Revision homework rarely stops because someone decided it wasn't working. It usually stops because of a sum nobody did in advance.
The arithmetic
Here's the sum, with illustrative inputs. They're not our measurements of anything, and they're not a claim about your school. Replace them with your own.
Four classes. Thirty students in each. One short revision task a week. Two minutes to read and mark each script. A thirteen week term.
| Marking every script | Self-checked, records reviewed | |
|---|---|---|
| Scripts to read each week | 120 | 0 |
| Minutes each week | 240 | 32 |
| Hours in a thirteen week term | 52 | 6.9 |
The thirty two minutes isn't nothing. It's three minutes per class to choose and share an activity, plus twenty minutes a week looking at what students recorded.
The difference over a term is 45.1 hours.
The size of that number depends entirely on inputs you should change, so it's worth paying more attention to what each column scales with. Marking scales with students: every extra child in the room adds two minutes. Choosing an activity scales with tasks: it costs the same whether the class is twelve or thirty two.
That's also why the answer can come out the other way. One class of four, a one minute mark, and a setup that takes ten minutes, and self-checking costs you more than marking did. It's worth doing the sum rather than assuming.
The routine that keeps it near zero
Choose once, use four times. The three minutes per task is per task, not per class. Setting the same activity to four classes costs three minutes if you choose it once.
Share a link rather than distributing anything. No printing, no collecting, no chasing the absent.
Let the answers land in the room. Feedback that reaches a student while they still remember the question is worth more than feedback you spent an evening writing.
Look at records, not scripts. Reading four students' record sheets at the start of a lesson takes two minutes and tells you what they think they can't do, which marking doesn't give you.
What still needs a human
This is the honest half, and it's the longer one.
Judging extended writing. A six mark answer, an essay, a piece of analysis. Nothing here touches those.
Working out why an answer was wrong. An automatic check knows that an answer is wrong. It doesn't know whether the student misread the question, misunderstood the concept, or knew it perfectly and made an arithmetic slip. Those three need different responses, and telling them apart is teaching.
Deciding what to reteach. A list of wrong answers isn't a plan. Choosing the one thing worth ten minutes of the next lesson is a professional judgement.
Noticing the student who has quietly stopped. No system reports this to you, by design, since the ones that do are collecting data about children in order to do it. Someone has to look at the room.
Anything that counts. Assessments that produce a grade need a person, a mark scheme and moderation.
What this does not save
It does not save planning. It doesn't save teaching, and it doesn't save the assessment cycle, reports, parents evening or any of the rest of it.
It saves one line: the reading of short revision scripts. That's the line with the most repetitions in it, which is why it's worth going after first, and it's also the one that quietly kills the revision routine in November.
Where to go next
The case for setting revision as a link rather than a task is on revision homework that marks itself. The evidence behind short recall practice, with its limitations, is on retrieval practice for GCSE. The record students keep is the printable revision record. To put a run of activities in front of one class, build a sprint, or browse by subject and board on the revision pages.