same logic applies to exam practice and workplace writing: define the real task, then specify criteria in terms of how they appear there. For IELTS or TOEFL practice, build criteria around what the official exam scores, then add one or two of your own for the skills a particular learner needs to practice. For business writing such as emails and reports, replace “thesis” with “purpose and requested action” and make audience awareness a heavier criterion. Keep the same structure: few criteria, parallel levels, observable descriptors.
Making rubrics faster with Writing Lab
Drafting from a blank grid is the slowest part. Writing Lab’s rubric generation tool gives you a starting draft, which you then check against the steps above: criteria matched to your outcomes, descriptors that describe evidence, and weights that fit the assignment. The same platform also offers AI essay grading and writing feedback, so a rubric you build can be used to assess student work. You can try it on the Writing Lab site.

Start with the free tier to see how a draft compares with the rubric you would have written by hand.
Our perspective
We built Writing Lab around a simple belief: a rubric is only as useful as the feedback it makes possible. A grid of boxes that justifies a grade does not help a student revise. A grid that names what the work shows, and what the next level would look like, does.
That is why we pair rubric generation with essay grading and feedback in one place. A teacher can draft criteria from an assignment, adjust the descriptors to their classroom, and then apply the same standards to student work without retyping anything. The time saved on repetitive marking goes back to teaching.
We also think rubrics belong to students as much as to teachers. When learners, including those preparing for IELTS and TOEFL, can see the criteria and practice against them, the rubric becomes a guide to improvement, not a verdict.
Sabba
FAQ
How many criteria should a rubric have?
Three to five is a safe range for most writing assignments. Guidance from MIT allows up to eight, ANU recommends three to six for analytic rubrics, and TU Delft suggests three to five. Fewer criteria make the rubric easier to apply and easier for students to understand.
What is the difference between a holistic and an analytic rubric?
A holistic rubric gives one overall judgment of the work. An analytic rubric breaks the work into criteria and describes each level separately. Analytic rubrics take longer to build but give students more specific feedback.
Should I use a generic rubric or write one for each assignment?
Write one for each assignment when you can. Generic rubrics can justify a grade but offer limited direction for revision. If you need a reusable version, keep the structure and rewrite the descriptors for each task.
When should students see the rubric?
With the assignment, before they start writing. Stanford recommends creating the rubric alongside the assignment, and Writing@UW recommends distributing it and discussing it in class.
Sources
- Elements of rubrics, ANU Learning and Teaching
- On Grading Writing, Writing@UW
- How to Use Rubrics, MIT Teaching + Learning Lab
- Rubric Design, Stanford TeachingWriting
- Design a Rubric, TU Delft Teaching Support
- How to Create a Rubric, University of South Carolina CTE
- Create an Analytic, Holistic, or Single-Point Rubric, ACUE
- Developing Good Rubrics for Writing, IIT
- Using Rubrics to Assess Student Learning Outcomes, UNC Assessment
- Assessment, University of Missouri Campus Writing Program