# Graded quizzes: answer first, then reveal, with a 90% mastery gate (Academy)
Harvested 30 Oct 2026 for the £49.99/month /learn/ Academy grade ladder (the 22:21 brief). 'pass mark', 'retrieval practice' and 'mastery' had 0 skill hits before.

## What the evidence states (sources read 30 Oct 2026)
- **Testing beats re-reading.** Adesope, Trevisan & Sundararajan (2017), a meta-analysis of 217 retrieval-practice studies, summarised by Yana Weinstein (The Learning Scientists): the benefit over restudy was **stronger when the gap between study and test was over 1 day**, and the result was judged "resistant to the file drawer problem".
- **Feedback adds a bit, and it has to come after the attempt.** The same summary: retrieval practice with feedback was "only slightly more effective" than without. Rowland (2014, Psychological Bulletin): the testing effect holds without feedback but is enhanced by it, **even when the answer was already right**. Yang et al. (2020): "Corrective feedback induces greater reexposure and larger learning gains." So the brief's "answer, then Reveal" order is the evidence-backed order. Showing the answer before the attempt turns the quiz back into re-reading.
- **Mastery thresholds.** Mastery learning (Bloom) sets the bar at 80–90% on a test, with re-teaching and retesting before moving on (Winget 2022, a practical review of mastery learning, PMC). Univ. of Nebraska teaching centre: 80% is the mastery level for a quiz on a single standard. Scholario (2026): "A retake should show new learning, not just familiarity with the same questions."
- **Format.** Smith & Karpicke (2014, Memory) compared short-answer, multiple-choice and hybrid retrieval. Mixing in some typed or short answers is worth it at higher grades.

## How to do it
1. **One attempt, then lock.** Record the member's answer before Reveal works. Once revealed, that question scores as attempted and the answer can't be changed. Otherwise "reveal, then answer" games the 90% gate.
2. **Reveal = answer + one-line why + source link.** Keep the explanation to 1–3 sentences and quote the record verbatim with its catalogue ID. Reveal on correct answers too (Rowland: the gain still applies).
3. **90% gate with a fresh retake.** At 10 questions, 90% allows one miss; at 20, two. Build each grade from a pool **about 2× the quiz length** and draw a new mix on retake, so a pass shows learning and not memory of the last attempt. Retake after a short cooldown (for example, the next day). The over-1-day benefit is the reason.
4. **Space it: carry-forward items.** Put 2–3 questions from earlier grades into each later quiz (the spacing gain). Count them toward the score.
5. **Harder by grade, by kind of question and not just by obscurity:** recall (dates, places) → read a primary-source extract and answer → compare two records → the open questions. The final grade has no single right answer: score it "attempted with a cited record" and show "what record would settle it" on Reveal. Never present an open point as fact; label it "our case, prospective, not a finding".
6. **Measure:** first-attempt pass rate per grade (aim for roughly 40–70%; near 100% means too easy, under 20% means a wall), the questions most often missed (rewrite or re-teach them), retakes needed per pass, and churn after a failed attempt.

## How we apply it to Cedwyn
- The grades build (a worker) should store `answered_at` before `revealed_at` per question, draw from a pool larger than the quiz, and add a 24 h retake cooldown. Both are cheap in D1.
- Grade 1 as a free preview lets visitors feel the answer → reveal loop before paying £49.99. Report its first-attempt pass rate weekly.
- Every Reveal links to the evidence page. That also feeds internal links and time on site.
- Add a "Most-missed questions" list for each grade to the growth report notes. Missed questions point to unclear pages on the site.

## Sources
- learningscientists.org (Adesope et al. 2017, 217 studies; read)
- curriculumteamleader.wordpress.com (Rowland 2014; Yang et al. 2020; read)
- teaching.unl.edu (80% mastery; search snippet, not opened)
- scholario.digital (retakes; snippet)
- Winget & Persky 2022, "A Practical Review of Mastery Learning", PMC (snippet; the page returned 403)