Guides · Mark schemes
How to write a mark scheme that marks consistently
A mark scheme is only as good as the agreement it produces. If two colleagues read it and land three marks apart on the same essay, the scheme is doing less work than it looks. Here is how to write one for extended writing that holds up across a class set, a department and a marking tool.
Start with what you are actually assessing
Before writing a single level descriptor, write one sentence for each skill the task is assessing. "Analyses how the writer uses language to create tension." "Uses evidence from the sources to support a judgement." If you cannot write that sentence, the task is not ready to be marked, and no scheme will rescue it.
Most essay tasks assess two to four skills. Exam boards call these assessment objectives; you do not need to. What matters is that each one is distinct. A scheme that rewards "analysis" in one strand and "depth of understanding" in another will double-count the same quality and produce marks that bunch at the top and bottom.
Use levels, not points
For extended writing, a points-based scheme ("1 mark for each relevant quotation") rewards length and listing. A levels-of-response scheme rewards the quality of the answer as a whole, which is what you are trying to judge. This is how GCSE and A-level essays are marked, and it is the model worth copying even for a Year 8 homework.
A workable shape
- Four or five levels. Fewer and every essay lands in the middle; more and nobody can tell the levels apart.
- A mark range for each level, with ranges that do not overlap. Level 3: 9–12 marks, not "around 10".
- One descriptor per level per skill, written so that each level describes a recognisably different piece of work.
- An instruction on placing within the level: top of the range if the answer securely meets the descriptor, bottom if it only just does.
Write descriptors you could point to
The most common weakness in a home-made scheme is descriptors built from adjectives. "Good analysis", "very good analysis" and "excellent analysis" sound like three levels. They are actually one level and two moods.
A useful descriptor names something you could underline in the student's work:
- Weak: "Shows good understanding of the text."
- Better: "Explains the effect of specific word choices, linking them to the writer's purpose."
- Weak: "Excellent use of sources."
- Better: "Uses details from both sources to support a judgement, and weighs one source against the other."
The test is simple: if you showed the descriptor and an essay to a colleague, could they point to the sentence that earns it? If not, the descriptor is a feeling, and feelings drift between the first script and the thirtieth.
Make each level a different kind of answer
The jump between levels should be a change in what the student is doing, not just how well. A common progression for analytical writing:
- Describes or retells, with little or no reference to how or why.
- Identifies relevant features or evidence, with simple comment.
- Explains how features or evidence work, with some development.
- Analyses in detail, connecting several points into a sustained line of argument.
- Evaluates or judges, with a perceptive, well-supported overall view.
That ladder will not suit every subject, but the principle does: each rung is a distinct behaviour, so a borderline essay can be argued about with evidence rather than instinct.
Indicative content is an example, not a checklist
Under each question, list the points a strong answer might make. Then write, in so many words, that other valid points must be credited. Without that line, indicative content quietly becomes a checklist, and the student who made an original, well-argued point the scheme did not anticipate gets marked down for the omission.
This matters even more when someone else is applying the scheme, whether that is a new colleague, a cover teacher or a marking tool. They will read the list literally unless you tell them not to.
Say what not to penalise
Every subject has features that are easy to notice and should not move the mark. Write them down:
- Spelling, punctuation and grammar, unless the scheme has a separate strand for them.
- Length, beyond what the task requires.
- A different structure from the one you would have chosen.
- For students writing in a second language: imperfect expression that does not obscure the argument.
One line of "do not penalise" prevents more inconsistency than a page of extra descriptors.
Anchor it with real scripts
A scheme on its own is an abstraction. Two or three marked examples, ideally a strong answer, a middling one and a weak one, with the mark you gave each and one line on why, turn it into a standard. This is what exam boards do with standardisation scripts, and it is the fastest way to bring a department into line.
Choose anchors that sit clearly inside a level rather than on a boundary. A boundary script teaches people where the line is only if they already agree where the levels are.
Test it before the class set
Before you rely on a scheme, take three scripts and mark them with it as written, reading only the scheme and not your memory of what you meant. Then ask:
- Did any script fit two levels equally well? The descriptors overlap.
- Did you reach for a reason that is not in the scheme? Add it, or decide it should not count.
- Did the marks bunch? The levels are not distinct enough, or the ranges are too narrow.
Ten minutes here saves the experience of discovering on script twenty-two that the scheme does not say what you thought it said.
Writing a scheme a marking tool can follow
Everything above applies doubly when software is applying the scheme, because software does exactly what the words say. In practice:
- State the level and the maximum mark for each question explicitly. "Year 11, GCSE standard, out of 20" changes how a strong answer is read.
- Give every band a numeric range. A range lets the score be checked against the level it claims to be in.
- Include the "credit other valid points" line and the "do not penalise" list.
- Add anchor scripts with your marks if the tool supports them; they do more to match your standard than any amount of extra wording.
How Paddle uses your scheme
Paddle has no template library. You paste your own scheme, with its levels, ranges and maximum marks, and every script in the class set is marked against it. Each criterion's score is checked against the band named beside it, and each criterion quotes the phrase from the student's own work the judgement rests on, so you can see straight away whether the scheme was applied the way you meant. On paid plans you can add marked examples to anchor your standard. Nothing reaches a student until you have read it and pressed publish.
Common questions
How many levels should an essay mark scheme have?
Four or five is usually right. With fewer, most essays land in the same level; with more, markers struggle to tell adjacent levels apart and agreement falls.
What is indicative content in a mark scheme?
A list of points a strong answer might include. It should be presented as examples, with an instruction to credit any other valid point, so it does not turn into a checklist.
Should spelling and grammar count in an essay mark scheme?
Only if the scheme has a separate strand for them. Otherwise state explicitly that technical accuracy should not move the mark for content or analysis.
Can I use my own mark scheme with an AI marking tool?
With some tools, yes. Paddle marks only against the scheme you paste in, so your levels, ranges and wording decide the mark rather than a built-in template.
Paste your scheme and see what it does
Paddle marks against the scheme you write, not a template. Run one script you have already marked and read the reasoning it gives back.
Try Paddle free →25 marks a month on the free plan. No card.