Objective tests are easy to score fairly. Performance tasks are not — a poster, a role play, a research output, a cooking demonstration. Two teachers can look at the same work and give it very different marks.
A rubric for a performance task is what removes that inconsistency. This guide shows how to build one that actually works: choosing criteria, weighting them, and writing descriptors specific enough that scoring stops being a matter of mood.
The 10-second answer: Pick 4–6 criteria, weight them to 100%, and write an observable descriptor for every criterion at every level. If two teachers can score the same output differently, the descriptors are too vague.
Analytic vs holistic
A holistic rubric gives one overall score with one description per level. It is fast, and it is fine for low-stakes work.
An analytic rubric scores each criterion separately. It takes longer but tells the learner which part was weak — and gives you data on what to reteach.
For performance tasks that carry real weight in the grade, use analytic. The rest of this guide builds one.
Step 1 — Decide what you are actually scoring
Before writing anything, finish this sentence: “A learner has done this well if they can ______.”
Your criteria come from that answer — not from what is easy to count. The number of slides is easy to count and tells you nothing.
Keep it to 4–6 criteria. Fewer than four rarely captures the task. More than six makes scoring slow and the weights meaningless.
Step 2 — Weight the criteria
Weights should reflect what the task is for. If the point is scientific reasoning, accuracy carries more than neatness.
| Criterion | Weight |
|---|---|
| Scientific accuracy | 40% |
| Organization of information | 25% |
| Visual presentation | 20% |
| Oral explanation | 15% |
| Total | 100% |
The most common error is silent weighting — four criteria all scored out of 4, which quietly makes neatness worth exactly as much as accuracy. If that is not what you intend, say so with the weights.
Step 3 — Choose your levels
Four levels is the practical standard:
| Level | Points |
|---|---|
| Excellent | 4 |
| Good | 3 |
| Fair | 2 |
| Needs Improvement | 1 |
Four is deliberate — an even number removes the safe middle that scoring drifts toward.
Step 4 — Write descriptors that can be observed
This is where most rubrics fail. Compare:
Weak: Excellent — content is very good. Good — content is good. Fair — content has some errors.
That is not a rubric. “Very good” is not observable, and the levels are only distinguished by adjective.
Strong: Excellent — all scientific concepts are correct and each is explained in the learner’s own words. Good — concepts are correct, with one minor inaccuracy or an explanation copied from the source. Fair — several inaccuracies, or explanations that restate the source without understanding. Needs Improvement — major misconceptions that would mislead a reader.
The test: could two teachers, scoring the same output separately, land on the same level? If not, the descriptors need more specifics.
Describe what is present, not what is missing. “Includes three supporting details” beats “lacks detail.”
A complete sample
Task: Science investigatory poster on plant growth
Total: 40 points
| Criterion | Excellent (4) | Good (3) | Fair (2) | Needs Improvement (1) |
|---|---|---|---|---|
| Scientific accuracy (40%) | All concepts correct; each explained in own words | Concepts correct; one minor inaccuracy | Several inaccuracies; explanations restate the source | Major misconceptions |
| Organization (25%) | Layout leads the reader logically from question to conclusion | Mostly logical; one section out of place | Cluttered in places; reader must search | No clear structure |
| Visual presentation (20%) | Visuals clarify the data and are labelled | Visuals present and labelled | Visuals present but unclear or unlabelled | No visuals, or decorative only |
| Oral explanation (15%) | Explains without reading; answers questions confidently | Explains with occasional reference to notes | Reads most of the presentation | Cannot explain the work |
Step 5 — Convert the score to a grade
Compute the percentage, then transmute it using the table your grade level requires. Our MATATAG transmutation guide covers that step.
Record it under the correct component — performance tasks usually have their own weight in the class record, separate from written work.
Give learners the rubric before they start
A rubric handed out with the grade is a justification. A rubric handed out with the task is an instruction.
Learners who see the descriptors beforehand produce noticeably better work, because you have told them what “excellent” means instead of leaving them to guess.
Frequently asked questions
How many criteria should a rubric have?
Four to six. Fewer misses important dimensions; more makes scoring slow and dilutes the weights.
Should I use 4 levels or 5?
Four is easier to score consistently. Five gives a middle option that tends to absorb everything you are unsure about.
Can one rubric cover several tasks?
For similar tasks, yes — a general “oral presentation” rubric can serve a whole quarter. For different task types, no; the criteria genuinely differ.
How do rubrics fit with item analysis?
Difficulty and discrimination indices are built for objective items. For rubric-scored work, look at the average score per criterion instead — it shows which criterion the class struggled with. See our item analysis guide.
Do I need a rubric for every performance task?
For anything that carries weight in the grade, yes. It protects you as much as the learner when a score is questioned.
Build a rubric in about a minute
Skoolari generates a complete performance task rubric — 4–6 weighted criteria totalling 100%, four performance levels, and an observable descriptor for every cell — from your subject, grade level, and a description of the task.
Choose Rubric as the material type in the Instructional Materials tool, then print it or download it as Word.