Quant Interview Diagnostic Test
How to run a quant interview diagnostic test and turn the result into a focused prep plan.
Candidates who are unsure which quant interview topics need the most work.
What a diagnostic should measure
A useful diagnostic measures methods, not confidence. Score sample-space setup, counting, conditional probability, expected value, mental arithmetic, estimation, statistics intuition, and explanation as separate lines, because a single overall percentage tells you nothing about what to do on Monday. The output you want is an assignment - "two sessions on conditioning" - rather than a grade.
Build a mixed set
Twelve to fifteen questions is enough to expose the obvious gaps, provided the set spans methods rather than topics. A workable spread is one ordered-versus-unordered counting question, one sampling-without-replacement question, one complement question, one Bayes base-rate question, one linearity-of-expectation question, one stopping or recursion question, one variance or covariance question, two mental arithmetic tasks, one Fermi estimate, and one prompt you have to explain aloud. Allow roughly three minutes each, and look nothing up mid-set.
Score misses by cause
For each miss, tag the cause: setup, method choice, arithmetic, time pressure, communication, or review discipline. These demand completely different fixes. A wrong answer caused by bad arithmetic is repaired with ten minutes of daily drills; a wrong answer caused by choosing the wrong sample space is repaired by slowing down the first thirty seconds of every problem. Tagging by topic instead of by cause is the single most common way to waste a diagnostic.
Concrete example
Suppose you miss three questions: a dice-maximum problem, a Bayes base-rate problem, and a five-card-hand problem. By surface topic that is dice, Bayes, and cards, which implies three separate study blocks. Tagged by cause, all three turn out to be denominator errors - you counted ordered outcomes over an unordered sample space, and you used the prior population instead of the post-test population. That is one problem wearing three costumes, and a single repair block on sample-space discipline fixes all three. This is precisely why cause tagging beats topic tagging.
Turn results into the next week
Take the top two causes and schedule one repair block each. If conditioning is weak, work Bayes tables written in counts, cards without replacement, and first-step analysis. If communication is weak, redo problems you have already solved, out loud, and score only the explanation. Then retest cold three to seven days later using fresh questions on the same methods, because re-solving the original problems measures memory rather than repair.
Common mistakes
Diagnostics fail when they are too long, so fatigue becomes the variable you are measuring; too narrow, so a probability-only set cannot tell you that probability is actually your strength; or too easy, so acing it tells you nothing. The most expensive mistake is running the diagnostic and then ignoring the result because it conflicts with the plan you already wanted to follow. If you are not willing to let it choose your next block, do not spend the hour.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.