Almost every candidate makes the same sequencing error. They open a question bank, switch on adaptive mode because adaptive sounds clever, and start answering. The engine dutifully targets their weaknesses, their percentage climbs, and they feel prepared. Then they sit the exam and meet an entire domain they have simply never seen. The problem is not the adaptive engine. It is that an adaptive engine can only target the weaknesses it has observed, and it has observed nothing at all about the topics you never attempted. Before you optimise, you must measure, and measurement has a specific method.
Key takeaways
- The first job of a question bank is measurement, not improvement, and the two need different modes.
- Record two separate figures per topic: your accuracy, and how many questions you have actually attempted.
- A topic with low exposure is not a strength, it is an unknown, and unknowns cause exam-day surprises.
- Only switch to adaptive practice once your exposure is broadly even across the whole syllabus.
- Rerun the audit periodically, because coverage decays as you drift back towards comfortable topics.
Accuracy and exposure are different questions
Every question bank dashboard will show you your accuracy per topic. Almost none will make you look at your exposure, and exposure is the number that matters first.
Consider two entries on your dashboard. Cardiology: 62 per cent, across 140 questions. Ophthalmology: 80 per cent, across 5 questions.
At a glance, ophthalmology looks like a strength and cardiology looks like a weakness. That reading is wrong, and dangerously so. The cardiology figure is a real measurement: you have sampled the domain properly and you know where you stand. The ophthalmology figure is noise. Four correct answers out of five tells you essentially nothing about whether you could handle the exam's ophthalmology questions, which will not be the five you happened to get.
Low exposure does not mean you are good at something. It means you do not know whether you are good at it, and that is a different and more dangerous position, because it does not feel like a gap.
The audit itself
The audit is not complicated and it takes a few sessions. Work in Standard mode, which lets you choose what you attempt rather than having it chosen for you, and sample deliberately across the entire syllabus rather than following interest or your current rotation.
Set a minimum exposure threshold per topic before you allow yourself to draw any conclusion from its accuracy. Twenty questions is a reasonable floor for a substantial domain, and even ten will tell you far more than five. Work through the syllabus, topic by topic, until every domain has cleared that threshold.
Then build the table. One row per topic, three columns: questions attempted, accuracy, and a verdict. The verdicts are only ever one of three things. Known strength: adequate exposure, good accuracy. Known weakness: adequate exposure, poor accuracy. Unknown: insufficient exposure, verdict withheld.
That third category is the entire point of the exercise, and it is invisible on a standard dashboard.
What the audit reliably reveals
Candidates who do this properly tend to find the same three things, and they are all uncomfortable.
First, their apparent weaknesses are simply the domains they have actually sampled. You cannot be bad at a topic you have not attempted, so the topics showing poor accuracy are, by definition, the ones you have engaged with. The genuinely neglected domains hide behind a small, flattering percentage.
Second, the distribution of their practice is far more skewed than they believed. Candidates are consistently surprised by how much of their question bank history sits in three or four favourite systems, and how little sits in the rest.
Third, entire domains are sitting at zero or near-zero. Almost every candidate has at least one, and in the exam-specific writing on this site the same domains recur: the clinical sciences block that no rotation teaches, the statistics and evidence component, the organisational and regulatory content, the specialties that carry fewer marks and therefore feel skippable.
When to switch to adaptive
Once every topic has cleared your exposure threshold, and not before, adaptive practice becomes genuinely powerful, because now the engine is working from a real map rather than a partial one. It can target your known weaknesses, it can space the material that is decaying, and it can keep returning the concepts you keep getting wrong.
Run it too early and you get something that feels like progress and is not: a well-optimised, efficiently targeted tour of the topics you were already doing. The engine is not at fault. It is answering the question you asked it, which was "where am I weak among the things you have seen me attempt", when the question you needed answered was "what have I not attempted at all".
Rerun it, because coverage decays
The audit is not a one-off. Left to your own devices over a twelve-week preparation, you will drift back towards the topics you find rewarding, because getting things right is more pleasant than getting things wrong, and your exposure will quietly re-skew.
So rerun the audit every three or four weeks. It takes an hour. Look specifically at whether any domain has stopped receiving attention, and at whether the domains you corrected have held. A weakness you fixed in week two and never revisited is not fixed, it is dormant, and it will reappear at the worst possible moment.
A reusable template
The whole thing fits on one page, and it works for any exam.
List every topic in the official syllabus or blueprint, taken from the exam's own published documentation rather than from a question bank's internal categories, because those are not always the same and the exam is sampling from the former. Against each, record questions attempted, accuracy, the date you last touched it, and a verdict of strength, weakness or unknown. Add a column for the blueprint weight if the exam publishes one, because a weakness in a heavily weighted domain matters more than one in a lightly weighted domain.
Sort by verdict. Attack the unknowns first, because they are the only category you cannot reason about at all. Then attack the weaknesses, weighted by their share of the paper.
Where iatroX fits
iatroX supports this sequence directly. Standard mode lets you drive the audit, choosing what you attempt so that you can systematically clear every domain rather than being served questions by an engine that has not yet met your blind spots. Once your coverage is even, the adaptive engine targets the weaknesses the audit exposed, and spaced repetition prevents the domains you have corrected from decaying back into weaknesses. Missed questions can be opened in the Socratic Tutor, which asks you to reason before it explains, so that a correction is understood rather than merely recorded. Try it with free sample questions at iatroX. For what your percentage actually does and does not tell you, see why your Q-bank percentage is not your exam score.
Frequently asked questions
Should I use adaptive mode from the start of my revision? No. An adaptive engine can only target weaknesses it has observed, so running it before you have sampled the whole syllabus produces an efficient tour of the topics you already attempt. Establish representative coverage first, then switch.
How many questions per topic count as adequate exposure? Twenty is a reasonable floor for a substantial domain, and ten will tell you far more than five. Below that, an accuracy figure is noise rather than a measurement, and should not be treated as evidence of strength.
What is the difference between a weakness and an unknown? A weakness is a domain where you have sampled adequately and scored poorly, so you know where you stand. An unknown is a domain you have barely attempted, so you have no idea. Unknowns are more dangerous, because they do not feel like gaps.
How often should I repeat the syllabus audit? Every three or four weeks. Coverage decays as you drift back towards comfortable topics, and a weakness corrected in week two but never revisited is dormant rather than fixed.