What Writing Cloud Certification Questions Taught Me About Good Practice Exams

Writing a good question is harder than it looks

For more than two years, I’ve contributed as a subject matter expert to question development for multiple AWS certifications.

The biggest lesson has been that writing a good multiple-choice question is much harder than it looks.

You aren’t simply writing one true statement and surrounding it with three false ones. You’re trying to find out whether someone understands a specific concept well enough to apply it.

That requires care at every stage.

A question needs a reason to exist

Before writing a scenario, you need to know what you’re assessing.

What skill or knowledge does this question test? Where does that fit in the exam outline? Why should a candidate at this level be expected to know it?

Without those answers, it’s easy to drift into trivia: facts that are technically correct but don’t tell you much about someone’s ability to do the work.

The exam outline matters because it defines the scope of the assessment. A foundational candidate and a professional-level candidate should not be expected to make the same decisions with the same depth of knowledge.

Good questions respect that difference.

The wrong answers require real work

A multiple-choice question has three basic parts:

• The stem: the question or scenario.
• The key: the correct answer.
• The distractors: the incorrect alternatives.

Distractors are where many weak questions fall apart.

If three options are obviously unsuitable, the learner can answer correctly without understanding the concept being tested.

Useful distractors reflect plausible mistakes. An option might solve part of the problem while missing a stated requirement. Another might work in a different situation but fail under the constraints in this one.

The question should provide enough information to explain why one answer is best. Difficulty should come from the decision, without making candidates guess what the writer meant.

Wording can accidentally reveal the answer

A learner shouldn’t be able to pass by choosing the longest option.

Other clues can be just as revealing: wording copied from the scenario into the correct answer, grammatical mismatches, or one unusually detailed option surrounded by vague alternatives.

Reviewing a question means looking for those clues as well as checking technical accuracy.

The objective isn’t to make every answer exactly the same length. It’s to ensure the formatting and language don’t do the thinking for the candidate.

We also look for teaching embedded in the stem or answer choices. If a question defines the exact behavior the candidate is meant to know, it can turn a knowledge assessment into a reading exercise. Give the candidate the scenario’s requirements; save the lesson about the tested concept for the explanation.

The aim is to distinguish someone with the minimum knowledge and judgment expected at that level from someone who can only eliminate obviously unrelated options. Careful reading should help a prepared candidate apply their knowledge. It should not replace the need for that knowledge.

The explanation is part of the product

For practice material, “B is correct” isn’t enough.

A useful explanation identifies the requirement that makes B correct, connects that requirement to the relevant service behavior, and explains why the other options fall short.

It should also point to authoritative documentation so the learner can investigate further.

That turns a wrong answer into something valuable: a specific gap the learner can understand and address.

Review matters because writers have blind spots

When you write a question, you already know what you intended it to mean.

Someone else may read it differently, and reasonably so.

My experience contributing to certification development has reinforced the value of multiple perspectives. Technical scrutiny, editorial feedback, and discussion among subject matter experts can reveal ambiguity that the original writer completely missed.

Sometimes the right response is to revise the question. Sometimes it’s to discard it.

The time spent writing a question doesn’t make it a good question.

Publication is not the finish line

Cloud services change. Documentation changes. A once-clear question can become outdated or develop more than one defensible answer.

Learner feedback and response patterns can also reveal problems worth investigating. An unexpected pattern doesn’t automatically mean the question is flawed, but it gives you a reason to look more closely.

A question bank needs maintenance, not just expansion.

How this shapes our approach

These are the principles informing the practice platform we’re building: alignment with the exam outline, meaningful scenarios, plausible alternatives, useful explanations, and ongoing review.

My background informs that approach. It does not make our material official, imply endorsement by a certification provider, or mean we reproduce their questions.

The aim is to help learners build the understanding an exam is intended to assess.

A good practice question earns its place by teaching you something about your readiness. A high score is only useful if you can trust how you earned it.