liveBook · Manning

Qualitative assessment

As explained in The Art of AI Product Development

Qualitative evaluation assesses a system through descriptive observations, judgments, and feedback rather than primarily numerical measurements.

Qualitative assessment evaluates a search system through observed user experience, manual judgments, and detailed feedback before teams rely on advanced metrics. It starts with quick vibe checks and can become more systematic as evidence and optimization needs grow. section 7.2.3

Useful approaches include:

  • User studies: Have people perform realistic tasks and describe whether results are relevant or require searching through irrelevant documents. section 7.2.3
  • Relevance assessments: Ask evaluators to judge whether results satisfy a query’s intent, creating structured feedback about what works. section 7.2.3
  • Task-based evaluation: Observe system performance while users complete live work, such as retrieving customer feedback for a case study. section 7.2.3

How The Art of AI Product Development covers it

How deep it goes
Progresses from manual vibe checks to user studies and structured judgments, while deferring advanced metrics until data and optimization mature.
Examples
Uses queries about advanced-plan feedback and API integration case studies, plus live retrieval of customer feedback for case studies.
What it assumes
Starts with limited data and postpones advanced metrics until the evaluation dataset and optimization effort grow.
What you can do after
Evaluate search usefulness, identify broad improvement areas, and establish a baseline for future changes.
How it connects
Connects search evaluation with user studies, relevance assessment, task-based evaluation, quantitative evaluation, and real-world evaluation.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of The Art of AI Product Development. The section links point to the parts of the book it draws on.