Based on the SAP-C02 exam guideReport an error
It's the last day of the 80 days. Today we don't learn new concepts. Instead, we break down how the exam is scored and the psychology with which question writers lay out wrong answers (distractors), compress the knowledge of the four domains you've built up into "keyword → direct answer" reflexes, and finish with a 12-question scenario mock exam of real-exam length. SAP-C02 is not a knowledge exam but a judgment exam — it measures your ability to pick, from among answer choices that all look correct, "the single one that meets the requirements most precisely with the lowest operational overhead."
🔍 Going deeper: SAP-C02 is 75 questions / 180 minutes, and the passing mark is 750 out of 1000 (about 75%). The key point is that it is a scaled score — not a simple percentage of correct answers but a score adjusted for question difficulty (based on IRT, Item Response Theory). Also, 10 of the 75 questions are unscored questions used to gather statistics on new items. You can't tell which ones are unscored, so you must treat every question the same. There is no partial credit, and there is no penalty for wrong answers (blank < guess). So always guess on questions you don't know — a 4-choice question has an expected value of 25%, and if you eliminate 2, it's 50%.
There are patterns in how distractors are designed in a multiple-choice exam. Typical ones are "plausible but missing one condition," "correct but excessive (over-engineering)," and "correct but with high operational overhead." Most of SAP-C02's traps are one of these three — for example, the correct answer is a managed service but they lay out "build it yourself on EC2" (excessive operational overhead), or the correct answer is an SCP but they lay out "IAM Policy" (missing the condition that it can't restrict root). So the scoring strategy is to eliminate choices not by treating them as answer candidates but by "which condition they violate."
⚠️ Trap: Pro exam question stems are long and noisy. They plant irrelevant information such as "company revenue, number of employees, history of past systems" and then ask about the key constraint. The strategy is to read the actual question at the end of the stem first → extract the constraint keywords (RTO/RPO, operational overhead, cost, isolation, regulation, downtime) → then scan the stem in reverse. If you read every sentence in order, you'll exceed 144 seconds.
The Pro exam doesn't state the answer directly; it signals with encoded keywords. The decoding dictionary:
| Question keyword | Answer signal |
|---|---|
| "Minimal operational overhead / no management overhead" | Managed/Serverless (Lambda, Fargate, Aurora Serverless) |
| "Most cost-effective" | SP, Spot, Lifecycle, serverless, avoid On-Demand |
| "Isolate failures" | The AZ → Region → Account ladder |
| "Without human intervention / automatic recovery" | EventBridge, SSM Automation, Auto Scaling |
| "Must block even the root user" | SCP (not IAM Policy) |
| "Absolutely no changes / regulation" | Object Lock/Vault Lock Compliance (not Governance) |
| "RPO in seconds/minutes, global SQL" | Aurora Global Database |
| "Local writes in each Region, eventual OK" | DynamoDB Global Tables |
| "Minimal downtime, whole OS, one-time" | MGN (DRS if it's ongoing DR) |
| "Heterogeneous DB + schema conversion" | DMS + SCT |
| "Tens to hundreds of TB + slow line" | Snow Family + online delta |
| "SaaS → many customer VPCs, IP conflicts irrelevant" | PrivateLink |
| "DX failure failover within 1 second" | BGP + BFD |
| "Highly reliable failover where a human decides" | Route 53 ARC Routing Control |
🎯 Scenario: When a stem says "a startup is growing fast, has only 3 engineers, and can't spend time on operations," this is a code not about cost or scale but telling you to pick "minimal operational overhead → the most managed/serverless choice." Even for the same architecture, if "there is a dedicated SRE team and fine-grained control is needed," it tilts toward more controllable choices like EC2 and EKS. Organizational context is the weight on the correct answer.
| Domain | Weight | Key tools | Answer reflex |
|---|---|---|---|
| 1. Organizational complexity | 26% | Org, SCP, Control Tower, TGW, DX, PrivateLink, Outposts | Account = isolation, SCP = ceiling, TGW = O(N), Snow = bandwidth |
| 2. New solutions | 29% | Lambda, Fargate, Aurora, DynamoDB, EventBridge, Kinesis | CAP/PACELC, decoupling, fan-out |
| 3. Migration and modernization | 20% | MGN, DMS+SCT, Snow, App2Container, Strangler Fig | Rehost first, CDC for zero downtime, no big bang |
| 4. Continuous improvement | 25% | CloudWatch, SSM, CodeDeploy, FIS, Compute Optimizer | SRE, automate toil, gradual deployment, FinOps |
SAP-C02 is an "optimal solution" judgment exam. All the depth of the 80 days converges into one reflex — read the question at the end of the stem first → extract the constraint keywords → decode the keywords into answer signals → if two look equally correct, pick the one with lower operational overhead. Scoring is a scaled score (750/1000) and a blank is a loss, so always guess even on what you don't know. Passing is decided not by the amount of knowledge but by the speed of reading constraints and judging trade-offs.
Now we solve 12 scenario questions of real-exam length to wrap up the 80 days. Each explanation breaks down even why the other choices are wrong.
Click a choice to reveal the answer and explanation.
Question 1
A multinational company operates more than 200 AWS accounts. The security team wants to block use of non-EU Regions in all accounts (including root), have new accounts automatically inherit standard logging and guardrails, and view security findings from all accounts in one place. Which of the following combinations meets all of these requirements?
Question 2
A global SaaS reads and writes users' session data from multiple Regions around the world. Each Region must read and write with local latency (single-digit ms), and it is enough for changes from other Regions to be reflected eventually (strong consistency is not needed). Operational overhead must be minimal. What is the most suitable design?
Question 3
You must migrate 100TB of on-premises Oracle within a 6-month deadline. You want to convert only the core analytics DB to Aurora PostgreSQL to eliminate licensing, and move the rest quickly and then optimize gradually in the cloud. What is the most suitable strategy?
Question 4
An order system triggers 4 microservices for each order: payment, inventory, shipping, and notification. If one service fails, the others must remain healthy, and the failed service must process the backlog of messages without loss after recovery. What is the most suitable pattern?
Question 5
A financial regulator requires transaction audit logs to be retained for 7 years so that no one (including administrators and root) can modify or delete them. Logs are generated in all member accounts. What is the most suitable design?
Question 6
You have 200 VPCs across 3 Regions connected to 2 on-premises data centers, but direct communication between EU VPCs and US VPCs must be blocked. If a DX circuit fails, traffic must automatically switch to the backup VPN within 1 second. What is the most suitable combination?
Question 7
You must apply security patches monthly to 500 EC2 instances. Security policy prohibits SSH and bastions, patches are applied incrementally by group within set time windows, and all access must leave an audit record. What is the most suitable configuration?
Question 8
A mission-critical SQL workload must recover in another Region with RPO under 1 second and RTO within 1 minute if the primary Region fails. Normally the secondary Region shares the read load. What is the most suitable design?
Question 9
You want to verify resilience by injecting deliberate failures every quarter, with the experiment automatically stopping if real user impact exceeds a threshold, and the experiment must run on an automatic schedule. What is the most suitable configuration?
Question 10
A company wants to convert a 20-year-old Java EE monolith (sharing a single Oracle DB) into microservices. Rewriting everything at once carries a large risk of business disruption. What is the safest incremental strategy?
Question 11
A SaaS company wants to build a RAG chatbot that answers based on its own documents while minimizing operational overhead. It doesn't want to run model hosting or vector infrastructure itself. What is the most suitable design?
Question 12
In the event of a full Region failure, you want to avoid false positives from automatic health checks and have the operations team assess the situation and shift traffic to another Region reliably. At the same time, cost should be a compromise, but RTO must be in minutes. What is the most suitable combination?
"SAP-C02 is an 'optimal solution judgment' exam. Read the question at the end of the stem first → extract the constraint keywords (operational overhead, cost, RTO/RPO, isolation, regulation, downtime) → decode the keywords into answer signals → if two look equally correct, pick the one with lower operational overhead. A blank is a loss, so always guess even on what you don't know."
After the exam, set either a Specialty (Security, Networking, DB, ML) or hands-on depth as your next goal.