Exam Simulation B: Second Practice Set, Trap Drills and Exam Day

Fifteen lessons gave you the content and Set A gave you a first simulation. This is the second sitting, and a second sitting measures something different: not whether you know the material, but whether you can hold two neighbouring services apart while the clock runs. Almost nobody fails CLF-C02 because they have never heard of AWS Config; they fail because the stem asked for AWS's evidence and they supplied their own, or because the stem said read traffic and the option said Multi-AZ. On a compensatory 100 to 1,000 scale with 700 to pass, discrimination is what separates a comfortable pass from a near miss.

By the end of this lesson you will be able to:

• state the one-line correct framing for each of the ten trap pairs most likely to cost you marks;
• spot a half-true option, the distractor that swaps exactly one attribute of a pair, and kill it in one clause;
• handle multiple-response stems with an all-or-nothing procedure that never assumes partial credit;
• run the arithmetic the exam actually asks for: RPO/RTO, discount ceilings and database-by-pattern selection;
• recite the official exam-day logistics: IDs, arrival, check-in, breaks, comment time, results timing;
• grade Set B by domain block against explicit ready-versus-not-ready thresholds, and defend the comparative verdict on booking now versus seven more days;
• run a weight-driven 7-day study plan that revises by domain dip rather than by chapter order.

---

1. Strategy: trap pairs and how the exam weaponises them

1.1 What a pair is, and what a distractor really looks like

AWS describes its own distractors as "generally plausible responses that match the content area". The practical consequence is that most options you face are drawn from the same decision as the key; which means most losses on this exam are not knowledge failures, they are pair failures: you knew one member of the pair but could not state the line between them under time pressure.

A pair becomes an item through exactly three moves:

1. Swap one attribute per option: stateful/stateless, allow/deny, push/pull, synchronous/asynchronous, grant/cap.
2. Swap the ask: the stem wants read scale, the option offers availability; the stem wants evidence about AWS, the option offers evidence about you.
3. Swap the owner: AWS's job versus your job, under the shared responsibility model.

Worked example 1: a half-true option, dissected before you read the stem. Take the security group / network ACL pair and build all four options the way the exam does:

Only B survives, and it survives because you evaluated both halves rather than recognising a familiar word. Options A, C and D each contain a phrase you have read in the documentation; that is the design. A distractor does not have to be false; it has to be false in the place the stem is asking about.

1.2 The three questions for any pair item

Slow candidates read options; fast candidates interrogate the pair first. Ask these three, in this order, before you look at a single option:

1. What does each member of the pair actually do? One clause each, from memory. If you cannot say what a read replica does in one clause, the options will answer it for you: badly.
2. Which stem qualifier selects between them? DENY, subnet, read traffic, fan out, maximum permissions, layer 7, AWS's own, interruptible. The noun in the stem is usually the decoy; the qualifier is the key.
3. Does the surviving option satisfy both halves? For a (Select TWO) stem this is literal: both statements must be correct, and AWS's published rule elsewhere in its certification material is that you must select all correct responses to receive credit.

• 📚 Did you know? AWS's own Official Practice Question Set for CLF-C02 contains exactly 20 questions, the same count as Set A and Set B here, is free, can be retaken, reshuffles the same items on each attempt, and gives detailed feedback for answer choices including recommended resources. It is the first step of AWS's four-step prep plan, and it is the closest official yardstick for the pacing you are training in this lesson (as of Oct 2026).

---

2. Strategy: the three item shapes that decide Set B

CLF-C02 publishes exactly two question types: multiple choice (one correct response plus three distractors) and multiple response (two or more correct responses out of five or more options). There is no ordering, no matching and no case-study item on this exam, so do not budget for them. What Set B adds on top of Set A is practice in the three shapes that cost the most marks: multiple response, calculation, and selection by pattern.

2.1 Multiple response without partial credit

The procedure that survives contact with five or six options:

1. Mark every option definitely yes / definitely no / maybe: in one pass, without choosing.
2. Resolve every maybe by re-reading the qualifier, not the option.
3. Submit only when no maybe remains. Two obvious answers plus an overlooked fifth correct one is the classic zero.

2.2 The calculation and selection shapes

Three families appear repeatedly, and each has one mechanical procedure:

Worked example 2: RPO/RTO arithmetic. A database fails at 14:00. The last successful backup completed at 08:00, and service is restored at 16:30.

$$\text{RPO} = 14{:}00 - 08{:}00 = 6\ \text{h of data at risk} \qquad \text{RTO} = 16{:}30 - 14{:}00 = 2.5\ \text{h of outage}$$

RPO = data, RTO = time. Compute both gaps first, then assign the letters: candidates who assign them first lose the item while doing the arithmetic correctly (see Set B item 17).

Worked example 3, discount ceilings, as of Oct 2026 (verify current before use). For an uninterrupted workload running 24/7 for three years in one instance family: Spot is up to 90 % but interruptible (two minutes' notice), a 1-year Compute Savings Plan is up to 66 % and commits for one year, a 3-year EC2 Instance Savings Plan is up to 72 % and matches the horizon, and On-Demand is the undiscounted baseline. The largest surviving ceiling is 72 %, the biggest headline number was disqualified by a constraint, not by arithmetic (see item 19).

Worked example 4: database-by-pattern selection. Translate the stem's nouns into a data model before reading service names:

Two of those rows become Set B items 15 and 16. The pattern never changes: noun → data model → service, and only then the options.

2.3 The Set B blueprint, its clock and its weights

Only domain percentages are published; the item counts below are arithmetic on 50 scored items: a planning aid, never an AWS figure.

Worked example 5, weights to items, twice over. On the live exam: $50 \times 0.24 = 12$, $50 \times 0.30 = 15$, $50 \times 0.34 = 17$, $50 \times 0.12 = 6$, and $12 + 15 + 17 + 6 = 50$. On Set B: the same proportions round to 5 / 6 / 7 / 2, and $5 + 6 + 7 + 2 = 20$. Domains 2 and 3 together are 64 % of scored content, that is why they hold 13 of these 20 items.

Worked example 6, Set B's own clock. Twenty items at the first-pass budget of 60 s each is $20 \times 60 = 1{,}200\ \text{s} = \mathbf{20}$ minutes. The full 83-second ceiling would give $20 \times 83 = 1{,}660\ \text{s} = 27.7$ minutes, comfortable, but it trains a rhythm you will not be able to keep for 65 items. Sit Set B on 20 minutes flat; if you finish with time left, spend it on flagged items, not on re-reading confident ones.

• 📚 Did you know? AWS's own published rhythm for a Foundational certification is 2 to 3 weeks of preparation in 30 to 60 minute chunks, with at least one full-length timed practice exam before the proctored one. The 7-day plan in section 7 is therefore a compressed plan: it assumes lessons 01 to 15 and Set A are already behind you, and it spends the week on discrimination and logistics rather than on new content.

---

Real-World Case Drills

Two AWS-published customer stories, converted into the only two moves this exam ever makes with a story: map it to a domain, then predict the stem. Every percentage below is a customer result that AWS chose to publish: never an AWS promise and never a number without its date (case-study pages carry no publish date, so the only timestamp available is the access date; all figures below accessed Oct 2026).

Case drill 1, Capital One: eight data centers to zero

The scenario. A regulated bank (Fortune 100) runs eight on-premises data centers under a hardware-refresh cycle while trying to scale machine-learning and real-time personalisation work. AWS's published account: Capital One exited all eight data centers by migrating to AWS (the last in 2020), built roughly 80 % of its ~2,000 applications in the cloud from the ground up, recycled 103 tons of copper and steel, cut the average time to build a development environment from three months to minutes, cut disaster-recovery test time by 70 %, and cut incident resolution and transaction errors by 50 %.

Which domain? D1 Cloud Concepts (24 %): the story is an argument about agility, opex versus capex and the migration strategies, not about a service. The hooks that reach into D3 (serverless, migration services) and the separate AWS resilience page for the same customer (critical-severity events down 80 to 90 %, recovery hours to minutes) are secondary: those numbers come from a different case study, so never blend the two pages in one answer.

What the exam would ask? The four stem shapes this story normally becomes:

Trap to pre-load: "AWS guarantees 90 % savings" is wrong twice; it converts a customer's result into a promise, and AWS never publishes such a guarantee. Rewrite every case percentage as "AWS reports that <customer> achieved …".

Case drill 2, Box: $2.23 million found in four line items

The scenario. An enterprise SaaS company wants to cut spend without weakening security, reliability or performance. It applies the Well-Architected Framework and expert reviews, then attacks four specific bills: S3 and Glacier storage tiering, EBS volume and snapshot hygiene, CloudTrail event filtering, and traffic routed around internet gateways. AWS reports $2.23 million in savings, decomposed as:

Which domain? D4 Billing, Pricing and Support (12 %), with a D3 hook on storage classes. The point of the story is the ranking: the biggest saving was not a discount negotiation, it was architecture.

What the exam would ask?

• "Which lever produced the largest single saving?" → data transfer out (egress), not a commitment discount.
• "Which free resource gives best-practice cost guidance before any code is written?" → the Well-Architected Framework review: free guidance, distinct from Trusted Advisor (checks your account) and from a paid Support plan (opens cases).
• "Where would a customer see inter-AZ charges accumulating?" → Cost Explorer / cost-allocation views: never the Pricing Calculator, which estimates spend that does not exist yet (the Q20 four-way split applies unchanged).

Trap to pre-load: Well-Architected review ≠ Trusted Advisor ≠ Support plan: three different questions wearing one word. And $2.23 million is Box's result; the exam loves an option that restates it as an AWS guarantee.

• 📚 Did you know? Box's headline total is arithmetic you can check in one line: 438,000 + 1,100,000 + 500,000 + 192,000 = 2,230,000. Case-study figures on this exam are almost always decomposable: when an option's components do not add up to its own total, it is wrong even if every individual number looks familiar.

2026 Updates (as of October 2026)

• 📚 Did you know? The current exam guide PDF prints no version number at all: only Copyright © 2026. The single stamped artefact is the launch-era Version 1.0 CLF-C02 PDF, and the uiVersion=2024.10 you may spot in the docs URL is the AWS Docs site template version, not a guide revision date (checked Oct 2026). Whenever a prep site claims "guide v1.3, March 2026", treat the claim as unverified until you see it on AWS's own page.

---

3. Practice Questions, Set B: 20 items in published weight order

Set B mirrors the live exam's dominant format, one correct response, three distractors, while drilling four multiple-response-style stems. Items run in domain-weight order: D1 five, D2 six, D3 seven, D4 two, which is the published 24 / 30 / 34 / 12 split rounded onto twenty items.

3.0 How to sit Set B

3.1 Domain 1: Cloud Concepts (24 %) · items 1 to 5

Teardown: Q1 · which CAF perspective. Why it is right: Governance is the perspective that owns IT governance, IT service management, policy and risk. Why the traps lose: A is the financial and value case (benefits, revenue, reduced business risk); C is the implementation foundation: migration and infrastructure; D is operations and services delivery. Trap: CAF's six perspectives are routinely confused with Well-Architected's six pillars: Governance exists only in CAF, Sustainability only in Well-Architected (see Q3). Never swap the two vocabularies. Source: seed Q03 · cloud-economics/migration digest F.3 (AWS CAF perspective list).

Teardown: Q2 · the two-way behaviour. Why it is right: the stem contains two verbs: adds and releases, and only elasticity covers the release half. Why the traps lose: A is the ability to handle growth, with no requirement to shrink; C is surviving a component failure across Availability Zones, which the stem never mentions; D is the aggregation argument: hundreds of thousands of customers lowering unit price: a pricing effect, not a scaling behaviour. Trap: "it scaled out" pulls candidates to scalability; the discriminator is the two-way behaviour, not the presence of scaling. Source: seed Q06 · cloud-concepts digest F.5 (elasticity ≠ scalability ≠ high availability).

Teardown, Q3 · pillar, not topic. Why it is right: Sustainability is one of the six published pillars. Why the traps lose: A is a Domain 2 headline, compliance sits under governance, not under the framework's pillars; B and D are real cloud benefits from Domain 1, which is exactly why they sound pillar-shaped. Trap: all four options are true statements about AWS; only one is a pillar. Memorise the six as a set: operational excellence, security, reliability, performance efficiency, cost optimization, sustainability. Source: seed Q08 · monitoring/support digest F.9 · exam-guide digest §B.2 task 1.2 (pillar list verbatim).

Teardown: Q4 · three of six, all from one task. Why it is right: rightsizing, BYOL-versus-included licensing and benefits of automation are all named in task 1.4: benefits of automation is explicitly an economic benefit there, not a technical one. Why the traps lose: B mixes one true item with a reliability design choice (disaster recovery, task 3.2 territory) and a network security control (task 3.5): both are off-list; C pairs two true items with an absolute claim AWS never publishes: availability is a shared responsibility, so no guarantee exists; D is the nastiest of the four: two of its three items are on the list, but "eliminating the shared responsibility model" is false, so the triple fails. Trap: a 2-of-3 triple is the standard MR distractor; you must validate every member, not the first two you recognise. Live-exam form: three checkboxes, all-or-nothing. Source: seed Q09 · exam-guide digest §B.2 task 1.4 (knowledge and skills bullets verbatim).

Teardown: Q5 · extends, not replaces. Why it is right: the stem has two halves: still running VMware on premises and extends to AWS with a consistent experience, and hybrid is the only model that keeps both. Why the traps lose: A is AWS-hosted infrastructure reached over the public internet, which ignores the on-premises half; B is single-organisation infrastructure (AWS illustrates it with Amazon VPC) and ignores the extends to AWS half; D requires two or more public providers, and your own data centre is not a provider. Trap: "still running on premises" tempts private cloud; extends is the operative verb. Source: seed Q10 · deployment/operations digest §B.2 (deployment models in AWS wording).

3.2 Domain 2: Security and Compliance (30 %) · items 6 to 11

Teardown, Q6 · memorise the pair, not the words. Why it is right: stateful-plus-auto-reply for security groups and stateless-plus-explicit-return-rule for NACLs are the two properties that never swap. Why the traps lose: A reverses both properties, NACLs are the stateless, first-match ones; C grafts NACL deny onto SG evaluation, because security groups cannot express deny at all; D reverses scope (security groups attach to an elastic network interface, NACLs to a subnet) and ignores NACL deny. Trap: everything in A, C and D is half-true, the exam swaps one attribute per option, so hold the pair as a four-line block: SG = stateful, allow-only, ENI, all rules evaluated · NACL = stateless, allow+deny, subnet, first match stops. Source: seed Q14 · security/IAM digest F.4, F.7 · networking digest F.3, F.5.

Teardown: Q7 · date-stamp every MFA answer. Why it is right: SMS MFA is retired as of Oct 2026, and the supported set is passkeys/FIDO plus TOTP devices. Why the traps lose: B repeats the retired method and is the single most common stale-fact answer in study material written before the retirement; C is backwards: protecting the root user with MFA is the first recommendation; D confuses who may act (IAM policies) with who is acting (identity proofing): least privilege does not prove identity. Trap: older notes and older practice tests still say "SMS MFA". Date-stamp every MFA answer to the year you sit, and re-read the current identity documentation if a fact smells old. Source: seed Q17 · security/IAM digest F.19 and §B (root MFA registration).

• 📚 Did you know? ESL accommodation adds +30 minutes to the CLF-C02 exam and is requested once through the before-testing policy (as of Oct 2026). If you are entitled to it, request it well before your appointment: accommodations are not granted at the desk, and the same policy page says prices are valid for one exam attempt.

Teardown, Q8 · the qualifier is WHICH principal. Why it is right: which principal is an actor question, and an audit trail of API events is exactly what CloudTrail stores. Why the traps lose: A answers "is it healthy / is CPU high", metrics, alarms and dashboards, not actors; C records the configuration state of resources over time, which tells you what the table looked like, not who deleted it; D advises on best-practice gaps and cannot produce API history at all. Trap: the word monitoring appears in many stems and pulls candidates to CloudWatch; ask whether the stem wants who acted (Trail), what the state was (Config) or is it breaching a threshold (CloudWatch). Source: seed Q18 · governance/compliance digest F.1, F.3.

Teardown: Q9 · two traps in one item. Why it is right: both statements are the published job of each service: events for CloudTrail, configuration state and rule evaluation for Config. Why the traps lose: A is half-right then claims Config can do CloudTrail's job: the official walkthrough pairs CloudWatch with CloudTrail precisely because sign-in alerting needs the event stream; B is half-right then hands Config GuardDuty's job (threat detection), and the replaces wording is the giveaway; D smuggles in the "90 days free covers everything" myth: the free Event history covers management events, and longer retention or all-account coverage needs a trail, while its second statement is true. Trap: an option with one true statement and one false statement is wrong, not "mostly right", and the two myths this item attacks (Config can do CloudTrail's job, 90 days free covers everything) are both extremely common. Live-exam form: two checkboxes, all-or-nothing. Source: seed Q19 · governance/compliance digest F.3: F.4 · AWS re:Invent exam-prep walkthrough 2.2.

Teardown: Q10 · the answer is always on the in-scope list. Why it is right: all three named services appear verbatim in task 2.4, and all three are on the current in-scope list. Why the traps lose: B substitutes AWS Network Firewall, which is on the official out-of-scope list: legal to read as an option, never the key; C substitutes Amazon Cloud Directory, also out of scope, and swaps Shield for Inspector (a real in-scope service, but not one task 2.4 names); D pairs one correct name with CloudFormation and Batch, which belong to entirely different categories. Trap: out-of-scope services are chosen as distractors precisely because they sound more advanced than the real key. Correct answers on this exam are always in-scope services: confirm against the current in-scope list rather than your intuition about what "sounds serious". Source: seed Q21 · exam-guide digest §B.2 task 2.4 + §E.1 / §E.2 (in-scope and out-of-scope lists).

Teardown, Q11 · whose evidence is it? Why it is right: the stem says Amazon Web Services' own reports, that phrase is the whole item, and Artifact is the on-demand repository for AWS's attestations. Why the traps lose: B records your resource configuration over time, which is the mirror-image answer; C builds a landing zone with guardrails; D scans your workloads for known vulnerabilities. Trap: "compliance evidence" is a two-sided question, Artifact holds AWS's evidence, Config holds yours. If the stem ever says "gather our audit evidence", the key flips to Config, CloudTrail or Security Hub. Source: seed Q22 · governance/compliance digest §B + F.6.

3.3 Domain 3: Cloud Technology and Services (34 %) · items 12 to 18

Teardown, Q12 · the phrase and its impersonators. Why it is right: without provisioning or managing servers is Lambda's definition, you upload code, the service runs it on demand. Why the traps lose: A is infrastructure as a service: you still choose, launch and patch the operating system; C is a bundled virtual-private-server product with a fixed bundle; D is platform as a service that sounds serverless, you just upload code, yet it provisions and manages EC2 or container resources inside your account. Trap: Elastic Beanstalk's marketing is the closest impersonation of serverless on this exam; the discriminator is who manages the underlying instances. Source: seed Q23 · compute digest §B (Lambda: serverless, per-request and duration billing) · deployment/operations digest §B.10 (Beanstalk = PaaS).

Teardown: Q13 · unknown pattern plus no retrieval fees. Why it is right: two qualifiers decide it: unknown and changing (so automatic movement) and refuses to pay retrieval fees (so a class with no retrieval charge on tier moves). Intelligent-Tiering is the only option that satisfies both. Why the traps lose: B is the impostor: a lifecycle rule looks automatic, but it is a rule you write against a pattern you guessed, and access from Standard-IA does carry retrieval fees; C puts objects in a single Availability Zone, which is about cost versus durability, not about an unknown pattern; D adds a restore step measured in minutes to hours on top of the same guessing problem. Trap: "automatic tiering" tempts the lifecycle-rule answer; Intelligent-Tiering is a storage class, not a lifecycle rule, and its documented difference at this stem is the absence of retrieval fees. Note the minimum durations in the IA and Glacier classes (30 / 90 / 180 days: verify current before use). Source: seed Q26 · compute/storage digest §B.6, C5 and F.3.

Teardown, Q14 · shared, POSIX, several AZs. Why it is right: all three qualifiers, shared, POSIX, several Availability Zones, point at the file-services answer, and EFS is the in-scope managed NFS service. Why the traps lose: A is block storage attachable to one instance in one Availability Zone (multi-attach exists but must be deliberately configured, and it is not a shared file system); C stores files, which is the bait, but it is object storage reached over HTTP, not a mountable file system; D is ephemeral and local to one instance in one AZ, so it disappears when the instance stops. Trap: "it stores my files" is the most reliable lure on this domain; the qualifier mountable file system is what matters, and it maps to block (EBS) / file (EFS) / object (S3) before it maps to any service name. Source: seed Q27 · compute/storage digest §E (task 3.6 block, file and object split).

Teardown, Q15 · the word relational forbids the flagship. Why it is right: the stem's noun phrase is relational database, MySQL- or PostgreSQL-compatible, managed, which is Aurora's exact design point. Why the traps lose: A is AWS's flagship database, fast, managed and scalable, all true, but it is NoSQL key-value/document, and no amount of adjective-matching overrides relational; C is block storage, not a database at all; D is a networking traffic-optimisation service, a pure category error. Trap: DynamoDB is the most recognisable database name in the option list, so "records new customer orders" pulls candidates toward it. Database-by-pattern rule: translate nouns to a data model first (worked example 4), relational → Aurora or RDS, key-value → DynamoDB. Source: seed Q28 · database-services digest §B.6 · AWS re:Invent exam-prep walkthrough 3.4 (same distractor family: Global Accelerator / DynamoDB / EBS).

Teardown, Q16 · microsecond plus sorted sets. Why it is right: microsecond reads and strict ordering for leaderboards are the two discriminators, an in-memory engine with sorted-set data types and replication with automatic failover. Why the traps lose: B is object storage with request latencies far from microseconds and no ordering primitives; C is a genuine key-value store and a plausible "session" answer, but its headline is single-digit milliseconds and it has no leaderboard data types; D orchestrates workflows and stores no application data at all. Trap: session store tempts DynamoDB, the two services fight over exactly this stem, so key on microsecond and sorted set. And remember ElastiCache is an ephemeral cache, never your system of record. Source: seed Q30 · database-services digest §B.10 (Valkey / Redis OSS versus Memcached; microsecond reads).

Teardown, Q17 · compute both, then assign the letters. Why it is right: $14{:}00 - 08{:}00 = 6$ hours of data at risk (RPO) and $16{:}30 - 14{:}00 = 2.5$ hours of outage (RTO), see worked example 2. Why the traps lose: B does the arithmetic correctly and then swaps the definitions; C adds the two intervals together ($8.5$ h), which is neither metric, and asserts zero downtime that no restoration implies; D measures the outage from the backup rather than from the failure, which double-counts the pre-failure window. Trap: you will compute both numbers correctly and still lose this item if you assign the letters first. Memorise RPO = data, RTO = time, compute, then label. The same arithmetic underpins the recovery ladder: backup/restore (RPO hours, RTO up to 24 h) → pilot light (minutes / tens of minutes) → warm standby (seconds / minutes) → active-active (near zero / potentially zero). Source: seed Q32 · global-infrastructure digest §H.5 worked example + REL13-BP02 recovery bands.

Teardown, Q18 · the AZ minimum is the buried constraint. Why it is right: both halves are published behaviour of the Application Load Balancer, Layer 7 host/path routing and WAF association. Why the traps lose: A pairs a true statement with a numeric falsehood: an Application Load Balancer requires targets across at least two Availability Zones, so "only one" kills the pair; C is wrong twice, AWS WAF web ACLs attach to Layer 7 resources such as Application Load Balancer, CloudFront and API Gateway, not to a Network Load Balancer, and an NLB works at Layer 4; D pairs a true WAF statement with the Layer 7 claim on the wrong balancer. Trap: you probably know "ALB = Layer 7", the item is built so that you stop reading after the first half. In an MR stem, every statement is a separate question, and one false clause zeroes the option. Source: seed Q33 · networking digest F.10 (Application Load Balancer / Network Load Balancer split, ≥ 2 AZs).

3.4 Domain 4: Billing, Pricing, and Support (12 %) · items 19 to 20

Teardown, Q19 · the biggest number loses. Why it is right: three constraints stack, cannot be interrupted, three years, one instance family in one Region, and the largest ceiling that survives all three is the 3-year EC2 Instance Savings Plan at up to 72 % (as of Oct 2026; verify current before use). Why the traps lose: A has the biggest headline percentage but Spot capacity can be reclaimed with about two minutes' notice, which the cannot be interrupted constraint forbids outright; B commits for one year while the workload runs three, and its ceiling is lower because Compute Savings Plans buy flexibility across instance family, Region and OS; D is the undiscounted baseline you only accept when nothing else fits. Trap: two traps fire at once, the largest percentage is disqualified by a constraint, and the more flexible product carries the smaller ceiling. Read the qualifiers before the percentages (worked example 3). Source: seed Q36 · pricing digest C1, F.1, F.4 · compute/storage digest C3, C5 (72 % / 66 % / 90 %, as of Oct 2026).

Teardown: Q20 · before any resources exist is the whole item. Why it is right: only the calculator estimates spend that has not happened yet, and only it works with no AWS account: the stem states both constraints. Why the traps lose: A analyses actual historical and forecast spend inside an account, so it has nothing to analyse on day zero; B alerts you when actual or forecast spend crosses a target you set: a target that requires spend to exist; D is the raw line-item export of bills already incurred. Trap: all four are cost words, so candidates pick the one they use most often. The four-way split is the whole domain in one line: Calculator = estimate before you build · Cost Explorer = explore and forecast actuals · Budgets = alert and act on a target · Cost and Usage Report = raw line-item export. Source: seed Q37 · pricing digest C12 · billing/organizations digest F.9.

3.5 Bonus items 21 to 22: case and 2026-update stems (outside the 5 / 6 / 7 / 2 count)

Set B's blueprint is 20 items in weight order; these two extension items are built from the Real-World Case Drills section and the 2026 Updates box. Grade them separately; they never count toward the 18/20 benchmark in section 6.

Teardown, Q21 · the story's noun, not the story's fame. Why it is right: the stem's only measurable is lead time (three months → minutes), and agility is the benefit that buys lead time. Why the traps lose: B is the aggregation argument, hundreds of thousands of customers lowering unit price, a pricing effect, and the stem never mentions price; C needs a component failure and Availability Zones to appear, neither of which is present; D needs a two-way demand response (a spike and a release), and a build-time story is not a scaling story. Trap: all four options are true statements about the cloud, which is exactly the point of section 1.2, the stem's qualifier selects, familiarity does not. A second trap hides in the number: the 70 %, 80 to 90 % and 50 % figures attached to this customer are published customer results, never AWS guarantees (case drill 1). Live-exam form: a domain-1 benefit stem wearing a real customer's story. Source: case drill 1 · digest 17 §D.1 case 1 and §F.1 (customer result ≠ guarantee).

Teardown: Q22 · the exam tests the guide you download today. Why it is right: option A reproduces the current guide's own example list (accessed Oct 2026), and all four names are current-plan names. Why the traps lose: B is the launch-era Version 1.0 wording almost verbatim: Developer, classic Business and Enterprise On-Ramp stopped taking new subscriptions after 2025-12-02 (existing ones run to 2027-01-01), so it is the single most likely stale-fact answer; C keeps only one retired name but still misses Business Support+ and Unified Operations; D mixes two correct names with a retired one and with AWS Free Tier, which is not a support plan at all. Trap: the older your notes, the more right these options feel; this is a date item, not a knowledge item. When a support-plan stem smells like recall from an old PDF, re-derive it from the current exam guide's Domain 4 text (as of Oct 2026). Source: digest 18 §A (S2 task 4.3 quote), §B.6 and §C23 · support-plans announcement 2025-12-02.

• 📚 Did you know? Every AWS account gets 56 AWS Trusted Advisor checks for free, and Business Support+ and above unlock 426 more, for 482 in total (the plan page also says "more than 500": as of Oct 2026). That matters for the trap drill in section 4: Trusted Advisor answers "does my configuration deviate from best practice", which is a third question again, distinct from Config's "what was the state" and CloudTrail's "who acted".

---

4. TRAP DRILL: the ten pairs, framed correctly

Ten pairs do most of the damage on this exam. The rule for this table is strict: the framing column must be sayable in one clause, because that is the only thing that survives contact with a 60-second budget. Read the row, cover the framing column, say it out loud, then check.

---

5. EXAM DAY: the official logistics checklist

Every row below comes from AWS's certification policy pages or the exam guide as retrieved in October 2026. Nothing here is folklore, and nothing that AWS does not publish is asserted: see section 6.4 for the list of those.

5.1 Before the day

5.2 Check-in

5.3 During the exam

5.4 After the exam

• 📚 Did you know? The exam interface gives you a review screen that can filter all, incomplete, or flagged questions, so a flag is an official, documented mechanic, not a trick. Combined with the fact that "unanswered questions will be marked as wrong" while "there are no penalties for incorrect responses", the only rational endgame is a sweep that finds zero blanks and changes an answer only when you can state the reason in one clause.

---

6. Grading Set B, and the ready-versus-not-ready verdict

6.1 The rubric

Worked example 7, turning a raw Set B score into a signal. $\frac{17}{20} = 0.85 = \mathbf{85\,\%}$, $\frac{14}{20} = \mathbf{70\,\%}$, $\frac{11}{20} = \mathbf{55\,\%}$. Those three numbers are study signals, never forecasts, the live cut is scaled, the 15 unscored items are hidden, and no published conversion exists. What the percentages are good for is comparison against yourself: Set B versus Set A, and this week versus last.

6.2 Reading your Set B score

Grade by domain block, not by total. A 17/20 that puts three of its misses into the 7-item Domain 3 block, that is $4/7 \approx 57\,\%$ in the exam's heaviest domain, is worse than a 15/20 spread evenly, because compensatory scoring punishes the dip, not the average. Then classify every miss as a knowledge gap (you did not know) or a reading error (you knew it and picked the distractor): knowledge gaps go back to the lesson, reading errors go back to section 1.2.

The three comparisons that actually decide your date:

6.3 The routing table: miss → lesson

6.4 What this lesson deliberately does not assert

A digest marked (G) is never a fact. These points are flagged rather than taught:

1. Partial credit on multiple response. The CLF-C02 guide says "two or more correct responses out of five or more response options" but never writes "select all correct responses" or "partial credit". AWS's all-or-nothing sentence comes from its AI Practitioner guide and a 2024 AWS blog post. Practise all-or-nothing; do not quote it as CLF-C02 text.
2. How many multiple-response items the live exam carries, and whether every stem states the count: unpublished. Set B's 4-of-20 split is a teaching choice, not an AWS figure.
3. Pass rates. AWS publishes none. Any first-time pass rate you read online is a community estimate.
4. Scratch paper at a test centre. The retrieved online rules list notes and paper as prohibited; test-centre rules defer to the Pearson VUE Candidate Rules Agreement, which was not retrieved here. Never claim it is provided, or banned.
5. On-screen pass/fail. Official text is only that the exit screen "provides information on exam results", while results take up to five business days.
6. Raw-to-scaled conversion, and per-objective weights. Only domain percentages exist; objectives carry no published weight; exam forms are equated.
7. Per-question pacing. 83 seconds is arithmetic (5,400 ÷ 65). AWS publishes no pacing guidance; the 45 to 60 second reading cap in a 2021 AWS blog is one author's practice.
8. Support-plan rendering. The HTML exam guide renders one plan name as AWS Business Support\+: a source-page artifact whose meaning is unconfirmed. Teach the current names, date-stamped, and re-check the support page before you sit.
9. Whether CLF-C02 adopts other question types (ordering, matching, case study). The current guide lists only multiple choice and multiple response.

---

7. The final 7-day study plan

This plan is weight-driven and dip-driven: it does not re-read the syllabus in order, it spends the week on the blocks that lost you marks and on the logistics you cannot afford to re-derive. AWS's own published rhythm for a Foundational certification is 2 to 3 weeks in 30 to 60 minute chunks: assume that work is done, and treat these seven days as discrimination practice.

Worked example 8, the hours. Six study days at 45 / 60 / 45 / 45 / 30 / 40 minutes is $45 + 60 + 45 + 45 + 30 + 40 = 265$ minutes ≈ 4.4 hours of focused work, plus 20 minutes on exam morning. That is deliberately small: hours spent after the content is learned are worth most when they are spent on discrimination (the ten pairs), arithmetic (RPO/RTO, ceilings) and rehearsal (the clock and the checklist), not on another pass through a video.

• 📚 Did you know? AWS states that "you need to achieve a passing score, not a perfect score". On a 20-item set that permission matters: 18 of 20 is 90 %, and it is a better signal than a nervous 20/20 achieved by spending four minutes on one item and eating your review window. The exam rewards coverage of all 65 items and zero blanks: depth on a single item you already had is worth nothing.