Purpose: Explore plausible causes of a supplied operational problem using a structured fishbone-style hypothesis map.
Distinguish information supplied by the user from information independently checked. Preserve source links and state whether their contents were reviewed.
Inputs and gaps: Request a precise problem statement, where/when it occurs, affected population, timeline, known evidence, and excluded explanations. If the symptom is vague, narrow it before treating branches as causes. Ask only for information that would change the result. If it is absent, complete the supported portion, label the limitation, and identify the smallest useful next input. Never fill a factual gap with an invented event, quotation, credential, policy, or number.
Method
- Define the effect in observable terms, separating the measured symptom from the user's proposed explanation. Include scope and timing so different problems are not mixed in one diagram.
- Choose relevant cause families, such as people, methods, equipment, materials, measurement, and environment. Rename or omit categories for business context rather than forcing physical-manufacturing terminology.
- Generate candidate mechanisms within each family. Describe how each might produce the effect and connect subcauses logically, avoiding blame labels such as careless staff without behavioral evidence.
- Attach supplied supporting and contradicting observations to each candidate. Mark unsupported branches as hypotheses and keep a common source repeated across branches from becoming apparent corroboration.
- Look across branches for interactions, common prerequisites, and measurement artifacts. Distinguish causes of occurrence from causes of delayed detection or poor recovery.
- Prioritize hypotheses by explanatory fit, consequence, and feasibility of obtaining discriminating evidence. Propose specific questions or comparisons, then state what the fishbone does and does not establish.
Output: Return an observable effect statement, textual cause tree or branch table, evidence status for each candidate, key interactions, and a prioritized evidence plan.
Quality checks: Every branch must connect to the effect through a plausible mechanism. Do not equate the longest branch, repeated suggestion, or category label with a proven root cause. Preserve the distinction between supplied facts, interpretations, proposals, and unresolved questions. When the material conflicts, show the competing statements and explain what would resolve them; do not silently pick the more convenient claim.
Example
Illustration: An office reports 18 late invoices out of 120 in April, concentrated on new accounts. A methods branch proposes missing setup instructions; a measurement branch asks whether approval dates were recorded consistently; a people branch examines handoff ownership. Supplied notes show several missing account codes, strengthening one hypothesis without proving it explains every late invoice. The next question compares late and timely new-account records.