A rubric is a function

A rubric looks like a table and behaves like a formula. You mark one level per criterion; the marks become points; the points become a percent. Whatever else it does — align expectations, speed up feedback, survive an appeal — it is first a function, and you can read it.

The arithmetic, in one line

Each criterion contributes its weight times the points of the level you marked, out of its weight times the top level:

$\text{score} = \frac{\sum w_i \cdot p_i}{\sum w_i \cdot p_{\max}}$

That's the same weighted average a course grade uses — which is not a coincidence and is worth saying to students: the rubric isn't a different system, it's the gradebook applied to one piece of work.

Work one through. Four criteria on the classic 4·3·2·1 scale:

Earned 29, top possible 9 × 4 = 36, so 29 ÷ 36 = 80.6% — a B.

Weights are relative, not percentages

The weights above are 3, 3, 2, 1. They could be 30, 30, 20, 10 or 6, 6, 4, 2 and the score would not move by a hair, because they only ever appear as ratios. This is the single most useful thing to know when building one:

Don't try to make the weights add to 100. Ask instead: is thesis worth three times mechanics? Answer that, write 3 and 1, and let the arithmetic normalize. Rubrics get bad when a teacher forces the column to total 100 and ends up assigning 17% to something because that's what was left.

What the weights are really saying

Every weight is a public claim about what the assignment is for. Two rubrics with identical criteria and different weights are two different assignments:

• Thesis 3, Mechanics 1 → an argument task where surface errors cost something but can't sink you.
• Thesis 1, Mechanics 3 → a copy-editing task wearing an essay's clothes.

Neither is wrong. But if you tell a class the essay is about argument and then weight mechanics heaviest, the rubric is telling the truth and you aren't — and students read the rubric.

A useful test before you publish one: cover the criteria names and read only the weights. If the shape doesn't match what you'd say the assignment is about, fix it now, not in the comments afterward.

Partial marking is a real state

Half a rubric marked is a real situation — you graded thesis and evidence on the first pass and stopped. The builder grades what has been judged: unmarked criteria are left out of both sides of the fraction rather than counted as zero.

That distinction matters. Counting an unmarked criterion as zero doesn't report an incomplete grading job; it reports a failing student. The number should say "here's the score of what I've read so far", and then say how much that was.

Run it

Open the rubric builder with the essay rubric loaded:

1. Confirm the 80.6% and the 29 of 36 from the table above.
2. Change the weights from 3/3/2/1 to 30/30/20/10. Nothing moves — the weights are ratios.
3. Now flip them: mechanics 3, thesis 1. The marks on the page never change, and the grade falls from 80.6% to 69.4% — a B to a D. Eleven points of grade lived entirely in the weighting, which is another way of saying the weights decided what the assignment was.