How to Write a Rubric: A Step-by-Step Guide (With Examples)
A good rubric does two jobs at once: it tells students what excellent work looks like before they start, and it makes your grading faster and more consistent after they finish. A bad rubric does neither — it just adds paperwork. The difference is almost always in the criteria and the descriptors. Here's how to write the kind that earns its keep.
The four parts of a rubric
Every analytic rubric is a grid built from four things:
- Criteria — the dimensions you're judging (e.g. thesis, evidence, organization, conventions). These are the rows.
- Performance levels — the scale (e.g. Beginning → Developing → Proficient → Exemplary, or 1–4). These are the columns.
- Descriptors — the cell text describing what each level looks like for each criterion. This is the part that does all the work.
- A scoring rule — how the levels turn into a grade (points per criterion, or a holistic judgment).
Step 1: Start from the objective, not the assignment
Ask: what should a student be able to do by the end? Your criteria should measure that, and nothing else. If your objective is "construct an evidence-based argument," then thesis, use of evidence, and reasoning are criteria — but neatness and word count are not. A common mistake is grading what's easy to count instead of what you actually taught.
Keep it to 3–5 criteria. More than that and you're not assessing, you're auditing — and neither you nor your students will use it.
Step 2: Write the top descriptor first
Counterintuitive, but it works: describe Exemplary before anything else. That forces you to define your target in concrete, observable language. Then write Beginning, then fill the middle. Working from both ends inward keeps your levels evenly spaced.
Make descriptors observable. "Good use of evidence" is invisible — two teachers will score it differently. "Each claim is supported by a specific, relevant quotation that is explained, not just dropped in" is something you can actually see on the page.
Step 3: Use parallel language across levels
The only thing that should change from one level to the next is the degree, not the vocabulary. This makes the rubric fair and fast to use.
| Criterion | Beginning (1) | Developing (2) | Proficient (3) | Exemplary (4) |
|---|---|---|---|---|
| Thesis | No clear claim, or claim is just a fact | States a claim but it isn't arguable | Clear, arguable claim | Clear, arguable, and precisely scoped claim |
| Evidence | Little or no evidence | Some evidence, loosely connected | Relevant evidence for each point | Well-chosen evidence, explained and integrated |
| Reasoning | Claims asserted, not explained | Some explanation, gaps in logic | Each point reasoned through | Reasoning anticipates and answers counterpoints |
Notice that "evidence" runs none → some → relevant → well-chosen and explained across the row. That parallel structure is what makes grading a row a quick left-to-right scan instead of a re-read.
Analytic vs. holistic: which to use
- Analytic (the grid above) scores each criterion separately. Use it when students need specific feedback on what to improve, or when you're teaching the skill. It's slower to grade but far more useful to learners.
- Holistic gives one overall score from a single paragraph per level. Use it for quick sorting, timed exams, or when the whole is more than the sum of parts. It's fast but tells a student little about why.
For most classroom assignments where students will revise, analytic wins.
Step 4: Share it before they write — and grade with it in hand
A rubric handed out with the assignment turns into a planning tool: students self-assess against it as they draft. A rubric revealed only after grading is just a justification for the score. The research on criteria transparency is consistent — students produce better work when they know the standard in advance.
When you grade, mark the cell, don't re-invent the wheel. If you find yourself writing the same comment repeatedly, that comment belongs in a descriptor for next time.
A faster way to build one
Writing parallel, observable descriptors for every criterion is the slow part. EvidenceLesson generates a criterion-referenced rubric from your objective and grade level — with levels that stay parallel — and cites the assessment-design research behind how it's structured, so you can defend it to a coach or adapt it in minutes rather than build it from a blank grid.
Related method: Criterion-Referenced Rubric Generator — see the research and how to apply it.