From symptoms to diagnosis

Lessons 10 through 15 introduced three specialized retrieval strategies, hybrid, graph, and corrective RAG, plus the metrics (Lesson 15) that would tell you, empirically, whether any of them actually helped. The skill this tier is building is not memorizing the three names, it is the diagnostic habit of reading a failure and knowing which fix it points to.

A worked diagnostic pass

Suppose users report that a support-document RAG system frequently misses the exact product model they asked about, even though a document about that exact model exists in the knowledge base. First question: is this a precision problem (wrong chunks retrieved) or a recall problem (right chunk never found)? Testing shows the correct chunk simply never appears in the top-k list, so it is a recall problem, specifically one where an exact model number is not being matched. That symptom, exact terms failing under semantic search, is precisely the gap hybrid RAG (Lesson 11) closes, not graph RAG, which would not help here since there is no relationship-traversal question involved.

Practice the same pass yourself

Given: a RAG system correctly retrieves relevant chunks (high context precision and recall by the Lesson 15 metrics) but its answers still sometimes state facts that are not in any retrieved document. Which stage is failing, retrieval or generation, and which technique from Lessons 11 to 14 (if any) actually addresses this specific symptom versus which one would not help at all?

Try this yourself

  • Work through the practice scenario above: name the failing stage, and explain why hybrid RAG and graph RAG would not fix it while a faithfulness check (closer to what CRAG's grading step provides) would.
  • Describe a realistic support or documentation scenario where graph RAG would be the correct fix and hybrid RAG would not help at all, and vice versa.
  • For a knowledge base you are familiar with (work, school, a hobby project), name which of naive, advanced, hybrid, or graph RAG best matches its actual structure, and why.

If you can make these diagnoses confidently, you're ready for the Advanced tier, starting at Lesson 17.