"I've finished the bank" is the most dangerous sentence in exam preparation, because completion and coverage are not the same thing. A bank reflects its authors' choices about what to write; the exam reflects an official blueprint about what to test; and the gap between them is invisible in any completion percentage. The fix is a Blueprint Coverage Matrix: a simple grid that maps your actual practice against the official blueprint, exposing the domains you have under-practised and the ones the bank under-covers. Build it once for your exam and you replace a comforting completion figure with an honest map of where you are actually ready — and it works for any exam with a published blueprint.
Why completion misleads
Three mechanisms make bank completion a poor proxy for coverage. First, banks are uneven: authors write more questions on topics that are easy to write and popular, and fewer on the dry, the rare and the awkward — so "100% complete" can mean thorough on the well-covered topics and thin on the rest. Second, your practice is self-selected: even within a complete bank, you attempt, repeat and remember unevenly, so your effective coverage is more skewed than the bank's. Third, blueprints are weighted: the exam samples domains in fixed proportions that no bank is obliged to match, so a bank can be complete and still misaligned with what the exam actually tests. Put together, a completion percentage can coexist with a domain you have barely touched and the exam samples heavily — which is precisely how well-prepared-feeling candidates fail on the parts they never mapped.
The Blueprint Coverage Matrix
The matrix is a grid with one row per blueprint domain and four columns: the domain's official weighting, your attempted-question count, your first-attempt accuracy on unseen items, and a status flag. Filling it in takes half an hour and turns a vague sense of readiness into a specific action list.
| Blueprint domain | Official weight | Your attempts | Unseen first-attempt accuracy | Status |
|---|---|---|---|---|
| (per official blueprint) | e.g. 15% | e.g. 320 | e.g. 74% | On track |
| … | e.g. 15% | e.g. 40 | e.g. 55% | Under-covered + weak |
| … | e.g. 4% | e.g. 8 | n/a (sample too small) | Blank spot |
Status follows simple rules: on track (weight matched, accuracy stable), under-covered (attempts well below weight-implied share), weak (accuracy low on an adequate sample), or blank spot (too few attempts to measure). The matrix's power is that it separates two things a percentage conflates — how much you have practised a domain, and how well you actually perform on unseen questions in it — and shows both against what the exam will demand.
How to build it in half an hour
Get the official blueprint first, from the exam body, and list its domains and weightings — not the bank's topic list, which is the thing you are auditing against. Then pull your attempted-question counts per domain from the bank's analytics. Then, for each domain, get an unseen first-attempt accuracy figure — this is the step candidates skip, because the bank's own repeated-question accuracy is inflated; use a separate unseen source or a fresh timed block. Finally, apply the status rules and read off the action list: under-covered domains get quota, weak domains get targeted practice, blank spots get their first proper sessions. Rebuild the matrix fortnightly; it is a living instrument, not a one-off.
Worked examples across exam types
A broad written exam (e.g. a licensing or membership Part 1). The matrix routinely exposes that the dry, high-weight domains — basic and clinical sciences, pharmacology — are under-practised relative to the enjoyable organ-specialty questions, and that low-weight domains (ophthalmology, palliative care) are blank spots. The completion percentage hid both; the matrix makes next fortnight's quotas obvious.
A situational-judgement or professional component. Here the matrix has to include the non-clinical domain explicitly, because clinical banks make it invisible. A candidate "90% through the bank" may have a blank cell for the professional-judgement paper that carries a third or half of the marks — the matrix forces that cell onto the page.
A content-mapped exam (e.g. one built on a published condition-and-presentation map). The matrix rows come straight from the official map, and the audit reveals which areas of clinical practice the adaptive feed or your own choices left thin — professional content and low-volume specialties being the usual culprits. Because the content universe is published, coverage here is unusually checkable, which makes the matrix especially powerful.
Failure modes and limits
The matrix is a coverage instrument, not a pass predictor — a fully green matrix means you have practised the blueprint broadly and perform on unseen items, which is necessary but not sufficient, because pacing, test-day conditions and standard-setting still apply. It depends on an honest unseen-accuracy figure; fill the accuracy column with repeated-question percentages and the matrix flatters you exactly where it should warn you. It assumes the blueprint is available and current — check the exam body's latest version, especially where blueprints have recently changed. And it is domain-level, so it will not catch narrow within-domain gaps; pair it with error-type review for those. Used within those limits, though, it is the single best antidote to the false comfort of completion.
The evidence and logic behind it
The matrix formalises a basic measurement principle: you cannot infer performance on a weighted, blueprinted assessment from an unweighted, self-selected sample. Exam bodies publish blueprints precisely because domain weighting is part of the construct being measured; ignoring the weighting when you audit your own readiness discards information the exam itself considers essential. And using unseen accuracy rather than repeated-question accuracy reflects the testing-effect literature: recognition of previously-seen items overstates durable, transferable knowledge, which is what the exam tests. The matrix is not a novel theory; it is standard assessment logic applied to your own preparation.
An iatroX workflow that demonstrates the framework
iatroX supports the two columns candidates find hardest to fill honestly: unseen first-attempt accuracy and blank-spot coverage. Because iatroX's exam banks are free for the UK core exams and select questions adaptively across the blueprint, you can use them as the unseen-measurement source for the accuracy column while learning in whatever bank you prefer — so your matrix's accuracy figures come from questions your main bank never showed you. When the matrix flags a blank spot, iatroX's adaptive selection can drive practice into it and then measure whether the gap has closed. The learning bank and the measurement source stay separate, which is what keeps the matrix honest.
Implement it this week
Build the matrix once, this week. Download your exam's official blueprint, list its domains and weights, pull your attempted counts from your bank, and get one unseen first-attempt accuracy figure per domain from a fresh timed block. Apply the status rules and write next week's quotas from the under-covered, weak and blank-spot rows. It will take thirty minutes and it will replace "I've nearly finished the bank" with a sentence that actually predicts your exam: "here are the three domains I have not yet covered to the blueprint."
A fully worked matrix
Take a candidate eight weeks out who reports being "85% through the bank". They build the matrix and it tells a different story. Against a blueprint weighted, say, 25% basic and clinical sciences, 15% pharmacology, and the remainder across organ systems, their attempts are heavily skewed to the organ systems they enjoy: cardiology 380 attempts at 78% unseen accuracy (on track), gastroenterology 320 at 75% (on track), but clinical sciences 90 attempts at 56% (under-covered and weak — a quarter of the exam), pharmacology 110 at 58% (under-covered and weak), and ophthalmology 8 attempts (blank spot). The 85%-complete figure was true and useless: it averaged thorough coverage of the enjoyable third with thin, weak coverage of the highest-weight domains. The matrix converts that into next fortnight's quotas — the largest going to clinical sciences and pharmacology, ophthalmology getting its first real block — and the candidate stops mistaking "nearly finished" for "nearly ready".
The accuracy column is where candidates cheat themselves
The single most common way the matrix is filled in dishonestly is the accuracy column. It is tempting to use the bank's headline per-topic accuracy — but that number blends first attempts with repeats, so it is inflated by recognition and overstates exactly the domains you have practised most. A domain can show 80% in the bank's analytics and 60% on cold unseen questions, and it is the 60% the exam will find. So the discipline is strict: fill the accuracy column only with unseen first-attempt figures, from a reserved source or a fresh timed block the bank's feed did not select. This is where the coverage matrix and the two-bank rule interlock — the second bank exists precisely to supply honest accuracy figures for the first bank's matrix. A matrix filled with repeated-question accuracy is a comfort object; a matrix filled with unseen accuracy is a readiness instrument, and the difference is entirely in that one column.
When the matrix reveals a bank problem, not a you problem
Sometimes the matrix exposes that a whole domain is thin not because you avoided it but because the bank barely covers it. This is a real finding, and it is the coverage matrix doing a job no single bank can do for itself: showing you where your chosen resource under-serves the blueprint. When that happens, the fix is not to grind the bank's few questions in that domain harder — that just drives recognition — but to add targeted material from another source for that domain specifically, or to lean on your measurement bank if its coverage there is better. The matrix, in other words, audits your bank as well as your effort, and a domain that stays a blank spot after diligent practice is telling you the bank is the limitation. That is exactly the kind of gap a second bank, chosen to complement the first, is meant to fill.
Frequently asked questions
Does this framework work for every medical exam? Yes, for any exam with a published blueprint or content map — the rows come from the official weighting, so the matrix adapts to any exam while the method stays constant. Exams without a public blueprint are harder to audit and should be approached with extra caution.
How often should the framework be updated? Rebuild the matrix fortnightly during active preparation, and immediately if the exam body publishes a new blueprint version, because coverage is measured against the current standard.
Which metrics are valid across different exams? Official domain weighting and unseen first-attempt accuracy are the two portable, valid inputs; bank completion percentage and repeated-question accuracy are not valid coverage metrics and should never fill the matrix.
How should AI-generated feedback be verified? If an AI tool claims to report your coverage, check its domains against the official blueprint and its accuracy figures against an unseen source — an AI's "coverage" summary is only as valid as the blueprint and sample behind it.
How does iatroX implement the framework? iatroX provides the unseen-measurement source for the accuracy column and adaptive selection to drive practice into blank spots, keeping the learning bank and the measurement source separate so the matrix stays honest.
Why "how many questions have you done?" is the wrong question
The cultural default in exam preparation is to measure effort by volume — questions done, banks finished, hours logged — and the coverage matrix exists to replace that with something the exam actually rewards. Volume is an input; coverage against the blueprint, measured on unseen items, is the outcome. Two candidates who have each "done 3,000 questions" can be in completely different places: one broadly covered and honestly measured, the other deep in a comfortable third and inflated by repeats. The matrix makes the difference visible, and in doing so it changes the question you ask yourself from "have I done enough?" — which has no honest answer — to "is every blueprint domain covered above floor and stable on unseen questions?", which does. That shift, from volume to weighted coverage, is the single most important mental move in preparing for any blueprinted exam, and it is the reason a completion percentage should never be the number you trust.
The matrix as a conversation with your future self
There is a practical reason to build the matrix early and rebuild it fortnightly: it is a running argument against the optimism that sets in near the exam. In the final weeks, a familiar bank's rising percentage and a sense of having "seen it all" conspire to produce confidence, and the matrix is the instrument that keeps that confidence honest — a domain flagged under-covered in week four does not become covered because you feel ready in week eight. Candidates who keep the matrix live tend to catch their blank spots in time; candidates who build it once and abandon it drift back to the completion illusion. Treat it as a fortnightly appointment with the version of you sitting the exam, and let its status flags — not your mood — decide where the next quota goes. The half-hour it takes each fortnight buys the one thing volume never can: an honest map of where you are actually ready.
The exam-specific content-gap checklists in this series apply this matrix to named blueprints, so once you have built your first matrix here, those child articles will feel familiar — they are the same instrument, pre-populated with the official domains for a given exam. Build it once as a habit and every future exam becomes a matter of swapping in the new blueprint's rows, which is exactly the portability a good framework should give you.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. This is a cross-exam framework article; exam-specific content-gap checklists in this series apply the matrix with named blueprints. Disclosure: iatroX operates free and paid question banks and positions itself here as the unseen-measurement source, not the sole learning tool. Corrections via the feedback route on iatrox.com. References: exam-body blueprints cited in the relevant child articles; testing-effect and assessment-weighting literature; related reading: why your Q-bank percentage is not your exam score.
Copy the matrix, fill it for your exam, then test the blank spots in iatroX →