River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Rubric Template and Grade Weight Check

Four documents and three sheets that measure each criterion's declared weight against its share of the marks a cohort actually lost.

Free download  ·  No account needed

Grade Distribution Check

What the weights claimed, against what they did

One row per criterion. Declared weight is the published claim. Realised weight is the criterion's share of the marks the cohort lost.

RowCriterionDeclaredMarks lostRealisedGapVerdict
C1
C2
C3
C4
C5
C6

Checks

CheckWhat it countsValue
CHK-1Declared weights sum
CHK-2Realised weights sum
CHK-3Criteria over five points from declared
CHK-4Marks the reweighting moves

Realised weight is the column a rubric template has no room for, and it is the one that says whether the declared weights described the marking.

Every rubric template is the same grid: criteria down one side, Excellent to Poor across the top, and you writing the adjectives into the cells. Both halves break. An adjective cannot be applied twice the same way, and applying it twice the same way is the one thing a grade appeal asks for. A declared weight nothing checks afterwards is a claim about the assignment rather than a description of your marking. So this pack anchors the descriptors, then computes what the weights did.

MGMT 315 Case Analysis Memo at Fenmore State University: six criteria, four levels, 24 scripts, two markers. Declared weight is a criterion's share of the marks available. Realised weight is its share of the 551.25 marks the cohort lost. Both sum to 100, so the gap between them is arithmetic. Citation and formatting was declared at 8% and decided 1.1%, because the portal enforces the file name and the citation style before a marker opens the file.

Analysis and framework use was declared at 25% and decided 35.1%, so six marks move off the floor criterion and the published rubric starts describing the marking. Clearly stated criteria, consistently applied, is the standard NECHE holds an institution to. The marks are what a reviewer eventually reads, since an accreditor's own standards must set expectations for success with respect to student achievement. One level up, the course design template crosses outcomes against the items measuring them.

The criterion declared at 8% that decided 1.1% of the marks

The Grade Distribution Check, the Moderation Record, and the Rubric itself.

Grade Distribution Check

MGMT 315, after one marking round

Fenmore State University. Six criteria, four levels, 24 scripts, 551.25 marks lost across the cohort.

RowCriterionDeclLostRealGapTop lvlVerdictNew
C1Problem identification15%52.509.5%-5.512/24Slightly over15%
C2Analysis and framework use25%193.7535.1%+10.15/24Underweighted33%
C3Evidence from the case20%135.0024.5%+4.54/24Close enough20%
C4Recommendation, feasibility20%125.0022.7%+2.76/24Close on weight20%
C5Structure and clarity12%39.007.1%-4.912/24Slightly over10%
C6Citation and formatting8%6.001.1%-6.921/24Floor, not a test2%
SumBoth weights partition100%551.25100%0.0Balanced100%

Six marks off C6 and two off C5, both onto C2. That leaves C2 at 33% against a realised 35.1%, which moves the published rubric toward the marking without overshooting one cohort's numbers.

Moderation Record

The same 24 scripts, read by marker

Mean level per criterion, twelve scripts each. Tolerance is half a level, or 2.50 marks on a 20-mark criterion.

RowCriterionRuttenbergVasquez-HaleLevelsMarksIn tolerance
M-C1Problem identification3.423.420.000.00Yes
M-C2Analysis and framework use2.672.75+0.08+0.52Yes
M-C3Evidence from the case2.832.92+0.08+0.42Yes
M-C4Recommendation, feasibility2.423.50+1.08+5.42No
M-C5Structure and clarity3.503.42-0.08-0.25Yes
M-C6Citation and formatting3.833.92+0.08+0.17Yes
SumCohort means, 73.90 and 80.17+6.271 of 6

What the C4 gap decides

ScriptMarkerTotalA minus floorShort byOne level up on C4
S01Ruttenberg83.7587.003.2588.75, over
S06Ruttenberg48.7587.0038.2553.75, no band

The six mark gaps sum to 6.27, exactly the difference between the two markers' cohort means. That reconciliation is what makes one row a finding rather than an average.

Rubric

What each level is worth, and what it says

A criterion worth w marks on four levels awards w times the level over four, published with the rubric.

CriterionWeightLevel 1Level 2Level 3Level 4
Problem identification15%3.757.5011.2515.00
Analysis and framework use25%6.2512.5018.7525.00
Evidence from the case20%5.0010.0015.0020.00
Recommendation, feasibility20%5.0010.0015.0020.00
Structure and clarity12%3.006.009.0012.00
Citation and formatting8%2.004.006.008.00

C2, the two levels most marks sit between

LevelWhat a second reader can point at
3Reaches a conclusion, with one gap in the reasoning the reader has to fill.
2Applies a framework correctly and concludes nothing the framework did not contain.

C4, before and after the moderation finding

VersionLevel 4 descriptor
OldRecommendation is well-supported and realistic.
NewOne recommendation, naming the first action and who takes it, the resource it consumes, and one condition checkable against the case.

Two adjectives cost 1.08 of a level between two markers. Three observable features replaced them, and the level 3 feedback line writes itself: no stated condition.

What's in the pack

01

Grade Distribution Check

Each criterion's declared weight against its share of the marks lost, with a reweighting that still sums to 100.

02

Moderation Record

Per criterion, each marker's mean level and the gap converted into marks, reconciled against the difference in their cohort means.

03

Rubric

Twenty-four cells, each naming a countable feature, a specific failure or a checkable threshold rather than an adjective.

04

Criterion to Outcome Map

Every criterion traced to the sentence in the brief that asks for it and the course outcome it evidences.

05

Feedback Bank

One entry per criterion and level, naming what put the work there and the change that moves it up.

06

Assignment Brief

The brief students receive, written so every failure mode it names maps to a level on a named criterion.

07

How the Weight Check Is Computed

The arithmetic written out, the three reasons a criterion drifts, and which fix each one calls for.

How to use it

  1. 1

    Open it in River, or download it

    Open the pack in River and let the agent write the descriptors and run the check, or download the seven blank Word and CSV files with no account.

  2. 2

    Send the brief and the weights

    Paste the assignment brief and the criteria with the percentage each one carries. Weights summing to 100 are what make the check possible.

  3. 3

    Anchor the middle two levels

    Level 4 and level 1 write themselves. Almost every contested mark sits between levels 2 and 3, so those two get written first.

  4. 4

    Add the marks, then reweight

    Send per-criterion levels for scripts already marked. The check names the floor criterion and the one carrying the assignment, and the transfer between them.

Frequently asked questions

Is this template free?

Yes. Download all seven files as Word documents and CSV sheets with no signup and no card. Edit with AI is the optional path, where the agent writes the descriptors from your own brief and runs the weight check on your own marks. The rest of the packs sit in the template library.

What format are the downloaded files?

Word documents (.docx) for the rubric, the assignment brief, the feedback bank and the method explainer, and CSV (.csv) for the three sheets, zipped together. They open in Word, Pages, Google Docs, Excel, Numbers and Sheets. In River the same content opens as live Docs and Sheets.

How is this different from a rubric generator?

A generator produces the grid and stops at the claim. This one anchors every cell to something a second reader can point at, and publishes the marks each level is worth. Then it goes back to the marks awarded, to check whether the weights you declared are the weights that decided the grades.

Do I need marks already awarded for this to work?

No, though it is the difference between a rubric and a checked one. Without marks you still get the descriptors, the outcome map and a starter feedback bank. Scored scripts add the realised weights and, where two people marked, the criterion they disagree about. A multiple-choice exam reads from item analysis instead.

Will it grade student work for me?

No. The judgement and the accountability are yours, and a mark produced by something that has not read the course is not defensible. What it does instead is name the two descriptors a script sits between and what the work would have to contain to clear the higher one. Where the task itself is the problem, assessment design to reduce cheating prices the alternatives.

What does 'Edit with AI' actually do?

It creates a free River account and installs this exact pack as a private workspace. The agent opens ready to read an assignment brief, an existing rubric, or a gradebook export carrying per-criterion levels, and to build the check from whichever of those arrives.

Can I use it on a course somebody else set up?

Yes, and the weight check is often more useful there, because you did not choose the weights. Where the whole course needs reading rather than one assignment, the inherited course audit reports what the materials assess against what the syllabus claims.

Find out which criterion your rubric only claims to weight

Download the blank pack as Word and CSV files, or open it in River and have the descriptors anchored and the weight check run on your own marks.

Edit with AI