I've looked at a lot of SAA-C03 failure patterns. The ones that show up again and
again aren't from people who don't know AWS. They're from people who know AWS
well but are treating the wrong exam.
Here's what I mean.
## The exam isn't testing AWS knowledge
SAA-C03 is a scenario exam. Every question gives you four options that are
technically correct. The answer is determined by two or three words buried in the
scenario stem: "cost-effective," "minimal operational overhead," "existing
on-premises LDAP directory," "millisecond latency."
Those words are the exam. Not the service names.
A developer who has shipped production workloads on AWS knows what S3, RDS,
Lambda, and SQS do. That knowledge gets you to 60-65% on SAA-C03 without trying.
It does not reliably get you to 72% (720 scaled) because the hard questions aren't
asking what a service does. They're asking which combination of services satisfies
the constraint described in the scenario.
The symptom that tells me this is someone's problem: after the exam, they remember
questions where they narrowed to two answers and picked the wrong one. Both answers
used correct AWS services. One of them satisfied the scenario constraint. They
picked the one that was more technically interesting.
## The second failure mode: domain-specific gap that isn't where you think
AWS provides a score report broken down by four domains:
- Design Secure Architectures (30%)
- Design Resilient Architectures (26%)
- Design High-Performing Architectures (24%)
- Design Cost-Optimized Architectures (20%)
Candidates who fail with a score report showing near-passing across all domains
usually have one of two problems. Either the domain label is masking a subtopic gap
(Route 53 routing policies under Resilient Architectures, for example, which trips
candidates who drilled EC2 HA patterns exclusively), or they're dealing with
practice bank fatigue: a 78% score on a question bank they've seen twice is a
memory score, not a capability score.
Switching to a fresh question source in the final two weeks changes the picture
significantly. A lot of candidates who "should have passed based on practice scores"
are measuring recall, not readiness.
## The third one: running out of time
130 minutes for 65 questions is two minutes per question on average. Some questions
are 50 words. Some are 200. The budget is uneven and candidates who don't actively
pace run out of time on the last ten questions.
AWS doesn't differentiate between a wrong answer and an unanswered one. Both score
zero. Candidates who time out on the final ten with a 710 scaled score are losing
to the format, not the content.
The fix is specific: 90 seconds from when you start reading a question, if you
haven't narrowed to two answers, mark it and move on. Select your current best
guess so you have something if time runs out. Come back during the review period.
The question you'll spend five minutes on early in the exam is almost always harder
than the one you'd rush at the end. Mark and move is a rule, not a suggestion.
I wrote a longer breakdown of all three failure modes with the per-domain
diagnostic at doc.claudelab.me if you're working through a retake.
What's your experience been? Curious whether the constraint-reading problem
resonates with others who've sat SAA-C03 more than once.
