5 mistakes people make when studying for the SAA-C03

Every month we see the same questions in the community: "I studied for three months and barely failed, what was I missing?". The answer is almost never "you needed to study more", but "you studied the wrong way". These are the five mistakes that come up most often when preparing for the SAA-C03, and what to do instead.

1. Memorizing instead of understanding architectures

The most common mistake: turning your study into a stack of flashcards with "service X is for Y". It works for direct questions, but the SAA-C03 has almost no direct questions. Most give you a scenario with constraints (cost, latency, high availability, compliance) and expect you to combine two or three services to solve it.

If you memorized that "S3 is for storing objects" but don't understand why you'd choose S3 with Cross-Region Replication over a solution with EC2 and EBS for a disaster recovery case, you'll fail the scenario even if you know the service by heart. The fix is simple but uncomfortable: for every service you study, ask yourself "what business problem does it solve, and which two other options does it compete with?". That's what's really being assessed.

2. Ignoring the Well-Architected Framework

Many people treat the Well-Architected Framework as an optional appendix at the end of the syllabus, when in reality it's the logic behind almost every long scenario question. The six pillars (operational excellence, security, reliability, performance efficiency, cost optimization, and sustainability) are literally the criteria the exam uses to rule out the "almost right" options.

When a question gives you four technically possible answers, the difference between the right one and the distractors almost always comes down to a pillar of the framework: the right option minimizes cost without sacrificing the required SLA, or improves reliability without adding unnecessary operational complexity. If you don't internalize this logic, you'll "guess right" on easy questions and fail systematically on the hard ones.

3. Not practicing with mock exams

Reading documentation and watching video courses gives you knowledge, but it doesn't train you for the exam format: 65 questions in 130 minutes, many with three- or four-paragraph prompts you have to read quickly without missing the detail that changes the answer. Without practicing in that format, it's common to get to the real exam and run out of time — or worse, misread the scenario by rushing.

Mock exams do two things no theory course can replace: they show you where your real gaps are (not the ones you think you have), and they get you used to the reading pace the exam demands. If you're going to do just one thing differently in your preparation, make it this: take full, timed mock exams starting several weeks before the exam, not just in the last week.

4. Underestimating long scenario questions

Related to the previous point, but it deserves its own: there's a question pattern in the SAA-C03 that packs multiple requirements into a single prompt ("the company needs X, plus Y, but with constraint Z, and the security team requires W"). Many people read the prompt looking for the first keyword they recognize (say, "high availability") and jump to the option they associate with that word, ignoring the other constraints.

The result is choosing an answer that's technically valid for one part of the problem but ignores another explicit constraint. The practical fix: underline (mentally or on the scratch pad you're given) every constraint in the prompt before looking at the options, and rule out any option that doesn't meet all of them, not just the most obvious one.

5. Confusing similar services

The exam is designed to exploit confusion between services that sound alike but solve different problems. The most common pairs:

• ▸SQS vs SNS: SQS is a queue (each message is processed by one consumer, letting you decouple producer and consumer so they run at different speeds). SNS is pub/sub (a message can reach multiple subscribers at the same time). If the scenario says "multiple systems must react to the same event", it's SNS (or SNS+SQS combined). If it says "process tasks from a queue without losing messages if the consumer fails", it's SQS.
• ▸ALB vs NLB: the Application Load Balancer operates at layer 7 (HTTP/HTTPS, path- or host-based routing). The Network Load Balancer operates at layer 4, with ultra-low latency and static IP support. If the scenario talks about "millions of connections per second" or "a fixed IP for allowlisting", think NLB. If it talks about "routing by path to different microservices", think ALB.
• ▸VPC peering vs Transit Gateway: peering connects VPCs in pairs, with no transitivity; Transit Gateway centralizes connectivity between many VPCs and on-premises connections. If the scenario mentions "dozens of VPCs" or "a hub-and-spoke architecture", it's almost always Transit Gateway.

Don't memorize these differences as standalone definitions: practice with questions that put them in real scenarios, which is how they show up on the exam.

How to fix these mistakes before your exam

None of the mistakes on this list is solved by "studying more hours" in the sense of piling up more reading. They're solved by changing your method: understanding the why behind each service, internalizing the Well-Architected Framework as decision logic, and measuring your progress with real mock exams instead of the subjective feeling of "I've got this".

If you're in the middle of your preparation, check out our complete SAA-C03 guide for a structured study plan, and use the SAA-C03 cheat sheet to quickly review the key services and comparisons in the final days before the exam.