Back to Articles
    Teaching
    7 min read
    August 23, 2026

    How to Interpret Moodle Quiz Statistics With Fewer Than 30 Learners

    By Kira DrenwickUpdated October 26, 2026

    A guitar theory class of twelve gets a quiz report that looks just as sure of itself as one for a course of ten thousand. Moodle prints a tidy Facility of 82% and a Discrimination of -0.2 either way. With a small class, those numbers are hints to look closer. Treat them as verdicts, and good questions get rewritten while real faults survive.

    A small class changes what quiz statistics can prove

    With 8, 15 or 29 completed attempts, Moodle quiz statistics tell you where to look, and not much more. That is a normal theory group at a guitar school: an evening class on building triads, or fifteen adult beginners learning the notes on the fretboard. A high Facility may just mean a few students already knew the material, like the two who learned piano as children. A negative Discrimination may come from one unusual student, an unmarked manual question or too little difference in scores to say anything reliable.

    So first sort the questions: which need checking today, and which just need another class. With fewer than 30 students, base that on four things together: the figure in the report, how much it could easily move, the actual attempts, and what the question is in the quiz for.

    Use the number of attempts in the report, not the number enrolled. Excluded or unfinished attempts, regrades and filters can all reduce the real count. Repeat attempts need care too: 20 attempts from 12 students are not 20 separate pieces of evidence.

    The two figures answer different questions

    The Moodle Quiz statistics report gives you figures for each question and leaves the judgement to you.

    The Facility index is the average mark on a question as a percentage of the maximum:

    Facility index = mean item mark / maximum item mark × 100

    For right-or-wrong questions, it is roughly the percentage who got it right. With part marks, it is the average share of marks earned. A Facility of 82% means the question was easy for this group. It does not mean 82% chose one correct option.

    The Discrimination index works like a correlation: did students who did well on the whole quiz also do well on this question? A positive value suggests the question separates stronger and weaker students in this set of attempts. A value near zero adds little. A negative value means you should check for a wrong answer key, an ambiguous question, a marking problem or content unrelated to the rest of the quiz.

    Neither figure tells you whether the question tests the right thing. A simple recall question that every prepared student gets right will discriminate weakly, simply because there is no difference left to find. A question with strong discrimination can still reward clues in the wording rather than the skill you meant to test.

    A single number hides how unsure it is

    Nothing magical happens at thirty students. But below that, one or two attempts can swing the figures sharply. The swing is biggest when nearly everyone scores very low or very high.

    How far discrimination could move

    You can work out a range around the Discrimination figure (treated as a correlation r). Moodle does not show it, but a spreadsheet can calculate it with the standard Fisher method:

    z = 0.5 × ln((1 + r) / (1-r))

    SE of z = 1 / √(n-3)

    Take z ± 1.96 × SE and convert both ends back with the hyperbolic tangent (TANH in a spreadsheet). You need more than three usable, independent attempts. The range shows how unsure the figure is. Then still check the marking, the answers and the purpose of the question before deciding anything.

    For a Discrimination of 0.30, the range is roughly -0.41 to 0.78 with 10 attempts, -0.16 to 0.66 with 20, and about -0.07 to 0.60 with 29. The figure is positive, but the real relationship could be slightly negative or clearly positive. Rewriting the question because 0.30 looks modest would be an overreaction.

    Treat a negative figure differently, and look at the attempts before calling the question a failure. One strong student who got it wrong, or one weak student who got unexpected part marks, gives you a case to check. In a theory quiz the cause is often dull. Say a short-answer question asks for the minor third above C and accepts Eb but rejects E♭ typed with a proper flat sign. That is a marking fault to fix today, and it tells you nothing yet about the question itself.

    Facility needs a different range

    For right-or-wrong questions, Facility is a simple proportion. With small groups, use the Wilson method for the range. The textbook formula p ± 1.96 × √(p(1-p)/n) gives silly results near 0% and 100%.

    Say 7 of 10 students correctly spell a D7 chord as D, F♯, A and C. Facility is 70%, but the likely range runs from 40% to 89%. With 20 attempts, the same 70% spans about 48% to 85%. The question could be fairly easy or something almost everyone knows. A small class cannot tell the two apart.

    Do not use this range for part marks, penalties or complex marking. Keep the individual marks instead. For most teachers, the spread of scores and the options students picked are more useful than a precise-looking range.

    Both methods are long established: Fisher's dates from 1915 and 1921, Wilson's from 1927, and in 1998 Agresti and Coull showed that Wilson's works well for small groups where the textbook formula does not. One practical point follows. Combine attempts from two classes only if the question version, marking and quiz settings were identical. Otherwise you are measuring two different questions.

    A decision table for small-class item reviews

    Use this once grading is complete. The cut-offs are working thresholds for a small class; the table exists to stop you treating every low Discrimination index as a rewrite request.

    Report pattern Likely interpretation under n < 30 Immediate action Evidence still needed
    Facility at 0% or 100%, discrimination unavailable or unstable Little or no score variation Check key, availability, wording, and whether intentionally diagnostic Larger cohort or another delivery if it must distinguish performance
    Mid-range Facility, low or negative discrimination Weak relationship, or one attempt driving it Inspect responses, marks, feedback, and outlying learners' score patterns Comparable repeat before rewriting for statistics alone
    Negative discrimination plus visible scoring anomaly Possibly miskeyed, ambiguously scored, or misaligned Correct defect, regrade where appropriate, document change Later confirmation that revised item behaves sensibly
    High or low Facility with clear instructional purpose Doing its assigned job rather than separating learners Retain if confirming mastery or early diagnosis Blueprint evidence, not stronger correlation
    Same warning in comparable deliveries Signal survived one cohort's noise Review wording, options, scoring logic, and outcome alignment Decision record showing what changed and why

    "Comparable" matters here. A different course level, a changed question bank or a new marking rule spoils the comparison, because the students and the quiz have both changed.

    Read facility next to the actual answers

    Facility counts marks. It says nothing about how students got them. A 50% Facility on a four-option multiple-choice question can mean half-understanding, a confusing wrong option, a broken image or guessing.

    The answers themselves tell these apart. If most wrong answers pick the same option, that often shows a predictable misconception, such as naming the relative major when the question asked for the relative minor. Answers scattered across all options point to unclear wording or a missing basic skill. If nearly everyone gets part marks on a multi-step question, it is measuring progress through steps, not one skill. Calling it "medium difficulty" loses that.

    Time matters too. Normal Facility with an unusually long answer time can mean the question is genuinely demanding, badly worded, has audio or images that do not load, or uses an unfamiliar format. Moodle shows you which question. The attempt review and the question text explain why.

    Keep the course plan in mind. A final check on a basic skill can rightly have high Facility. A demanding end-of-course question can rightly be harder, with mixed marks. Judge each question against its job. There is no ideal row every question should match.

    Manual marks, retries and variants change the count

    The statistics quietly assume every attempt was the same experience. In a real course that is rarely true. Questions still waiting for manual marking distort total scores and discrimination, so finish marking and any regrades before you treat the report as evidence about questions.

    A student who improves on a later attempt after feedback cannot be compared with a first attempt. The report still shows what it was set up to show, but you need to know which attempts it includes before you rely on it.

    Random questions and calculated variants need checking too. One row may combine versions with different numbers, wording or difficulty, and the combined Facility can hide one faulty version. Look at examples of each version, especially when a small-class warning comes with an odd score pattern.

    Do not treat blank answers as wrong without thinking. A blank can mean confusion, lack of time, a browser failure or a deliberate skip. Logs cannot prove what the student meant, but they can stop a technical interruption being mistaken for a hard question.

    Review in an order that keeps the evidence

    A fixed review order avoids two mistakes: deleting a question after one awkward class, and keeping a real fault because the class was small.

    1. Note the report settings, the number of usable attempts, the question version, whether marking is finished and which repeat attempts are included.
    2. Read Facility and Discrimination together, and work out a likely range where that makes sense.
    3. Look at unusual answers, blanks, part marks and the strongest and weakest students who move the figure.
    4. Compare the question with what it is meant to test, and check the key, the marking guide, the media and the feedback before changing what you teach.
    5. Label it: fix now, watch, keep on purpose, or wait for another similar class. Keep the label with the question.

    The label pays off months later, when someone else edits the quiz and needs to know whether a flagged question was a real fault or just one small class's wobble.

    Gather more evidence before decisions become permanent

    A class of under 30 can still reveal faults that need fixing straight away: a wrong key, a missing diagram, broken marking or wording that can clearly be read the wrong way. These are visible problems and need no statistics.

    The subtler judgements, about discrimination, difficulty and whether a question fits the quiz, work the other way round. If you cannot see a fault, the honest label is "watch" until a similar class sits the same version. Read like this, the report tells you which questions to open next. Deciding what is wrong inside them is still your job.

    Related Articles