Building knowledge checks that actually stick

A typical training quiz: "Which of the following is a step in the closing procedure?" The frontliner re-reads the bullet points from the module and picks the matching answer. They scored 100%. They will forget all of it within a week. Recognition-based knowledge checks teach almost nothing. The checks that actually stick — that produce real retention months later — are built on retrieval and decision-making, not multiple-choice fact-matching. ## The recognition trap Multiple-choice questions where the right answer is visible in the module the learner just read are recognition tasks. Recognition is the easiest cognitive operation; passing a recognition check tells you almost nothing about whether the learner can use the information. A classic example: a food-safety module that ends with "At what temperature should chicken be stored?" with options 4°C, 7°C, 10°C, 15°C. The learner just read "chicken stored at or below 4°C" three paragraphs above. They tap 4°C. They forget the actual figure within days because they never had to retrieve it. Replace the question with: "You are putting away a delivery and notice the cooler is reading 8°C. The chicken pallet is in there. What is your first action?" Now the learner has to retrieve the temperature standard, recognize that 8°C exceeds it, and reason about the action. Same content, completely different cognitive operation. ## What separates a sticky check from a forgettable one Four properties. **1. Retrieval, not recognition.** The learner has to recall information from memory, not match it from the screen. Time delay between content and check helps; even a 30-second buffer matters. **2. Application, not definition.** The question asks "what would you do" or "what does this mean for the situation," not "what is X?" Application requires understanding; definition requires only memory. **3. Plausible distractors.** The wrong answers reflect realistic misconceptions, not nonsense. "4°C, 7°C, 10°C, 15°C" all sound like food-safety temperatures; the learner has to know which one. "4°C, 47°C, 412°C, 0K" gives the answer away. **4. Feedback that teaches.** A wrong answer triggers a brief explanation of why the right answer is right and why the learner's choice was tempting but mistaken. The feedback is the second teaching moment, often more memorable than the original content. ## What to avoid Four failure patterns common in homemade training quizzes. **Trick questions.** Negative phrasing ("Which of the following is NOT a step?"), double-negatives, options that are technically correct on a technicality. These test test-taking skill, not the underlying knowledge. **Long stems.** A question that takes 90 seconds to read is not a knowledge check; it is a reading comprehension test. Keep the stem short. **True/false on consequential decisions.** True/false halves the cognitive work and teaches almost nothing. For shift-floor decisions, use scenario or multiple-choice with realistic distractors. **Score-as-grade.** Treating the check as a pass/fail exam with 80% threshold creates anxiety without learning. The check is a learning opportunity; the score is informational, not punitive. ## Where in the course they belong Three placements for knowledge checks. **End of module (immediate).** Reinforces what was just covered. Useful but produces only short-term retention if used alone. **Spaced retrieval (1, 3, 7, 21 days later).** The version that produces durable retention. The 21-day check on a critical compliance fact is what makes the fact stick six months later. (See the spaced-retrieval guide for the mechanism.) **Pre-module (priming).** A check before the content, to surface what the learner already knows or thinks they know. Generates a productive mental gap that the module then fills. Underused but powerful. Aristotl's pedagogy approach combines all three placements automatically — modules end with knowledge checks, spaced-retrieval checks fire on schedule, and certain modules use pre-module priming for content where misconceptions are common. ## What the data tells you Knowledge-check data is diagnostic. Three patterns to watch for: 1. **Network-wide miss on the same question.** The question is unfair, the content is unclear, or the right answer is wrong. Investigate the content. 2. **Location-specific miss patterns.** One location keeps failing the same check. The local manager is teaching it wrong, or the location-specific context is different. Investigate the location. 3. **Drift over time.** A check that has been at 90% correct slips to 70% over months. Either the content has aged out, the workforce has changed, or the topic needs refresh. Investigate the trend. The checks are not just for the learner. They are also a continuous quality signal for the L&D team about which training is actually landing.

Ready to put this into practice?