Domain 7 — Troubleshooting and Optimisation (10%)
10% of the exam, roughly 6 questions. The smallest domain, and the most mechanical — diagnose the cause, apply the matching fix.
7.1 Diagnosing a poor result
| Symptom | Usual cause | Fix |
|---|---|---|
| Generic, could-be-about-anything output | No audience, context or purpose given | Add who it is for, what it is for, and what makes your situation specific |
| Wrong format or length | Format was never specified | State the structure and length; better, supply an example of the target |
| Confident but wrong facts | Claude was asked to recall rather than read | Supply the source material and restrict it to that; require citations |
| Forgets earlier instructions in a long chat | Conversation has grown past comfortable working memory | Start fresh with a summary, or move the standing context into a Project |
| Ignores one instruction among many | Instructions are buried, or two of them conflict | Separate them, put the critical ones prominently, and remove the contradiction |
| Right content, wrong tone | Register never specified | Name the register and give an example of house voice |
| Shallow treatment of a big multi-part request | Too much asked at once | Decompose into ordered steps with a check between them |
| Declines a legitimate request | Ambiguous phrasing that reads as something else | Restate the professional context and purpose plainly |
The default wrong answer in this domain is "regenerate" or "switch to a more capable model". Both occasionally help; neither is a diagnosis. The exam wants the cause identified first, and a model change is the right answer only when the task genuinely exceeds the tier you chose.
7.2 Working efficiently
- Reuse what worked. A prompt that produced a good result is an asset — save it, or promote it into Project instructions if the task recurs.
- Correct rather than restart. Targeted follow-ups keep the parts that were already right.
- Front-load the context. One well-built prompt beats six rounds of clarification, and costs less of your usage allowance.
- Right-size the model. Routine work on a heavier tier is slower and consumes limits faster for no gain.
- Keep conversations focused. A new topic belongs in a new chat; long unfocused threads degrade and are hard to reuse.
7.3 Improving over time
Individual fixes are worth little if they stay in one person's head. The pattern the exam rewards is capturing what works: a shared library of prompts for recurring tasks, Projects that encode the corrections people kept having to make, and a habit of noting which tasks needed heavy review so the workflow can be adjusted. Feedback from the people receiving the output is part of this loop, not an afterthought.
← Previous: Domain 6 — Governance, Risk and Responsible Use | Next: The 12 anti-patterns →
Explore Other Certification Study Guides
Switch to another exam guide to prepare for your next Claude certification:
Claude Certified Architect — Foundations
Core architectural principles: agentic loops & orchestration, tool design & MCP, Claude Code workflows, prompt engineering, and context reliability.
Claude Certified Architect — Professional
Advanced enterprise architecture: solution design, system integration, evaluation frameworks, governance, safety & developer enablement.
Claude Certified Developer — Foundations
Build software on Claude: API integration, model selection, tools & MCP servers, agents vs workflows, security, safety & debugging.