How do you align feedback to success criteria?

What makes a success criterion markable?

A success criterion is markable when two teachers reading the same response would agree on whether it was met. That is a high bar and most criteria fail it. “Analyses the text” is not markable. “Explains how at least two language choices shape the reader’s view of the central character” is, because it names what has to be on the page.

The test is simple. Read your criterion, then ask what you would underline in a script to prove it. If you cannot point at anything, the criterion describes a quality of the student rather than a feature of the work, and the feedback will be unusable.

Keep the set small. Four to six criteria for a single task is usually enough, and each one should map to something you taught. If a criterion has no teaching behind it, either drop it or teach it before the task goes out. Long lists tend to be aspirational: they name everything good work might contain, then get ignored at eleven at night with thirty scripts left.

Write them in the language students hear in class, not the language of the achievement standard. The achievement standard is your reference point for judging the level. The success criteria are the translation of it for this task, this text, this cohort.

• Not markable: "shows understanding of the text". Markable: "explains what at least two quotations suggest about the narrator's motive".
• Not markable: "writes with flair". Markable: "varies sentence openings so no two consecutive sentences begin the same way".
• Not markable: "uses evidence well". Markable: "follows each quotation with a sentence explaining the effect of the word choice".

How do you turn a rubric line into a comment a student can act on?

The most reliable comment shape has three parts: the criterion it relates to, the evidence in the student’s work, and one next step. Naming the criterion tells the student which part of the task you are talking about. Pointing at the evidence stops the comment being an assertion. The next step gives them something to do on Monday.

Compare “needs more analysis” with “Analysis of language: you quote the storm scene twice, but both quotations sit at the end of the paragraph with nothing after them. Next time, follow each quotation with one sentence on what the word choice suggests.” The second is longer to write and far more likely to change the next piece of work.

One next step per criterion is the ceiling, and across a whole piece two or three in total often beats six. Students act on what they can hold in their heads. Decide which two changes would most improve the next task and say those loudly.

Separate the mark from the message where your reporting allows it. Many departments release comments first and the mark a lesson later, because a number at the top of the page tends to end the reading.

How do you keep the standard steady from the first script to the fortieth?

Drift is the standard problem. The first few scripts get careful attention, the middle of the pile gets pattern-matched against whatever you just read, and the last few get whatever energy is left. Order effects are real and they are not a character flaw. Noticing and correcting for them is part of what expert essay marking looks like in practice, and the correction is mostly procedural.

Three habits help. First, mark by criterion across the whole class rather than by student, so you are holding one standard in your head at a time instead of six. Second, pull a spread of five scripts before you start, read them without marking, and decide where the boundaries sit for this particular cohort. Third, re-mark the first three scripts at the end, once your standard has settled.

For a department, moderation is the real check. Criterion-labelled comments make moderation quick, because a colleague can see the basis for a judgement instead of inferring it from a number. When comments are just impressions, moderation turns into a negotiation about taste, and nobody’s marking changes as a result.

What actually belongs in a comment bank?

A comment bank pays back more than anything else you can build, and it works whether you keep it in a document, a spreadsheet or software. What you are doing is writing your best explanation of a recurring problem once, properly, instead of writing a worse version of it thirty times.

Bank three kinds of comment. The recurring misconception, written out in full: what students believe, why it does not hold, what to do instead. The sentence-level fix, with a before and after so the student sees the change rather than reading about it. And the next-step prompt tied to a named criterion, phrased as an instruction rather than a judgement.

Do not bank praise that could apply to anything. “Great effort” costs you a line and teaches nothing. Praise earns its place when it is tied to evidence: “your topic sentences now carry an argument instead of a summary” tells the student exactly what to keep doing.

Review the bank once a term. Comments you never use are noise. Comments you paste twenty times a week are pointing at a gap in your teaching sequence, which is worth knowing before the next cohort reaches that task.

Should students see the success criteria before they submit?

Yes, and earlier than most of us manage. Criteria handed out with the task beat criteria revealed with the mark. Criteria used in class against real examples beat both.

The exercise that pays off: give students two anonymous responses, one strong and one weaker, and have them mark both against the criteria before they write their own. They argue about where the boundaries sit, which is the thinking you want them doing while drafting. Use the same wording in your study guides and exemplars so the language is familiar by the time it turns up on their own work.

When students have already worked with the criteria, feedback stops reading as a verdict and starts reading as a next step. It also shortens the “why did I get this mark” conversation, because the answer sits in language they have already applied to someone else’s writing.

How does an AI marking tool fit into this?

Everything above works with a printed rubric and a pen, so treat this section as optional. What software adds is speed on the mechanical parts: checking every criterion against every submission, and surfacing the banked comment that matches the gap it finds.

A tool built for this reads your materials before it writes anything. In Jeddle, the JeddAI engine takes your success criteria, your rubric, your saved comment banks and the task instructions, then drafts a comment per criterion along with the evidence it relied on. Absences show up rather than slipping past: when a criterion has no supporting evidence in the response, the draft says so instead of skipping it.

The draft is a starting point, not a decision. You read each comment, keep what is accurate, rewrite what misses, and delete what does not apply before anything reaches a student. That review step is the point of the design. A model can misread a deliberately understated argument, miss context you hold about a particular student, or apply a criterion too generously, and only you will catch it.

Two limits are worth stating plainly. Vague criteria produce vague drafts, in JeddAI or in anything else, so the criteria work in the first section is not optional groundwork. And consistent application is not the same as correct application: applying the wrong standard evenly across forty scripts is still the wrong standard. Mark a sample by hand and compare before you trust a batch.

The time saving comes from editing rather than composing, and from not re-reading a script three times to check three criteria. It does not come from handing over judgement. The teachers who get the most out of a tool like Jeddle are the ones who did the criteria and comment-bank work first, which is the work that improves their marking anyway.