Humanities academics judge an essay as a whole; every Cadmus rubric format forced criterion-by-criterion scoring anyway. I shaped a fourth format that decouples the rubric from the score entirely.
Three formats, different maths, same requirement: a number against every criterion, and a grade derived from those numbers.
The assumption is invisible. Parts genuinely add up to a total.
It fights the marker. They already know the grade — then have to reverse-engineer per-criterion numbers to reach it.
Not a missing feature — a missing assessment philosophy. Every format assumed a grade is built up from parts. Whole disciplines assume the opposite.
Raised by the UK team all year. Rather than build from the ticket I ran calls with academics at two institutions — and they wanted it for genuinely different reasons.
Easy to absorb into a pitch, hard to ship. Naming it as a separate problem kept this bet small enough to land.
Rubric weighting was already in flight. A 0% weighting would have behaved almost like a qualitative rubric for a fraction of the effort. I argued against it.
Cheap and fast — but it overloads a field that already means something, leaves scoring machinery visible in a rubric with no scores, and confuses weighting for everyone using it properly.
More work, but the format states its own intent. Scoring inputs are removed rather than zeroed, the existing three formats are untouched, and a marker can't mistake what they've opened.
Zero isn't the same as absent. 0% is a scoring decision; qualitative is the absence of scoring as a concept. Encoding the second as a special case of the first breaks the existing flow to protect a shortcut.
Three of the four are things a reviewer would otherwise have reasonably assumed were included. Writing exclusions down was as valuable as writing requirements.
One format, three surfaces.
The “score out of” value has to come from a different place depending on the assessment type.
Design prototype, not shipped product — weighted beside qualitative, the two removals marked, and where the score moves to.
Open the full prototype →Canvas, Moodle and Turnitin aggregate criterion scores into a grade. A “no points” mode usually hides the display while the scoring model stays intact.
Scoring machinery absent from the criteria rather than concealed, and the final grade an independent, deliberate input by the marker.
Discovery, shaping, the requirements spec and the design prototypes are complete across all four assessment types. Awaiting a build slot.
I shaped requirements before defining how we'd know it worked — the success-metrics section is still empty. For an additive format “did anyone choose it?” is measurable from day one. With designs done, it's the one gap left before build.
Refusing the cheap path. Zero-weighting would have shipped sooner and quietly degraded rubrics for every teacher already using weightings correctly. Writing down the four things I deliberately wasn't building kept the bet reviewable.