Back to Articles
    Teaching
    6 min read
    August 23, 2026

    Moodle Quiz Discrimination and Facility Indices: Check the Key Before You Rewrite

    By Milan KeldrowskiUpdated October 26, 2026

    On an ear-training quiz, a low facility figure usually gets blamed on a hard interval, and a negative discrimination figure on a confusing question. The reflex is to open the editor and start rewording. Both guesses can be wrong. With audio questions the answer key can break in one more way: the sound file and the marked answer can disagree. Reword before checking that, and one wrongly marked question becomes two.

    What the statistics report can and cannot tell you

    Open Moodle's Quiz statistics report once attempts are graded, to decide whether each question deserves to count towards the score. It shows how a question behaved for this group, with this marking, on this date. Whether the wording is good in itself is a different question.

    Low facility does not prove a question is bad. A hard question may test an advanced skill, and an easy one may check a basic skill everyone needs. Negative discrimination is more serious. It means students who did well on the rest of the quiz did worse on this question.

    Look at three things together: the report figures, the answers students actually gave, and what the question is meant to test. Moodle's Quiz statistics report documentation explains the fields and how they are calculated. The report tells you where to look. It cannot tell you whether the teaching was fair or whether a question should count in an important grade.

    Facility shows how hard a question was, nothing more

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

    Facility Index = mean question mark / maximum question mark × 100

    A Facility Index of 80% means the group earned 80% of the available marks on average. It says nothing about how well the question was written. It may reflect good preparation, an obvious answer, a clue in an earlier question, generous part marks or an undemanding task.

    As a rough guide, 0-30% is hard, 31-70% middle, 71-90% easy and 91-100% almost everyone got it. Moodle does not attach pass or fail to these bands. They just help you read the numbers. An advanced question may rightly fall below 30%, and a check on something essential may belong above 90%. Ask whether the difficulty fits what the question is for.

    Marks matter here. A two-mark question with part marks can show middle facility even when most students picked the same wrong option and earned marks elsewhere in their answer. Look at the spread of marks and at individual answers before you decide students half understood. Facility tells you how many marks were earned, not how.

    Negative discrimination: check the marking first

    Moodle's discrimination index asks whether doing well on one question goes with doing well on the whole quiz. A positive value means students with higher overall scores tended to do better on this question. A negative value means the opposite.

    Do not rewrite the question first. Check the answer key, the marking rule and the feedback students saw. A wrong key, an imported mark setting, a penalty or a question version that does not match the course material can all turn the result upside down. For numerical questions, check the accepted tolerances and units. For listening questions, play the audio and name what you actually hear before you trust the key.

    Here is a made-up interval quiz with 24 students. Question 7 plays a rising interval from C4 and offers four options: perfect fifth, minor sixth, major sixth, minor seventh. Facility is 75%, but discrimination is negative. The answers show that stronger students chose minor sixth, while weaker ones leaned towards the keyed major sixth. Playing the file settles it. The second note is A♭4, eight semitones above C4, so the interval is a minor sixth. The key had been copied from a similar question built on A4, nine semitones up. The best listeners were marked wrong for hearing correctly. Moodle works out its index from all the answers, so a quick split of this small group will not reproduce the exact figure.

    Negative discrimination can also mean the question tests something other than what you meant. A chord-quality question played on a heavily distorted guitar may reward students who know that sound, more than students who can hear the third and the seventh. Then the question is not testing the right skill, and being harder is no reason to keep it.

    Read the two figures together

    Look at either figure alone and you get false alarms: a hard question looks broken, an easy one looks useless. Reading the two together gives a fairer first view.

    The bands below are a working guide for after the quiz. The discrimination ranges are local rules of thumb; Moodle does not publish quality labels like these. Be stricter when scores decide progression, certificates or external reports.

    Report pattern Likely reading Post-assessment decision
    Any Facility Index, discrimination below 0 The item conflicts with the quiz pattern. Check the key, the marking, what the question tests and the actual answers before reusing it or changing any scores.
    Facility 0-30%, discrimination of 0.15 or higher Difficult, but may distinguish attainment. Keep if the learning outcome requires this challenge; inspect whether distractors reflect teachable misconceptions.
    Facility 31-70%, discrimination of 0.15 or higher Middle-range difficulty with usable separation. Keep it for now and watch it in the next comparable group.
    Facility 71-90%, discrimination of 0.15 or higher Easy item that still tracks wider performance. Retain when it checks foundational knowledge; do not expect strong ranking of top performers.
    Facility 91-100%, discrimination of 0.15 or higher Near-universal success. Keep only when universal mastery is the intended signal; otherwise it adds little evidence to a scored quiz.
    Any Facility Index, discrimination from 0 to 0.14 Weak separation without obvious reversal. Check group size and score spread first; a rewrite may be justified, but the figure alone does not prove it.

    A second made-up question shows why low facility is not a fault. It plays a seventh chord in close position and asks for its quality: major seventh, dominant seventh, minor seventh or half-diminished. Facility is 24% and discrimination 0.34. Students who did better overall are more likely to get it right. The right answer is major seventh. Weaker students pick dominant seventh, a sensible wrong answer, because it has the same major third. The question is hard, but it may be doing its job. The real question is whether hearing seventh-chord qualities was taught and expected at this stage. If it was, keep it and check the explanation students see afterwards. If it was not, it may be testing material too early.

    Positive values also come in grades. Teachers have long used rough bands made popular by Robert Ebel: 0.40 or more is very good, 0.30 to 0.39 reasonably good, 0.20 to 0.29 borderline and usually worth revising, and below 0.20 poor. They were written for larger, more even groups than most Moodle classes. Treat them as a reading aid, like the facility bands above, not as a reason to delete questions.

    Small groups make every figure less reliable

    Statistics are fragile in small groups, or when almost everyone gets a similar total. One answer can noticeably change both figures. A negative result in a small class deserves attention, but look at the answers before you decide the question is always bad.

    A narrow range of scores causes a similar problem. If nearly everyone scores highly, good questions can show weak discrimination, because the quiz shows little difference between students. The same happens when the whole quiz is too hard. A question cannot separate levels that the quiz has not revealed.

    Do not combine attempts just to steady the numbers. Mixing different quiz versions, dates, teachers, time limits, languages or marking rules can hide the cause. Combine only runs that are really comparable, and write down what changed. For a small group, "check again after the next similar run" is fairer than forcing a verdict.

    Check the evidence before you change a question

    A careful review after the quiz, with the marked attempts kept, protects students and your question bank, and gives every change a reason.

    1. Check what the report covers. Compare completed attempts, attempt settings, question versions, maximum marks and any regrading with the quiz you are reviewing.
    2. Check the marking. Confirm the correct answer, part marks, penalties, numerical tolerances, units and any manual grading.
    3. Read some real attempts. Compare stronger and weaker students. Look for two reasonable readings of the question, accidental clues, or course terms that differ from the key.
    4. Check what the question tests. Match it to one stated learning goal. If a correct answer depends on an unstated assumption or a skill you did not teach, fix that first and leave the wording for later.
    5. Write down what you did and check again. Note whether the question was kept, retired, given a new key, regraded under an approved policy or revised for next time. Then look at the figures again after a similar run.

    Order matters. If you edit before checking the key, you can create a new version of the same marking error. If you retire every hard question, the quiz ends up testing only memory. Regrading needs separate approval from your school or institution: the report points to a problem but does not decide what is fair.

    A growing bank shows version and group effects

    As a Moodle question bank grows, a question does not have one fixed reputation. Its figures change with the order of the course, random question selection, whether feedback is shown and who the students are. A question that works well in practice may behave differently in a timed final, because the other questions change the overall scores.

    For each question you review, note its version, learning goal, the quiz it was in, the date, a description of the group, the Facility Index, the discrimination figure, your decision and the reason. This keeps a record and helps you tell a faulty question from a gap in teaching. If several well-matched questions get harder after a change to the course, the bank may be showing you something you stopped teaching.

    With a big bank, look for patterns by learning goal and question type rather than chasing one odd number. Weak discrimination across many scenario questions may mean a problem with how they are written. Negative values that appear after one import may mean a technical or keying problem. Fix the problem at the level where the pattern appears.

    What the next similar run should show

    Every question you changed comes with a prediction for the next similar run, and it only counts if you wrote it down first. A question with a corrected key should move to positive discrimination. A reworded question should keep its facility band while the wrong option it was revised for stops attracting students. A hard question you kept should still separate stronger students from the rest.

    If the same pattern comes back with a similar group, the wording was probably never the cause. Move on to what the question tests, the audio or the teaching. Negative discrimination always means "check this", and firm conclusions wait for a group with enough spread in scores to support them.

    Related Articles