How this question bank is actually built
Most exam-prep sites tell you their questions are high quality. Here is the method, the measurements, and the parts we are not happy with — computed from the live bank each time this page loads, not typed into a marketing paragraph.
What is in the bank
Counted from the database as you loaded this page.
| Trade | Active questions | Bilingual | Per-distractor breakdown |
|---|---|---|---|
| 310S | 8,251 | 8,251 | 8,112 |
| 310B | 3,397 | 3,397 | 3,129 |
Every question, in both languages
All 11,648 of our 11,648 active questions carry a full Arabic stem, Arabic options and an Arabic explanation written to the same depth as the English. Not an interface translation with an English bank behind it. This is the single hardest thing here to copy, and it is the reason we exist.
The answer keys were independently re-checked
Every question in the 8,112-question Automotive Service Technician corpus was answered blind — with no key and no explanation shown — and the result compared to the stored key. Blind agreement was 86.97%. Every disagreement was re-examined; in 507 cases the blind answer was wrong and the stored key was right, which is the main lesson of the exercise.
186 answer keys were corrected. 123 remained genuinely uncertain and were deliberately left alone rather than changed on a machine's opinion — those are still open, and they are excluded from everything we publish publicly.
What that audit does not cover
Read this scope carefully: that audit covers the 310S corpus only. It says nothing about the 310B bank. Anyone claiming every answer in every trade is verified — including us — would be overstating it.
What we have not fixed
The honest part. If you only read one section, read this one.
On 310S, 8,112 of 8,251 questions explain why each wrong option is wrong, individually. On 310B, that number is 3,129 of 3,397. The remaining 310B explanations tell you why the right answer is right without walking through every wrong option. That is a real gap in the product, and it is the difference between the two banks.
The 123 uncertain answer keys above are still uncertain. We publish the list rather than quietly picking one.
The 310B bank has not had the blind answer-key audit that 310S had.
Difficulty, and what it is not
Every question carries a 1-5 difficulty rating that genuinely drives spaced repetition and adaptive ordering. It came from a model-assisted content audit against a rubric — not from a human expert review, and not from learner response data.
Level 1
298
Level 2
3,655
Level 3
5,696
Level 4
1,859
Level 5
140
Specifically: we do not calibrate difficulty from learner response data, the way a psychometrician would. The columns for it exist in our schema and the estimator is written, but they are populated on 0 of 11,648 questions, because real calibration needs answer volume we do not have yet. Any site claiming statistically calibrated questions without that data is claiming something it could not have done.
Built to the published exam weighting
Questions are mapped to the Major Work Activities Red Seal publishes, and mock exams are drawn to the published percentages rather than at random. We show that arithmetic openly, including the part where Red Seal publishes percentages and a total but no per-topic question count.
Check it yourself
The numbers above are computed live. The audit method, the corrections and the list of still-uncertain keys are written up in the repository rather than summarised here, so a claim on this page can be traced to the artifact behind it.
Judge the bank on the questions
Try one and read the explanation. That is the honest test of any of this.