Fishbone (Ishikawa) analysis for a multi-cause problem
intermediateClaude SonnetOperationsProject Managementframeworkmethodologyroot-causeoperationsproblem-solving
Use case
Use this when a problem has multiple plausible causes contributing simultaneously and a single root-cause analysis would oversimplify. The structure forces Claude to populate every category and then weight, instead of producing a balanced-looking diagram with no decision.
The prompt
You are running a fishbone analysis. The discipline is to populate every category honestly (resisting the urge to stuff everything in one), then weight which causes are doing the real work. <context> Problem (the "head" of the fish):{{problem}}Scope (where this problem shows up, where it does not):{{scope}}What has been tried or ruled out:{{prior_thinking}}Time horizon for fixing this:{{horizon}}</context> <task> Step 1 — Populate each category. For each of the six standard categories, list 2 to 5 specific contributing causes. Be specific. "Communication" is not a cause; "weekly cross-functional sync was canceled three weeks running" is. - People: skills, roles, training, motivation, attention - Process: workflows, handoffs, sequencing, dependencies - Tools/Tech: software, integrations, automation, gaps - Environment: physical or organizational context, time pressure, distractions - Inputs: data quality, materials, upstream signals - Measurement: how performance is measured, what is visible vs. invisible If a category is genuinely empty for this problem, say so explicitly rather than inventing a cause. Step 2 — Score each cause. For every cause, score: - Likelihood it is contributing (1 to 5) - Impact if removed (1 to 5) - Cost to address (1 to 5, where 5 = expensive) Combined score = (Likelihood × Impact) / Cost. Step 3 — Identify load-bearing causes. Highlight the 3 to 5 causes that score highest. Note which categories they cluster in. If they cluster in one or two categories, that is signal — the problem may not be multi-cause after all. Step 4 — Interventions. Propose one specific, testable intervention per load-bearing cause. Each must be deliverable within{{horizon}}. Step 5 — Sequencing. Recommend the order. Start with the cause whose intervention also weakens other causes (a "leverage move"), even if it is not the highest-scoring on its own. </task> <output_format> ## Fishbone (six categories, with causes) ## Scoring table ## Load-bearing causes (and clustering observation) ## Interventions ## Recommended sequence </output_format> <constraints> - Each cause must be specific enough that someone reading it for the first time knows what to look at. "Bad communication" is too vague. - If you find yourself stuffing every cause into one category, the problem may be single-cause. Note that and run 5 Whys instead. - Do not invent causes to fill out empty categories. Better to say "Measurement: no contributing causes identified." - Cost scores should reflect organizational cost, not just dollar cost. Re-organizing a team is expensive even if it costs no money. </constraints>
Variables
{{{{problem}}}}Replace with your {{problem}}{{{{scope}}}}Replace with your {{scope}}{{{{prior_thinking}}}}Replace with your {{prior thinking}}{{{{horizon}}}}Replace with your {{horizon}}
Example
Input
problem: Our customer onboarding NPS has fallen from 62 to 38 over two quarters scope: All segments; worse in mid-market; not seen in self-serve prior_thinking: We assumed it was the new onboarding flow that shipped Q1; rollback didn't restore the score horizon: 90 days
Output
## Fishbone **People** - CSM team turnover: 4 of 11 CSMs are <90 days tenured - New onboarding lead reorganized handoff but did not retrain frontline - Product specialists deprioritized onboarding sessions in favor of expansion calls **Process** - Sales-to-CS handoff doc is filled out inconsistently (~40% complete based on audit) - Kickoff scheduling slipped from "within 5 days" to "within 14 days" median - No defined "stuck" check-in between weeks 2 and 4 of onboarding **Tools/Tech** - Onboarding portal launched in Q1; mid-market customers report 3+ confusing steps - Notification flow does not surface stuck customers to CSMs proactively **Environment** - Sales is being pushed harder on Q3 quotas; deals close with thinner CS prep - Q2 reorg moved CSMs under sales-aligned pods; cultural reset still ongoing **Inputs** - Sales-provided customer goals are vaguer than they used to be (30% of handoffs lack a measurable goal) - Pre-kickoff customer data is incomplete in 25% of cases **Measurement** - NPS is measured at day 30; we do not measure leading indicators (time-to-first-value, # of stuck checkpoints) ## Scoring (top causes only) | Cause | Cat | L | I | Cost | Score | |---|---|---|---|---|---| | CSM tenure | People | 5 | 4 | 4 | 5.0 | | Handoff doc inconsistent | Process | 5 | 4 | 2 | 10.0 | | Kickoff slip 5→14 days | Process | 4 | 5 | 2 | 10.0 | | Sales goals vaguer | Inputs | 4 | 4 | 3 | 5.3 | | No stuck check-in | Process | 4 | 4 | 2 | 8.0 | | No leading indicators | Meas. | 5 | 3 | 2 | 7.5 | ## Load-bearing causes Three of the top five are in **Process**, with one each in Inputs and Measurement. Clustering signal: the problem is mostly process, not people or tools, despite the team's instinct to blame the new portal. The portal is contributing but is not load-bearing. ## Interventions - Handoff doc: enforce required fields; deals cannot move to "closed" until handoff is complete. Implement in 2 weeks. - Kickoff slip: SLA reset to 5 days; weekly dashboard publishes median to leadership. 1 week. - Stuck check-in: add a Day-14 check-in to the playbook; CSM owns it. 1 week. - Leading indicators: instrument time-to-first-value and stuck-checkpoint count; report alongside NPS. 3 weeks. ## Recommended sequence Start with the handoff doc fix. It is the cheapest, and it directly improves three other causes (kickoff prep quality, vague sales goals, stuck check-in usefulness). Then the SLA, then the check-in, then instrumentation in parallel.
Tips for best results
- 1The clustering observation is what makes this stronger than the typical fishbone — without it, the diagram looks balanced and the team does not realize the problem is concentrated in one or two categories.
- 2If every category is full, double-check. Real problems usually live in 1 to 3 categories; symmetric fishbones are often padded.
- 3The 'leverage move' sequencing rule is high-value. The intervention that weakens multiple causes at once is almost always the best place to start, even if it scores lower individually.
- 4Pair with five-whys-root-cause if the load-bearing causes all live in one category. Fishbone surfaces multi-cause problems; 5 Whys digs into single-cause ones.
Related prompts
Disciplined 5 Whys with explicit causal chain
intermediateRun a 5 Whys analysis that follows a real causal chain and refuses to stop at 'human error,' surfacing the systemic cause and the testable countermeasure.
Operationsframeworkmethodologyroot-cause
MECE breakdown of a problem space
intermediateDecompose a problem into Mutually Exclusive, Collectively Exhaustive buckets so analysis cannot double-count or quietly skip a branch.
Personal Productivityframeworkmethodologydecomposition
SWOT with rigor — weighted, evidenced, decision-driving
intermediateRun a SWOT that is more than a bullet list — every entry is evidenced, weighted, and tied to a strategic move, with explicit cross-quadrant interactions.
Operationsframeworkmethodologystrategy
Need help implementing this prompt in your workflow?
Book a call