Module 3 • Lesson 22• 45 mins
Balance Recall and Precision When Defining Retrieval Quality
Understand how top-k and similarity threshold shape retrieval, and write a requirement stating whether to lean recall or precision based on business consequences.
Explain how top-k and similarity threshold each shape retrieval results
Analyze the recall vs. precision trade-off through concrete business consequences
Write a retrieval-quality requirement framed around error consequences, not tuning parameters
Balance Recall and Precision When Defining Retrieval Quality
Once ingest and chunking are done, the system has a store of chunks. When a user asks a question, retrieval has to pick which chunks are "most relevant" to hand to the AI. Two parameters govern that choice: