CCA Exam Day: What to Expect and How to Pace 60 Questions

On CCA exam day you have 120 minutes for 60 scenario-based questions, which is two minutes each, delivered by Pearson VUE either online (OnVUE) or at a test center; check-in involves ID verification and a room scan, and nothing but you and the screen is allowed. The strategy that works is a first pass at roughly 90 seconds per question, flagging anything uncertain, then a second pass on the flags.

This guide covers what the day looks like and how to handle the question types. Logistics change, so read Pearson VUE’s own OnVUE policy page and Anthropic’s exam page before your appointment; treat everything below as orientation, not the rulebook. Exam facts are as of mid-2026.

What happens at Pearson VUE check-in?

If you chose online proctoring, the day generally runs like this:

• Before the appointment. Run the system test Pearson VUE provides on the machine you will use. Do it days ahead, on the same network, so a webcam or firewall problem is not discovered with ten minutes to spare.
• Check-in window. You can typically start check-in shortly before the scheduled time. You will photograph a government-issued ID and take pictures of your workspace, or a proctor will ask you to pan the camera around the room.
• Workspace rules. Expect: no notes, no second monitor, no phone within reach, no other people in the room, and no leaving the camera’s view. Some policies allow water in a clear container; some do not. Read the current rule rather than guessing.
• During the exam. The proctor watches through the webcam and microphone. Reading questions aloud, covering your face, or looking away for long periods can trigger a warning.

At a test center you skip the room scan but bring the same ID, store belongings in a locker, and sit at a provided machine. Either way the timer starts when the exam launches, not at check-in.

The point of knowing this in advance is not the rules themselves; it is removing surprises so your first ten minutes of attention go to question 1, not to a proctor chat window.

How should I pace 60 questions in 120 minutes?

Two minutes per question is the average, but scenario questions are uneven: some are one paragraph, some describe a whole multi-agent system with three constraints. Pace in blocks instead of per question:

Checkpoint	Target question	Time elapsed
Quarter	15	25–28 min
Half	30	50–55 min
Three quarters	45	78–82 min
End of first pass	60	100–105 min
Second pass on flags	—	remaining 15–20 min

If you are ahead, do not slow down; bank the time for the flags. If you are behind at the halfway mark, start answering flagged questions with your best guess immediately rather than leaving them blank. Every question should have an answer selected before you move on, even if it is flagged.

Our free mock exam uses 90 minutes for 60 questions on purpose. If you can finish that with time left, the real 120 minutes will feel generous.

How do flag-and-return and multiple-response questions work?

Flag and return. The interface lets you mark a question and come back to it. Use flags for two cases only: you are torn between two options, or the scenario is long enough that you want to answer the shorter questions first. Do not flag questions you simply dislike; you will run out of second-pass time. On the second pass, re-read the constraint sentence (the “must”, “cannot”, “without adding latency” clause), because it usually decides between the two options you were torn between.

Multiple-response. Some questions ask you to select two or more options, and the stem says how many (“choose two”). Treat each option as its own true/false against the scenario. The trap is picking one obviously right option and one that is merely plausible; if the question says two, look for the two that address the same failure from different layers, for example a hook that enforces a rule plus a structured error that lets the agent recover. Partial credit is not something you should assume; select exactly the number requested.

Multiple-choice questions on this exam rarely have a “textbook” answer that is simply correct. They have four designs, and you pick the one that survives the constraints.

How do I pick the “least bad” answer?

The recurring pattern across all five domains is: which mechanism guarantees the behaviour the scenario needs? A quick reading discipline helps:

1. Find the hard constraint. Words like “must never”, “irreversible”, “regulatory”, “100% of the time” mean the answer lives in the programmatic layer, not the prompt. See The Model Will Override Your Instructions 4-15% of the Time and Select by Consequence AND Verifiability.
2. Find the symptom. “Wrong tool selected” points to descriptions or tool count (The Tool Count Curve). “Retried five times” points to uniform error messages (Uniform Error Anti-Pattern). “Reported no results after an outage” points to access failure vs valid empty.
3. Eliminate the option that just adds prompt text for a problem the prompt cannot solve, and the option that adds a whole new system when a smaller change fits.
4. Between the last two, pick the one that respects every stated constraint, even if it feels less elegant. Elegant but violates “no added latency” is wrong.

If you genuinely cannot decide, prefer the option that is explicit and verifiable (a hook, a schema, a structured error, a persisted facts block) over the option that relies on the model behaving well.

What should the last 24 hours look like?

• The day before: one light pass over the cheat sheet, then stop. Cramming articles the night before rarely moves a score; being tired reliably lowers one.
• Confirm your appointment time and time zone, your ID matches your registration name exactly, and the room you will use passes the workspace rules.
• Two hours before: eat, close every other application, restart the machine, run the system check again.
• During: read the constraint sentence twice, answer, flag if torn, keep to the block checkpoints. Do not change a first-pass answer on the second pass unless you can name the constraint you missed.

Next step

Rehearse the day with the free 60-question mock exam under real conditions: no notes, one sitting, timer on. If your flags cluster in one area, spend the remaining days there; Domain 1 is the largest slice and the one most candidates flag most.