---
name: rubric-design
description: Use when asked to build a scoring rubric — criteria and performance-level descriptors for grading student work consistently — as distinct from [[formative-assessment]] and [[summative-assessment]], which decide when and what to assess; a rubric is the scoring instrument used once that decision is already made.
---

# Rubric Design

A rubric is a scoring instrument: a set of criteria, each with
descriptors for what performance looks like at each level, used to
grade student work consistently across students and across graders.
It doesn't decide when to assess or what task to assign — that's
[[formative-assessment]] or [[summative-assessment]] — a rubric is
what turns a completed piece of student work into a score once the
task itself has already been chosen.

## Key components

- **Criteria** — the distinct dimensions being scored (e.g.,
  "thesis clarity," "use of evidence," "organization"), each one
  something a grader can actually judge from the work in front of them,
  not a restatement of the assignment instructions.
- **Performance levels** — a fixed number of levels per criterion
  (commonly 3-5), consistent across all criteria so the rubric reads
  as one coherent scale rather than a different scale per row.
- **Descriptors** — concrete, observable language at each level for
  each criterion, specific enough that two different graders scoring
  the same work would land on the same level.
- **Weighting** — whether every criterion counts equally toward the
  final score, or some count more, decided and stated before grading
  starts rather than adjusted after seeing results.
- **Anchor examples** — sample student work exemplifying each
  performance level, used to calibrate graders and given to students
  in advance so they know what each level actually looks like.

## Descriptors that discriminate

The single failure mode that makes a rubric useless is descriptors
that don't actually distinguish between adjacent levels — if the
difference between "proficient" and "developing" on a criterion is
just the adjective ("strong evidence" vs. "some evidence") with no
concrete difference in what a grader should look for, two graders will
disagree on which level a given piece of work falls into, and the
rubric hasn't actually standardized anything. A descriptor needs to
name something observable: a specific count, a specific kind of error
present or absent, a specific feature the work does or doesn't have —
not just a stronger or weaker adjective describing the same vague
quality.

This is also what makes a rubric useful to students before they submit
work, not just to graders after. A student reading "uses evidence
effectively" at the top level has no way to check their own draft
against it; a student reading "cites at least three sources, each
directly supporting a distinct claim in the argument" can actually
self-assess against the criterion before turning the work in. Writing
descriptors concrete enough to grade consistently and concrete enough
to guide revision turn out to be the same exercise.

## Common pitfalls

- **Adjective-only descriptors** — levels distinguished only by
  stronger or weaker language ("excellent," "good," "fair") with no
  concrete, observable difference a grader can point to.
- **Criteria that restate the assignment** — a criterion like "follows
  the instructions" doesn't tell a grader what to actually look for in
  the work; criteria need to name the qualities that matter, not the
  fact of compliance.
- **Uncalibrated graders** — handing multiple graders the same rubric
  without anchor examples or a norming session produces inconsistent
  scores even when the rubric text itself is well written.
- **Too many criteria** — a rubric with a dozen fine-grained criteria
  slows grading down and buries what actually matters most in the
  assignment under equal-weighted minutiae.
- **Rubric written after grading has started** — deciding criteria and
  levels partway through scoring, rather than before any work is
  graded, lets early impressions of individual student work bias what
  the rubric ends up rewarding.

## Learn more

- [[formative-assessment]] for deciding when a check during instruction
  is warranted, before any rubric is needed to score it.
- [[summative-assessment]] for designing the end-of-unit task a rubric
  like this is often built to score.
- [[report-card-comments]] for translating a rubric's scores into
  narrative feedback a parent can actually use.