Guides

Multiple-response and case-format questions

Updated 20 Nov 2026. Unofficial guide, checked against the official exam guides.

Two item formats need their own technique: items where you pick more than one option, and sets of items that share one long scenario. Neither is harder in content than an ordinary item. Both punish habits that work fine on single-answer questions.

What the guides say

Every exam guide says the same thing about format: each item states how many responses to select. The Developer Professional guide and the version 1.0 guides name both multiple-choice and multiple-response items. The 7 Dec 2026 guides describe scenario-based multiple-choice items, each stating how many to select. Anthropic’s FAQ says all the exams use multiple choice and scenario-based multiple response.

Your form may have few of them. Szymon Paluch reports that every item on his Architect Foundations form had four options and one correct answer, and adds that the guide allows multiple-response items. Prepare for both.

Multiple response: assume no partial credit

The guides don’t say whether a half-right selection earns anything. When a rule isn’t published, prepare for the strict version: an item counts only if every option you pick is correct and you have picked the number asked for. The free forms here score multiple-response items that way.

That changes how you work through the options.

Judge each option on its own. Go through them one by one and decide true or false for this scenario, as if each were a single-answer item. Don’t look for the pair that “goes together”. Item writers can make two wrong options look like a natural pair.

Use the count as a check, not a guide. If the item says choose two and you have three trues, one of them is true in general but doesn’t address this scenario. Find it. If you have only one, reread the scenario for the constraint you skipped.

Watch for one idea at two strengths. Two options can describe the same fix, one scoped and one heavy-handed: stop the affected route, or stop all traffic. Only one of them can match the scope the scenario describes. If both made your list, you have counted one idea twice.

Don’t pick an option just to hedge. With all-or-nothing scoring, a hedge pick doesn’t protect you. If it’s wrong, it costs you the whole item.

Scenario blocks on Architect Foundations

Architect Foundations is built from scenarios. The guide publishes six, and each sitting draws four of them at random:

  • a customer support resolution agent
  • code generation with Claude Code
  • a multi-agent research system
  • developer productivity with Claude
  • Claude Code for continuous integration
  • structured data extraction

John Weidner describes his form as four blocks of fifteen items, each block built on its own production scenario.

Because the six scenarios are public, you can prepare for each one in advance. For each, work out its likely constraints: what must never happen, what has a latency target, where a human has to stay involved. Then the setup on the day only needs a careful read, not a fresh analysis.

On the day, budget about 25 minutes per block and keep 20 minutes for review. Read each scenario’s introduction once, carefully, and note its hard constraints before the first item. Those constraints decide many of the items in the block.

Case sets

Some practice material, including the Developer Professional bank here, groups items into case sets. A case is a fictional company described in 150 to 300 words: its product, architecture, scale, constraints and current metrics or incidents. Five or six items follow, in order.

The Developer Professional guide describes its items as scenario-based. It doesn’t spell out how scenarios are grouped on the exam, so treat case sets as practice in reading long scenarios fast, not as a description of the real form. The skill carries over either way:

  • Read the question before the case. Then read the case looking for what that question needs. On the next item in the same case, go back for different lines.
  • Keep a running list of the hard facts. Volumes, targets, versions, what was changed last week. If Pearson gives you scratch material, write them down once.
  • Expect some arithmetic. The Developer Professional cases on certdrill include items that need a quick calculation from the case’s figures, such as cost per request, a latency budget, or agreement between a judge and human labels. The guide’s evaluation objectives cover judge calibration and retrieval precision, so that kind of reasoning is in scope.
  • Don’t carry an answer from one item to the next. Each item in a case is scored on its own. A wrong reading in item two only spreads if you build on it in item three.

Try both formats

Every free form includes two multiple-response items. The Developer Professional free form also includes two complete case sets of six items each, with the case shown in a panel above each question.

All guides Changes log