The Question Quality report uses psychometric analysis to help you identify which questions are working well and which need improvement. It analyzes actual quiz attempt data to measure how each question performs.

This report is also available on the Reports screen of the front-end Teacher Dashboard, including its CSV export, when the Educator add-on is on version 3.1 or higher.

Accessing the Report

Navigate to PressPrimer Quiz > Reports and click the Question Quality card. Select a quiz or question bank to analyze, and the report will display metrics for each question. More information about the page and metrics is available on the report page by expanding the Understanding Item Analysis section.

Questions need at least 5 attempts before analysis is available. With fewer attempts, results aren’t statistically reliable.

Understanding the Metrics

Difficulty Index

The difficulty index (p-value) shows what percentage of learners answered correctly. Despite the name, higher values mean easier questions:

• 0.80 – 1.00: Very easy (most learners get it right)
• 0.40 – 0.79: Moderate difficulty (ideal range)
• 0.20 – 0.39: Difficult
• Below 0.20: Very difficult (most learners get it wrong)

Questions outside the 0.20-0.80 range may be too easy or too hard to effectively assess learning.

Discrimination Index

The discrimination index measures how well a question differentiates between learners who understand the material and those who don’t. It compares how the top 27% of performers did on each question versus the bottom 27%.

• 0.40 or higher: Excellent—high performers consistently get it right, low performers don’t
• 0.25 – 0.39: Good discrimination
• 0.15 – 0.24: Marginal—consider revising
• Below 0.15: Poor—the question doesn’t distinguish ability levels
• Negative values: Problematic—low performers are doing better than high performers, which usually indicates a confusing question or incorrect answer key

Point-Biserial Correlation

This metric shows how strongly getting a question right correlates with overall quiz performance. Values above 0.30 indicate the question aligns well with what the quiz measures. Low or negative values suggest the question may be testing something different from the rest of the quiz.

Distractor Efficiency

For multiple choice questions, this measures whether the wrong answer options (distractors) are doing their job. A distractor is “functioning” if at least 5% of learners select it. If a distractor is never chosen, it’s not helping assess understanding and should be replaced with a more plausible option.

100% efficiency means all distractors are being selected by some learners. Lower percentages indicate some wrong answers are obviously wrong and aren’t fooling anyone.

Question Classifications

Each question receives a classification based on its metrics:

• Good: Metrics are within acceptable ranges. No changes needed.
• Review: One or more metrics are slightly outside ideal ranges. Worth examining but not urgent.
• Revise: Metrics indicate significant issues—too easy, too hard, or poor discrimination. Should be rewritten or replaced.
• Problem: Negative discrimination detected. High performers are getting this wrong while low performers get it right. Check for confusing wording, trick questions, or an incorrect answer key.

Measured Difficulty on the Questions Screen

The difficulty index described above also feeds a Measured column on the Questions screen. Once a question has enough attempt data (at least 20 scored attempts), its measured difficulty appears next to the difficulty its author assigned. When the two disagree, the question is flagged, and a Mismatched filter shows every flagged question at once, so you can review ratings that real-world data contradicts. Questions without enough data yet simply show no measurement and continue to rely on the author’s rating.

For dynamic quizzes, a Use Measured Difficulty for Rule Matching option in the quiz editor makes difficulty-based rules match on the measured value where one exists, falling back to the author’s rating where it does not. This is useful when a quiz should genuinely draw “hard” questions by how students actually perform, not just by how questions were labeled. New sites work fine without it: until questions accumulate data, everything behaves exactly as authored.

Using the Report to Improve Questions

Start by filtering for questions marked “Revise” or “Problem”—these need the most attention. For each flagged question:

• Very easy questions (high difficulty index): Make the question more challenging or replace distractors with more plausible options.
• Very hard questions (low difficulty index): Check if the content was adequately covered. Consider whether the question is too complex or tests obscure details.
• Poor discrimination: The question may be confusingly worded, or the topic may not have been taught effectively to all learners.
• Negative discrimination: Immediately review the question. There’s likely an issue with how it’s written or the marked correct answer may actually be wrong.
• Low distractor efficiency: Replace obviously-wrong distractors with options that represent common misconceptions.

Run the report periodically as more attempts accumulate. Metrics become more reliable with larger sample sizes.