The least useful number in exam preparation is the one every practice test hands you at the end. Sixty-eight percent tells you how much you missed. It does not tell you what you missed, why you missed it, or what to do tomorrow morning, and with an exam date on the calendar those are the only three questions worth answering.
You can get all three answers out of a single practice set. It costs about an hour more than marking the test and closing the book, and it produces a short list of named gaps instead of a score. Here is the procedure I use, in the order that works.
1. Sit one timed set in the exam's own format
Not a chapter quiz, and not "some questions on the topic". Use the real format: a full IELTS Academic Reading paper, a past paper for a university module, the end-of-chapter questions for a school exam done cold with the book shut. Set the exam's real time limit and write down the clock when you finish each section.
What you should have at the end is an answer sheet with three kinds of marks on it: answers, blanks, and a time next to each section. All three are evidence, and the blanks and the times are the evidence most people throw away. It goes wrong in two ways. An untimed set hides every pacing problem you have, so you learn nothing about the last ten questions. A set you have seen before measures memory, not skill. If you recognise the passage, the score is not yours. Use material you have never opened, and the real clock, or skip the step and save the hour.
2. Log every miss by question type, not by topic
Topics do not repeat. The passage about coral reefs will never appear again, so "coral reefs: four wrong" is a fact with no use. Question types repeat every time. IELTS publishes a fixed list of Reading task types, from identifying information (True / False / Not Given) to matching headings and summary completion. A university exam repeats its verbs instead: define, compare, apply the rule to a new case, read the figure.
So the log looks like this. True / False / Not Given: four wrong out of seven. Matching headings: two wrong out of five. Summary completion: none wrong out of six. For a history paper: "explain a cause" is fine, "compare two sources" lost two marks out of three.
This is the filing system we built EduZen AI's practice around. A miss is recorded by what kind of question it was and what you did with it, so tomorrow's session can be built from it. But a notebook and a pencil do the same job. The filing rule is what matters, not the tool.
The way it goes wrong: you log by topic because that is how the textbook is organised, and you end up with a list of chapters to reread.
That list is the syllabus. You already had it.
3. Sort each miss into one of three causes
Go through the misses one at a time and put each into one of three piles. You did not know it. You knew it, but misread the question or the text. You ran out of time, so the answer is blank or was rushed.
The three piles need three different fixes, which is the whole point of sorting.
A knowledge gap needs study. A misread needs a slower first read of the question and a habit of underlining the word that limits it: always, only, most, before. A pacing miss needs a drill under the clock and nothing else. Studying the topic harder will not put those eight minutes back.
Take the commonest IELTS Reading miss as an example. In the test's own definitions, False means the statement contradicts the text, and Not Given means the text neither agrees with the statement nor contradicts it. A candidate who answers False when the passage merely mentions the topic does not have a knowledge problem. They found a sentence about the subject and treated its presence as a contradiction. That is a reading habit, and it goes in the second pile.
What you should see is that the first pile is smaller than you feared. Most people who sort honestly find that half their misses were never about knowledge.
It goes wrong when every miss is treated as "did not know", because that sends you back to rereading the chapter. Rereading is one of the techniques that Dunlosky and colleagues rated low utility in their review of ten study methods, while practice testing came out on top.
4. Check your high-confidence wrong answers first
This step happens during the test, not after it. Next to each answer, write one of three marks: sure, think so, guess. It takes two seconds a question and it is the most informative thing on the page.
Here is why. Koriat and Bjork showed that when you judge how well you know something while the material is in front of you, you judge it with information that will be missing at the test. They called the result an illusion of competence. A confident wrong answer is that illusion caught in the act. You would never have studied that item on your own, because you were sure of it.
The good news is in a second study. Butterfield and Metcalfe found that errors made with high confidence are the ones people correct most readily once they see the right answer, probably because the surprise makes them pay attention. So the "sure and wrong" list is both the most dangerous list you own and the cheapest to fix. Read the correct answer, say out loud why yours was wrong, and most of those items stay fixed.
Expect three to eight items on that list from one paper.
It goes wrong when you skip the marks because they slow you down, or add them after marking, when you already know which ones you got wrong.
5. Write tomorrow's session from the list, not from the syllabus
Pick one named gap and give it the whole session. Name it as a behaviour, not a score: "answers Not Given as False when the passage only mentions the topic", or "leaves the last eight questions blank", or "confuses the two sources in a compare question". A behaviour can be practised. "Reading: 28 out of 40" cannot.
The session itself is mostly questions of the type you missed, done under the clock, followed by the same log again. That is practice testing spread over days, the two methods at the top of Dunlosky's list, aimed at the one place you actually lose marks. After a week the log tells you which named gaps have closed and which have not, and that is the first time the score becomes useful: as a second reading on the same instrument, not as a plan.
It goes wrong in two familiar ways. Re-sitting the same paper makes the number go up because you remember the questions, and nothing has changed. And "studying everything to be safe" spends the week on the sixty-eight percent you already had.
When the list does not shrink
If a named gap survives three sessions, the sort was wrong, not the student. A "misread" that will not go away is often a knowledge gap hiding in the vocabulary of the question: you do not misread "contradicts", you do not fully know what it asks for. A "knowledge gap" that survives targeted study is often pacing in disguise: you know it, and you know it too slowly.
Move the item to the other pile and run the session again. If it still survives, do three questions of that type out loud in front of someone who can see where you go wrong. That is the one hour where a tutor earns their fee, and you now know exactly what to ask for.
The percentage keeps one job. It tells you whether the list is getting shorter. It never tells you what to put on it.
If you would rather not keep the log by hand, EduZen AI files every wrong answer as a named habit and builds the next session from it. You can start studying for free, no credit card required.