How scores are measured

The Claude Certified Architect – Foundations exam is 60 questions in 120 minutes with a pass mark of 720 out of 1000. It tests judgment in scenarios, not recall of terms. The app is built so that its numbers mean something on exam day: a score cannot be pushed up by repeating the easy activities, old work fades, and nothing says you are ready until you have shown judgment on every task and passed a timed mock with a margin.

Three numbers run the app, and each one answers a different question:

• Mastery, per task, 0 to 100: how well do I know this task, on the evidence so far?
• Readiness, for the whole exam, 0 to 100: how would I do tomorrow, weighted the way the exam is?
• Ready, a yes or no: have I cleared every bar that predicts a pass?

Everything else on screen (completion pills, streaks, XP, badges) is about effort and coverage, not about the score. This document explains all of it with the real formulas.

The exam has 30 tasks numbered 1.1 to 5.6. Each one has its own mastery score, computed from every answer you have given on that task. Mastery is never stored and never adjusted by hand: it is recomputed from the answer history every time it is shown, so every screen agrees.

Every answer gets a weight

An answer is a result with a value (1 = right, 0 = wrong, Explain This can be partial) and a weight. The weight says how much the answer tells us:

Difficulty multiplies the weight: easy ×1, medium ×1.25, hard ×1.5.

A wrong answer weighs more when it should worry you. Missing while you said you were Certain multiplies the miss by 1.75 (Fairly sure ×1.25, Guess ×1). Missing an item you had already missed before multiplies it by 1.5 again. A confident repeat miss is the most expensive answer in the app.

Old answers fade

Every result loses half its weight every 14 days. A card you got right a month ago still counts, but at a quarter of its original weight. This is why a task you have not touched in weeks slowly slides: the evidence that you knew it is getting old. A short session brings it back.

The score

Mastery = (sum of weight × value) ÷ (sum of weights), shown 0 to 100, then adjusted by two guards:

1. Evidence floor. Until the faded weight behind a task reaches 8, the score is multiplied by (weight ÷ 8). About eight easy flashcards, four medium quiz questions or three scenarios clear the floor. One lucky card cannot show 100. A task that shows 22 with two right answers is not 22% right; it is right so far with little evidence.
2. Depth cap. Recall drills alone (flashcards, vocabulary, fill in the blank) can take a task to 70. Matching and drag lift the cap to 85. Only judgment activities (quiz, scenario, Explain This, mock) can take a task to 100. The task page tells you which cap applies and what unlocks the rest.

Worked examples

Completion pills are not mastery

On a task page every activity row carries a pill: not started, 4 of 14 cards seen, or Done · all 14 cards. That is coverage: have you touched every item in that activity at least once. Mastery is correctness. You can finish every activity and sit at 55 if you missed a lot, and you can reach 100 on a task without finishing its flashcards if your quiz, scenario and Explain answers are strong. Use the pills to see what you have not tried; use mastery to see how well it went.

A note on sessions: flashcards are spaced repetition. A card you miss comes back a few cards later and due cards return across sessions, so a session of 19 cards may cover 8 new cards and 11 repeats. The pill counts distinct cards.

The exam has five domains with fixed shares of the questions: D1 27% · D2 18% · D3 20% · D4 20% · D5 15%.

A domain score is the plain average of its tasks' mastery. A task you have never practised counts as 0, so an untouched task drags its domain down until you open it. That is deliberate: the exam will ask about it whether or not you studied it.

Domain scores appear on Home, on the Readiness page and on the Mock page next to your last mock's percentage in that domain, so you can compare "what the app thinks" with "what a paper showed".

Readiness is the domain scores weighted by the exam's shares: readiness = (D1 × 27 + D2 × 18 + D3 × 20 + D4 × 20 + D5 × 15) ÷ 100. It is the app's estimate of how you would do on the exam tomorrow, with Domain 1 counting for more because the exam gives it more questions.

Example: domains at 80 / 70 / 60 / 50 / 40 give (80×27 + 70×18 + 60×20 + 50×20 + 40×15) ÷ 100 = 62.

The target and the forecast

The target is readiness 85. The Readiness page draws the last 14 days of readiness (recomputed as of the end of each day) and fits a straight line through it. If the line rises, the app says when it crosses 85 at your current pace; if it is flat, it says so and suggests a Daily Session. The forecast is a straight line, not a promise: it moves every day with your work.

The four Ready rules

Readiness alone is not enough. Ready is a verdict that needs all four of these at once:

1. Readiness at least 85.
2. No task below 60. One weak task can cost several questions on exam day, so every task must clear the floor.
3. At least 5 judgment results per task. Quiz, scenario, Explain This or mock answers. Recall drills do not count here, because the exam does not ask you to recall; it asks you to decide.
4. One mock exam scored 760 or more. The pass mark is 720; the app wants a margin on a timed paper before it calls you ready.

The Readiness page lists each rule as pass, fail or not yet, with the tasks that block it, so there is never a mystery about what to do next.

An item is an open mistake when your most recent answer on it was wrong. It stays open until you answer that same item right. The Mistake Journal lists open mistakes newest first, with the ones you missed while Certain marked, because those are the beliefs most worth correcting.

How a mistake affects the score

A miss is a result with value 0 and extra weight (see section 1), so it pulls the task's mastery down, and through the domain, readiness. Nothing is ever deleted: fixing a mistake does not erase the miss. Instead, a fresh right answer on the same item is added, and because it is newer and the miss keeps fading, the right answer outweighs it more every day.

Example on one task: three medium quiz questions right and one missed while Certain, three days ago, gives mastery 63. Answer that missed item right today and the task reads 70. The miss is still in the history; it just matters less.

Fix potential

The Journal and the Mistakes page show fix potential: what readiness would be if every open mistake were answered right now. It is the single fastest way to raise readiness, because each fix lands on a task that is already below where it should be. Fixing a mistake also pays 15 XP.

Drilling mistakes

"Drill my mistakes" serves open items first, hardest and most confident misses before the rest. A right answer closes the item; a second miss on it counts as a repeat miss and weighs ×1.5.

A mock is the full paper: 60 questions drawn by the exam's domain shares (D1 16, D2 11, D3 12, D4 12, D5 9), mixed difficulty, 120 minutes on a timer that does not pause. No feedback until you finish.

Scoring

Scaled score = right ÷ 60 × 1000, the way the real exam reports. 720 passes. The Ready rule asks for 760, a margin above the pass mark, because a timed paper on a good day should clear the line with room to spare.

What a mock does to mastery

Every mock answer is also a result on its task at ×3 weight, so a mock moves mastery across all 30 tasks at once. Blank questions count as wrong. If you press Finish after two questions, 58 misses land on your tasks and every domain drops. Leaving the page without pressing Finish records nothing, and the exam waits for you. If you did finish by mistake, each attempt on the Mock page has a Delete control (two taps) that removes the attempt, its answers and its session, and your scores return to where they were. XP already paid for the attempt stays.

Pacing

Mastery is a daily measure; the mock is the dress rehearsal. One every week or two is right. Taking one every day turns the paper into a drill and tells you less each time, since the bank avoids repeating your last mock only when it can.

These measure work, not knowledge, and they never feed the score. Readiness, mastery and Ready are computed from answers only.

• Daily goal and streak. Minutes of active study per local day; the default goal is 60 minutes and the minimum you can set is 30. Meeting the goal keeps the streak alive and pays 25 XP once a day. A freeze is earned on day 7 and spends itself on the first missed day.
• XP. 10 per right answer, a share for a wrong one (never less than 2), more for an Explain This graded well, plus bonuses: Daily Session 50, mistake fixed 15, Explain at 80+ 25, mock finished 200, mock at 760+ 500, domain boss 300. XP is lifetime and never decreases.
• Levels. Prompt Apprentice at 0, Tool Builder at 1,500, Orchestrator at 5,000, Systems Architect at 11,000, Principal Architect at 20,000 XP.
• Badges. Earned once and dated to the moment you crossed the line. Mastery medals (Bronze at 60, Gold at 85) fade with a Refresh link if the task slips under the line later; they are not taken away.
• Leagues. Weekly XP against learners who started the same week, capped at 1,500 a day so steady days beat a marathon.

If you want to know how you are doing, read mastery and readiness. If you want to know whether you are showing up, read the streak.

Hover any mastery number for its evidence: "72 · 14 answers · full evidence · last 10-09-2026".