AWS exams use two question formats, but within those there are recognizable patterns. Knowing what kind of question you’re looking at tells you how to attack it, and saves the time you’d otherwise spend flailing.

The two formats

**Multiple choice** gives you four options and exactly one is correct. This is most of the exam.

**Multiple response** gives you five or more options and asks you to choose two or three. The question always tells you how many to pick.

Multiple response is scored all-or-nothing. Two of three correct earns you nothing. That’s worth knowing because it changes the math on guessing: if you’re confident about two and unsure between two others, take the time. Partial knowledge earns partial credit on your practice score and zero on the real thing.

The patterns that repeat

“Least operational overhead”

Somewhere in the question, usually the last sentence. The answer leans toward managed and serverless options, because you’re being asked to minimize the amount of stuff a human has to run.

**The trap:** a solution that’s technically simpler but requires you to manage servers. Also, don’t take this so far that you pick a managed service which doesn’t meet a stated technical requirement. Overhead is the tiebreaker, not the only criterion.

“Most cost-effective”

Compare total cost for the whole workload, not the sticker price of one component. The cheapest individual service frequently isn’t the cheapest design once you account for data transfer, idle capacity, or the operational burden.

**The trap:** an option that’s cheap but fails a requirement stated earlier. A cheap answer that doesn’t work isn’t cost-effective.

“Highly available”

Look for failure domains. Multiple Availability Zones, health checks, automatic recovery. Ask what happens when one component dies.

**The trap:** confusing availability with durability. They’re different properties and the exam tests whether you know it. Durability is about not losing data. Availability is about still being able to reach it.

“Minimal downtime” / migration

Check compatibility, replication approach, cutover, and validation. These questions usually reward the option that keeps the old thing running while the new thing catches up.

“Secure” / “least privilege”

Narrow the permissions, use temporary credentials over long-lived keys, use roles over users, encrypt in transit and at rest, keep things off the public internet.

**The trap:** an option that’s secure but breaks the application’s actual access needs.

A method that works on any of them

**1. Read the last sentence first.** On a long scenario, the final line usually tells you what’s being optimized for. Knowing you’re solving for cost versus availability changes how you read the preceding four paragraphs.

**2. Find the hard constraints.** Compliance requirements, latency targets, “cannot change the application code,” “must stay in this region.” These eliminate options mechanically, before you have to think about which is *best*.

**3. Eliminate what’s impossible.** Usually one or two options are wrong on a technicality, not on judgment. Strike them.

**4. Then compare what’s left on the stated criterion.** By this point you’re usually choosing between two workable answers, and the question already told you the tiebreaker.

**5. Don’t fall for the familiar name.** A service you use daily appearing in an option isn’t evidence. The exam knows you’ll gravitate toward what you recognize, and distractors are built from exactly that.

Managing the clock

Roughly two minutes per question. Some take thirty seconds and some take four minutes, which averages out if you let it.

**Use the flag.** Pick your best answer, flag it, move on. Coming back with fresh eyes beats grinding. The review screen at the end shows everything you flagged.

**Check your pace at question 20.** If you’re behind, adjust then, not at question 55.

**Never leave anything blank.** There’s no penalty for a wrong answer. Guess on everything remaining if you run out of time.

Practicing the right way

Doing hundreds of questions isn’t the point. Understanding your misses is.

For every question you get wrong, write one line: was it missing knowledge, a misread requirement, a terminology slip, or time pressure? After fifty questions you’ll see your pattern, and the fix for each cause is completely different. [Turning practice questions into a learning system](/blog/aws-practice-questions-learning-system/) goes deeper on this.

And review the ones you got right. A question you answered correctly by eliminating two options and flipping a coin is a gap that hasn’t bitten you yet.