AWS BudgetsAWS Cost Explorer
#1Both deal with AWS spend, so candidates treat them as the same tool.
Deciding signal
Budgets is forward-looking: you set a spend threshold and get alerted when you approach or cross it. Cost Explorer is backward-looking: it shows historical costs, usage trends, and projections. "Notify me when monthly spend exceeds $500" is Budgets. "Show me where my EC2 costs increased last quarter" is Cost Explorer.
Quick check
Is the scenario setting a future limit, or reviewing past spending?
Why it looks right
Both appear under the same cost management category. Candidates who have not mapped each service to its temporal direction treat them as interchangeable.
AWS Compute OptimizerAWS Cost ExplorerAWS Trusted Advisor
#2All three appear in cost optimization questions, but they do different things.
Deciding signal
Compute Optimizer produces machine-learning-backed rightsizing recommendations based on actual utilization metrics for EC2, Lambda, EBS, and ECS. Cost Explorer shows cost and usage history — what happened, not what to change. Trusted Advisor runs static best-practice checks across multiple categories including cost, but at the advisory-checklist level, not utilization analysis. When a scenario describes underutilized EC2 instances and asks which service recommends instance changes, Compute Optimizer is the specific answer.
Quick check
Is the scenario asking for utilization-based rightsizing (Compute Optimizer), cost history (Cost Explorer), or advisory checklist checks (Trusted Advisor)?
Why it looks right
Trusted Advisor is a common wrong answer because candidates know it helps with costs — it does, but through static checks, not ML-based utilization analysis.
Amazon CloudFrontAWS Global Accelerator
#3Both improve global latency, so candidates default to CloudFront regardless of context.
Deciding signal
CloudFront caches content at edge locations. Its latency benefit depends on cache hits and is meaningful for cacheable assets: images, static files, video. Content that is unique per request cannot be cached, so CloudFront provides no latency advantage for it. Global Accelerator routes traffic over the AWS backbone network regardless of cacheability — it benefits dynamic, non-cacheable workloads where CloudFront would not help.
Quick check
Can the content be cached? Cacheable points to CloudFront; unique per request points to Global Accelerator.
Why it looks right
CloudFront is the more recognizable service and "global low latency" is how candidates recall it. The cacheability question is easy to skip when both options are framed around the same latency problem.
Amazon GuardDutyAmazon InspectorAWS WAFAWS Shield
#4All four are AWS security services. Candidates apply whichever one they have most recently reviewed without mapping each to its specific layer.
Deciding signal
GuardDuty continuously monitors API activity, network traffic, and log behavior to identify anomalies and known threat patterns. It detects; it does not block. Inspector scans compute resources for software vulnerabilities on a scheduled or event-driven basis — it evaluates configurations, not live traffic. WAF operates at the HTTP layer and actively blocks web requests matching defined rules such as SQL injection, cross-site scripting, or specific IP ranges. Shield provides DDoS protection: Standard is automatic and included at no cost; Advanced adds a response team and financial protections. Trusted Advisor runs periodic advisory checks and is not a threat detection service. When the scenario asks which service would stop or block a web-layer attack, WAF is the answer. When it asks which service detects unusual activity or behavioral anomalies, GuardDuty is the answer.
Quick check
Is the requirement to detect anomalies (GuardDuty), scan for vulnerabilities (Inspector), block web-layer attacks (WAF), or defend against DDoS (Shield)?
Why it looks right
GuardDuty is a common wrong answer on blocking questions because "threat detection" sounds active. It identifies threats; WAF is the service that actually intercepts and drops requests.
AWS CloudTrailAmazon CloudWatch
#5Both involve monitoring AWS, so candidates reach for CloudWatch first.
Deciding signal
CloudTrail records API calls: who made the call, when, from where, and what was affected. It is the account activity audit trail. CloudWatch collects operational metrics (CPU, memory, request counts), logs from services and applications, and enables alarms and dashboards. "Who deleted that S3 bucket" or "audit API calls for compliance" points to CloudTrail. "Alert when CPU exceeds 80%" or "monitor Lambda error rates" points to CloudWatch.
Quick check
Is the scenario about who did what at the API level (CloudTrail) or how a resource is performing (CloudWatch)?
Why it looks right
The word "monitoring" appears in questions about both services. CloudWatch monitors resource behavior; CloudTrail records account-level actions. These are different data sets with different purposes.
Amazon RDSAmazon DynamoDB
#6Both are managed databases — the confusion is usually in the data model signal.
Deciding signal
RDS manages relational engines: MySQL, PostgreSQL, Oracle, SQL Server, MariaDB. It suits structured data with a fixed schema, joins, and transactional requirements. DynamoDB is a key-value and document store. It suits high-velocity access patterns with variable item attributes and sub-millisecond latency at scale. The deciding signal is usually in the workload description, not just the word "database."
Quick check
Does the scenario describe structured relational data with joins (RDS), or flexible items accessed by key at high speed (DynamoDB)?
Why it looks right
Candidates default to whichever managed database they are more comfortable with when the scenario does not make the data model explicit.
AWS LambdaAWS FargateAWS Elastic Beanstalk
#7All three appear in "no server management" scenarios and are often described as serverless or managed.
Deciding signal
Lambda is event-driven and stateless, with a maximum execution duration of 15 minutes per invocation. Fargate runs containers without requiring EC2 instances — it suits containerized workloads that need more than Lambda allows, including longer runtimes. Elastic Beanstalk is a PaaS layer: you upload application code and AWS provisions and manages the underlying EC2 infrastructure. When the scenario describes short event-triggered tasks with no infrastructure, Lambda fits. When it describes containers without server management, Fargate fits. When it describes uploading application code and delegating all infrastructure decisions to AWS, Elastic Beanstalk fits.
Quick check
Is this event-triggered with a short runtime (Lambda), containers without EC2 (Fargate), or code upload with AWS managing the infrastructure (Elastic Beanstalk)?
Why it looks right
All three are described in marketing terms as managed or serverless. The distinction is in what is being abstracted: the function runtime, the container host, or the application infrastructure.
Amazon SQSAmazon SNSAmazon SES
#8SQS and SNS both handle messaging; SNS and SES both send email.
Deciding signal
SQS is a durable pull-based queue: messages are held until a consumer retrieves them, so worker-paced consumption and message retention are guaranteed. SNS is push-based pub/sub: it publishes once and delivers simultaneously to all subscribers — Lambda functions, HTTP endpoints, SQS queues, email addresses. Messages are not retained after delivery. SES is an email delivery service for transactional and marketing email at scale. SNS can send email, but SNS is for fan-out notification to multiple systems; SES is for high-volume dedicated email delivery.
Quick check
Does the scenario require message retention and worker-paced consumption (SQS), simultaneous fan-out to multiple systems (SNS), or high-volume transactional email (SES)?
Why it looks right
SNS is the common wrong answer on SQS questions because "notification" sounds like a general messaging service. The key difference is retention: SQS holds messages until consumed; SNS discards after delivery.
EC2 Reserved InstancesAWS Savings PlansEC2 Spot Instances
#9All three reduce EC2 costs. The workload description is what separates them.
Deciding signal
Reserved Instances provide the deepest discount for a specific instance type, size, and Availability Zone over a 1- or 3-year term. Best when the instance configuration will not change. Savings Plans provide flexible compute discounts across instance families and Regions in exchange for an hourly spend commitment — the right choice when the instance type may change during the commitment window. Spot Instances offer steep discounts but AWS can reclaim capacity with short notice. Spot only fits interruption-tolerant workloads such as batch jobs or rendering, where a mid-run termination is acceptable.
Quick check
Can the workload be interrupted? If not, eliminate Spot. Will the instance type change? Flexibility points to Savings Plans; fixed configuration points to Reserved Instances.
Why it looks right
Reserved Instances are the most familiar cost-reduction mechanism, so candidates apply them to all "save money" scenarios, including ones where interruptibility or instance flexibility changes the right answer.
AWS ArtifactAWS Audit ManagerAWS Config
#10"Compliance" is broad enough that candidates apply whichever service they reviewed most recently.
Deciding signal
Artifact provides access to AWS's own pre-existing compliance documentation: SOC reports, PCI DSS attestations, ISO certifications. Use it when the scenario asks to retrieve proof that AWS infrastructure meets a regulatory standard. Audit Manager automates the collection of the customer's own compliance evidence, mapping AWS usage to audit frameworks. Config continuously evaluates resource configurations against defined rules and flags noncompliant resources — it answers whether your resources are configured correctly on an ongoing basis.
Quick check
Is this about retrieving AWS's certifications (Artifact), collecting your own audit evidence (Audit Manager), or enforcing configuration rules on your resources (Config)?
Why it looks right
Config is the common wrong answer on Artifact questions. 'Compliance' sounds like a configuration problem, but when the scenario is asking to obtain AWS's existing audit reports, Config does not produce those documents.
10 CLF-C02 questions. Pattern-tagged with trap analysis. Free, no signup required.
Start CLF-C02 Mini-Trainer →