Most of what students call revision is rereading. It feels productive, it takes a long time, and it does much less for what they will remember in June than a few minutes of trying to recall the same material without looking.
What the evidence actually says
Three sources are worth knowing, and it is worth being precise about what each one covers.
Roediger and Karpicke (2006), "Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention", Psychological Science 17(3). Students who were tested on material retained more of it a week later than students who spent the same time restudying it. This is the finding the whole idea rests on.
Karpicke and Blunt (2011), "Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping", Science 331(6018). Retrieval beat a well regarded active study technique, not just passive rereading, which matters because the comparison is the honest one.
Dunlosky and colleagues (2013), "Improving Students' Learning With Effective Learning Techniques", Psychological Science in the Public Interest 14(1). A review of ten study techniques that rated practice testing and distributed practice as high utility, and rated rereading and highlighting, the two things students actually do, as low.
What they do not say
They do not say retrieval practice raises GCSE grades. The strongest studies measure retention under controlled conditions, often with undergraduates, over weeks rather than a two year course. The classroom evidence base is thinner, and anyone quoting a grade improvement from this literature is going beyond it.
We have written our position on citing evidence on the evidence page, including how we label our own numbers.
What that means on a Tuesday
The classroom version is unglamorous:
Short. Six to ten questions on one narrow thing. A retrieval task that takes a lesson is a lesson.
Closed book. If the answer is visible, it is reading. The notes come out afterwards.
Low stakes. Nothing recorded. The memory effect does not require the mark to count, and making it count changes what students do about getting things wrong.
Corrected. The recall does the work, but the correction is what stops a wrong answer bedding in. Feedback that arrives next lesson has usually missed.
Spaced. Coming back to a topic after a gap is harder and better than consolidating it the same week.
Where the product fits
The last two are where software genuinely helps. Answers are checked as the student works, so the correction lands in the same minute as the mistake rather than in a week. Each question records how many attempts it took to get right, so "nearly knew it" and "guessed it" are not the same thing.
Nothing here needs marking, which is the practical reason spacing survives contact with a real term: a routine you have to mark is a routine you stop.
Start with the question-dense activities below, or browse everything by specification on the revision pages. The official past papers are there for when you want exam questions rather than practice ones.