FirstTry

How we write and verify questions

Most practice sites tell you their questions are “expert-written” and leave it there. Here is exactly what we do instead, including the numbers, updated every time we deploy.

Where the bank stands right now

523

questions written

32

documentation-verified and live

491

held back from the public

32 questions have passed documentation verification and are the ones you can practise. The rest are written but held back: unverified, awaiting a human, or blocked because their source did not hold up. We would rather show you fewer questions we can stand behind than a big number we cannot.

1. Written from published objectives, never from exam content

Every question is original. We write each one from the exam objectives the certifying body publishes openly: CompTIA’s objectives documents, Microsoft Learn skills outlines, AWS exam guides, Cisco blueprints. We do not possess, reproduce, or paraphrase real exam content, and we do not buy question sets from anyone. Using leaked exam content can get your certification revoked, and it cannot prepare you for the performance-based items that decide real exams.

2. Checked against a ten-point rubric

Before a question is even a candidate, it has to pass ten checks. These catch the failures that make practice questions worthless: an answer you can guess from its length, a distractor nobody would pick, an explanation that restates the question.

  1. 1Maps to exactly one published sub-objective
  2. 2Original; never derived from a real exam item
  3. 3Scenario-based where it should be (bank target: 60%+ scenario)
  4. 4Exactly one defensible answer (or exact multi set)
  5. 5Every distractor plausible to a partially prepared candidate
  6. 6No clueing via length, grammar, or specificity
  7. 7Three-tier explanation complete and substantive
  8. 8Doc link is a real, current, authoritative page
  9. 9Difficulty rated 1-3 (bank target 25/50/25)
  10. 10Current exam version content; zero retired material

3. Verified against the vendor's own documentation

This is the part nobody else does, and it is the whole point. A question passing an internal rubric can still be confidently wrong. So every question is checked against an external authoritative source, its cited documentation, and assigned one of four states:

Doc verified

The source supports the answer unambiguously and contradicts no distractor. These are the only questions served to you without a human also approving them.

Needs human

The source is ambiguous, partly disagrees, or no adequate source was found. Held back until a human reviews it.

Unsourced

The cited link is dead, generic, or off-topic, so there is no evidence at all. Never served, under any circumstances, even if a human approved it.

Contradicted

The source actually disagrees with the marked answer. The highest-priority defect, and never served.

The rule is strict on purpose: contradicted and unsourced questions can never reach you, and no human approval overrides that. When we mark a question doc-verified, we store the exact passage the judgement rests on, so the decision can be audited rather than trusted.

4. Reviewed by a human where flagged

A human reviews every question the pipeline flags as needing one, and spot-checks a random sample of the doc-verified questions to confirm the label holds up. We do not claim every question has had a line-by-line human read: that would be a lie at our current volume. We claim what is true, which is that nothing reaches you without either a documentation source supporting it or a human approving it, and nothing reaches you at all if its source disagrees.

5. If you find a mistake, tell us

The report link sits on every question. It captures the question, the answer you chose, and your comment, and it goes to the top of the review queue, because a real candidate who thinks a question is wrong is the best signal we get. Or email issues@firsttry.app. We would rather retire a question than keep a wrong one.

Verified questions by certification

Where documentation verification has reached, so far:

Documentation verification currently covers 6% of everything written. The rest is real work in progress, not a number we are hiding.