You have completed 2,000 practice questions. The progress bar says you are most of the way through the question bank. Your last two mock scores are within a point of each other.
This pattern is common, and it is not a mystery. The number of questions completed measures activity. It does not measure what those questions changed in your understanding.
Question volume is a vanity metric. The learning happens in the error loop: diagnosing why an answer was wrong, fixing the specific gap, and testing it again later. A candidate who deeply processes 800 questions is better prepared than one who rushes through 3,000.
Why the Counter Rises While the Score Does Not
A question bank rewards completion. Every answered question moves a counter, and the counter feels like progress. That feeling is the problem.
When you answer correctly, you mostly confirm something you already knew. When you answer incorrectly, read the explanation and move on, you have seen the right reasoning once. Neither event reliably changes what you will do on the next question that tests the same idea in different wording.
The mock exam is a test of transfer. It asks familiar concepts in unfamiliar forms. If your practice has produced recognition rather than understanding, the mock score stays flat no matter how many questions sit behind it.
Questions completed measure effort. Errors resolved measure progress.
The Four Miss Types and What Each One Tells You
Not every wrong answer means the same thing. The first step of the error loop is classification, because each type of miss calls for a different fix. Treating every error as "I need to review this topic" is how candidates spend hours rereading material they already understand.
Concept gap. You did not know the idea, or you held a wrong version of it. You chose the distractor because it matched your mental model. The fix is to rebuild the concept from your notes, then test it on fresh questions.
Misread. You knew the concept but answered a different question from the one asked. You missed an "except", used a before-tax figure where the question wanted after-tax, or took the wrong time period. The fix is a reading habit, not more content: identify exactly what is being asked before you look at the answer choices.
Calculation slip. The method was right and the arithmetic or the calculator input was wrong: a sign error, the wrong compounding frequency, a payment entered in the wrong register. The fix is mechanical. Use one consistent calculator routine and check the magnitude of the answer before you commit.
Time pressure. You would have answered correctly with another minute, or you guessed because the clock forced you. If one topic is consistently slow, that is a fluency gap and belongs with concept work. If you are slow everywhere, it is a pacing problem and belongs in timed sets.
The distinction matters. A concept gap costs you on every question that touches the idea. Misreads and slips are cheaper to fix, but they recur until the habit changes. Without classification, you cannot tell which problem you actually have.
A Minimal Error Log That Takes 30 Seconds per Miss
Most error logs die within a few weeks. They turn into a second set of notes: copied explanations, full question text, color codes. The log becomes bureaucracy, and bureaucracy loses to fatigue.
A useful log needs four fields and nothing more:
- Reference: the question ID or a short label, so you can find it again.
- Concept: as specific as you can make it. "Fixed Income" is too broad. "Modified duration, price change estimate" is useful.
- Miss type: concept gap, misread, calculation slip or time pressure.
- Cause in one line: in your own words, the specific thing that went wrong. "Used D0 instead of D1." "Forgot to convert the semiannual yield."
That last line is the point of the exercise. Writing the cause in your own words forces a diagnosis. If you cannot write it in one line, you have not yet understood the error, and that is itself a signal to return to the explanation.
In practice, this takes about 30 seconds per miss. A session of 30 questions with 10 misses costs five extra minutes. They are the most productive five minutes of the session.
Redo Policy: Learning or Memorizing the Answer
Redoing questions is useful and dangerous in equal measure. A second attempt is only worth something if you are solving the problem, not recalling it.
A practical test: if you recognize the question and remember the answer before you have worked through it, the attempt tells you nothing. You are testing your memory of that item, not your command of the concept.
Three rules keep redoing honest:
- Space it. Redo a missed question only after enough time has passed for the specific answer to fade, typically a week or more. A next-day redo mostly measures short-term memory. The logic is the same as spaced repetition and active recall.
- Redo concept gaps, not slips. A calculation slip does not need the same question again. It needs a better routine applied to new questions.
- Explain before you answer. Before choosing, state why each wrong option is wrong. If you can do that, the concept is yours. If you can only find the right answer, it is not yet.
The stronger test is a new question on the same concept. If your log says "modified duration, price change estimate", real verification is a question you have never seen that tests the same idea. Passing that is evidence. Passing a repeat is, at best, a hint.
Reading Accuracy by Topic, and When to Act
An overall accuracy figure hides more than it reveals. Seventy percent across the bank can mean steady competence everywhere, or strong results in a few topics masking one that is collapsing. Track accuracy per topic, and per concept within a topic where you can.
Two rules make the figures meaningful. First, use recent data: your last few sessions in a topic, not a lifetime figure dragged down by your first week. Second, require a minimum sample. Accuracy on eight questions is noise; on 40 recent questions in one topic, it is a signal.
With that in place, simple thresholds turn tracking into decisions. The bands below are working triggers for your own study, not a prediction of any exam result:
| Recent topic accuracy | What it usually means | Action |
|---|---|---|
| Below 60% | The concept base is not there yet | Stop drilling. Rebuild from notes, then return to fresh questions. |
| 60% to 75% | Understanding is present but unstable | Targeted sets on the concepts your error log keeps naming. |
| Above 75% | The topic is working | Maintain with spaced review sets and move time to weaker topics. |
Watch the trend as much as the level. A topic rising from 55% to 68% over three weeks is responding. A topic stuck at 72% for a month despite heavy volume is telling you the method is wrong, not the effort.
Then weight the decision by exam importance. A weak result in a heavily weighted topic deserves your time before a weak result in a light one. The Level I topic overview shows where the weight sits.
The Anti-Pattern: Doing Questions to Feel Productive
There is a version of question practice that is mainly emotional. It happens late in the evening, after a long workday, when reading feels too hard and a quick set of 20 questions feels like study. You answer, glance at the score and close the tab.
This is not useless, but it is close. Without classification and a logged cause, the misses from that session are gone. You will make most of them again.
The signs are recognizable: sessions measured in questions rather than resolved errors, explanations skimmed for the right letter only, and a quiet preference for topics where you already score well. That last one is the most expensive. Practicing strengths feels good because the accuracy is high, and it is exactly where the next point is hardest to find.
A question bank is a diagnostic instrument. Used as a treadmill, it produces motion and very little change.
The practical implication is simple. End each session by naming what you now know that you did not know at the start. If the answer is nothing specific, the session was volume.
A Better Measure of Progress
Replace questions completed with three measures: errors logged, errors resolved (verified on a new question after a gap), and per-topic accuracy trends. They are less satisfying than a counter. They are much closer to what a mock exam actually measures.
This also answers the question of how many practice questions you need. There is no correct total. The right number is enough to surface the errors in every topic, plus enough fresh questions to verify each fix. For many candidates, that is fewer than the counter suggests, and each one is worked harder.
Wall Street Uni's study engine keeps this loop running in the background: it tracks which topics your own results have verified and schedules reviews at expanding intervals. Whatever tool you use, the principle is the same. Do fewer questions, and make each one teach you something.