Failure Class

A production problem named by what the user sees, not by which component is wrong.

Why it matters in a memory layer

A retrieval failure is usually diagnosable from one artifact — the ranked list. The document is missing or it is ranked low, and either way "why didn't it come back?" is answered by inspecting one thing.

Here the same symptom arrives from different stages. "A fact I told you is never recalled" has three causes — never extracted, extracted but unnameable so never arbitrated, or demoted out of the retrievable tier — living in three modules, and the store looks identical from the outside in all three. The read path is the last place to look, and the only place a retrieval intuition suggests looking.

Five of the seven classes in this course's guide have more than one cause. Organise by symptom, because the symptom is what arrives; a stage-indexed taxonomy is only usable by someone who already knows the answer.

Connections

Taught in: The Failure Field Guide

Do not confuse with: Differential Diagnosis