How to Make Revision Notes for University Exams: High-Yield Formula Sheets and Summary Tables

Build the sheet assuming you cannot take it in, which is the case unless your paper is explicitly open-book. One page per unit, drawn from the last five question papers rather than from the syllabus, with every formula carrying its conditions and units. Three to four hours per unit. The compression is the revision; the page is only the residue.

Students asking how to make a notes for the exam usually mean the last two weeks, not the semester. That is a fair question with an awkward answer, because the evidence on condensed revision sheets turns on one variable most guides never mention: whether you expect to have the sheet in front of you.

The revision notes kit at the end has the one-page planner, a past-paper tally sheet and a formula card spec.

The 1-page cheat sheet strategy: forcing 10 weeks of material into a single view

Start with what the research actually supports, because the popular version of this advice is stronger than the evidence behind it.

Work on crib sheets goes back decades and has not settled. A 2024 study opens by saying exactly that: research on crib sheets as a study aid "spans several decades and provides inconclusive results" (Whitehouse, Bredican, Heaford, De Regt & Plangger, 2024, Studies in Higher Education, 50(12), 2838–2856). That team analysed 168 crib sheets from UK business students across six undergraduate and graduate classes and interviewed fourteen of them. What the interviews found was mixed in an interesting way: students described the sheets as a source of security and as a stressor at the same time.

The sharper finding is older and comes from an experiment. Funk and Dickson ran two rounds with 51 students. In the first, students prepared a crib card expecting to use it, then sat an unannounced pretest without it, and did significantly worse than on the same questions when the card was allowed. In the second round students built a card for an exam they believed they could not use it on, and an unexpected posttest afterwards showed no difference whether they had the card or not. Their conclusion is the sentence to keep: "Students' expectations of crib card use fostered dependency and reduced learning, while expectations of non-use did not" (Funk & Dickson, 2011, Teaching of Psychology, 38(2), 114–117).

Read that against a closed-book end-semester paper and it stops being an academic curiosity. If you cannot carry anything in, you are already in the condition the experiment found harmless, and the sheet can only earn its keep through how it is made. So the instruction that follows is not motivational. Build the page as though it will be collected at the door, and check your examination rules rather than assuming either way.

That changes the construction rules in four ways.

Compression has to happen by decision rather than by shrinking. Six-point type and no margins produce a page you cannot read and did not think about, so the constraint has to bite at the selection stage instead: one page per unit, and a line earns its place only if you could not reconstruct it from another line already there.

What goes on the line is a trigger, not the content. "Bernoulli: steady, incompressible, along a streamline" is usable. A transcribed derivation is not, because you will never read it under time pressure, and copying it taught you nothing the textbook had not already given you.

Gaps stay visible. A unit you cannot fill a page for is not a finished page with white space; it is a unit you have not studied, and that blank is the most useful thing on the sheet.

Finally, build it twice: once about three weeks out, then again from memory a week later. The second attempt is the measurement. Whatever survives onto the blank page is what you have, and whatever you had to look up again goes on the short list for the final week.

One honest limit. None of this says the sheet itself raises marks, and the literature does not support promising that. What it supports is narrower: deciding what goes on one page is a demanding pass over the material, and doing it without expecting a safety net does not appear to cost you anything.

Past paper integration: reverse-engineering exam notes from previous questions

The syllabus tells you what may be asked. The question papers tell you what is asked. These are not the same document, and revision notes built from the first are usually the wrong length in the wrong places.

Work backwards from five years of papers for the same subject and the same university. Not solved guides; the question papers themselves.

The first pass is a tally, not a read. Go through the papers marking each question against the unit it belongs to, ten minutes a paper. What falls out is a distribution: units that appear every year, units that appear in alternate years, and units that have not been set since the syllabus changed. Let that distribution fix the page budget, because an hour spent on a unit worth four marks once in five years is an hour taken from somewhere else.

Then separate the command word from the topic. "Derive", "compare", "explain with a neat sketch" and "a problem with given data" demand four different kinds of preparation on the same material, so tag each tallied question with its verb. A topic that always arrives as a numerical needs worked steps and standard data on your sheet; the same topic set as "discuss" needs three argued points and an example, and no formula at all.

Structure repeats more reliably than wording, which is the third thing to look for. If the heat transfer question has opened with a steady-state conduction part in four years out of five, the conduction entry on your sheet is doing more work than its share of the syllabus suggests.

Last, carry the paper's own vocabulary into the notes. Where the papers say "with usual notations", your formula card should use the usual notations. Matching the examiner's phrasing is not gaming anything; it removes a translation step you would otherwise perform in the hall.

Two cautions, because this method fails in predictable ways. A recent syllabus revision invalidates older papers in part, so check the revision year before trusting a five-year pattern. And a unit's absence from past papers is weak evidence, not permission to skip it; the sensible response is a smaller allocation, not zero.

Formula and definition card layout

The structured component of this method is a card format, used identically for every entry so that your eye learns where to look. For a technical subject, five fields:

For a definition rather than a formula, the same five slots carry the term, the one-sentence definition in the textbook's own words, the boundary case that distinguishes it from the term students confuse it with, the standard example, and the appearance record.

The conditions field deserves the emphasis. A formula written without its assumptions is a formula you will apply to a question it does not fit, and the marks lost there are not arithmetic marks. Write the conditions first and the expression second if that helps; the order on the card can be whatever stops you skipping the field.

Keep the cards uniform in size and do not mix subjects on one card. When the page budget from the tally sheet says a unit gets twelve cards, twelve is the number, and the thirteenth card means something earlier was not compressed.

Where Sovi.AI fits

The accurate statement first, because the alternative is to let the tool sound like it does the hard part.

Sovi.AI does not extract a syllabus or predict exam topics. Nothing on the Live Recording page reads a course outline and returns likely questions; the structured outputs its FAQ describes are comprehensive lecture notes, summaries and action items. So the tally in the previous section stays manual work on real question papers, and no amount of note automation replaces it.

Where the product helps is the input side, which is where most of a semester gets lost in the first place. Live Recording transcribes a lecture as it happens and turns the transcript into those formats. The value for this workflow is narrow but real: a summary of week three is a faster starting point for a card than forty minutes of re-reading, and it keeps the worked examples the lecturer used, which is usually the part you have forgotten rather than the definition. Its AI Learning Assistant renders explanations with Markdown and LaTeX, so an expression you are copying onto a card arrives as an expression instead of mangled text. Check which spoken languages are listed before depending on it: the FAQ names English, Chinese, Spanish, French, German, Japanese and Korean.

Two limits worth stating plainly. A generated summary tells you what was said, not what is examinable, and only the past papers answer the second question. And the conditions field on a formula card should be checked against the textbook rather than a summary, because that field is exactly where a condensed version drops the qualifier you are being tested on.

The revision notes kit

• revision-notes-kit.pdf to print and fill in by hand.

Inside:

1. A one-page planner, one sheet per unit, with the page split into a selection column and a final-lines column, plus two date boxes so the first build and the from-memory rebuild are both logged.
2. A past-paper tally sheet with rows for units and columns for five years, a mark column, and a command-word column for tagging each question.
3. A formula and definition card spec with the five fields pre-printed and eight blank cards to a page.

Run the tally sheet before the planner. The page budget it produces is what stops the first unit from eating the week.

Frequently asked questions

1. How long should one unit take?

Three to four hours for the first build, including the tally, and under an hour for the from-memory rebuild a week later. A unit that takes a full day is being rewritten rather than compressed.

2. Is it worth making the sheet if my exam allows one?

Make it, but make it early and then revise without it. The experimental finding is about expectation, not about the paper itself, so the risk is that an allowed sheet quietly becomes the plan instead of the record.

3. How many past papers are enough?

Five years of the same paper from the same university. Fewer than three gives you noise, and going back further runs into syllabus revisions that make the pattern misleading.

4. Should definitions and formulas share a sheet?

Separate cards, same page. Mixing them on one card means you cannot cover one and test yourself on the other, which is the only use the cards have in the final week.

5. What if my course has no past papers available?

Use the question patterns from tutorials, internal assessments and the end-of-chapter exercises your department sets, and weight the tally towards the internals, since those are usually written by the same faculty.

6. Typed or handwritten?

Handwritten for the first build, because the slower medium forces the selection the method depends on. Typed is reasonable for the rebuild if your handwriting costs you reading time later.