business-problem-framing
Frame an ambiguous business request or a proposed solution as a decision question, bounded problem, testable issue tree and evidence plan. Use at project kickoff or when objectives, causes, scope or d
SKILL.mdskills/business-problem-framing/SKILL.md
GitHubBusiness Problem Framing
Turn a vague request into an agreed decision and the smallest analysis needed to resolve it. Preserve a sponsor's chosen scope; propose a reframing explicitly rather than silently changing the assignment.
Inputs and Missing Information
Start with the request, decision owner, deadline, affected business, baseline/target and known constraints. Ask only for missing information that changes the decision or feasible scope. If it is unavailable, produce a provisional brief with labelled unknowns rather than inventing a baseline, target or sponsor approval.
Read references/framing-methods.md when diagnosing solution-first requests, building a driver tree or choosing falsification tests. Copy assets/decision-brief.md and assets/hypothesis-plan.csv when the user needs editable deliverables.
Workflow
- Separate the stated request, observed symptom, claimed cause and proposed solution. An executive's diagnosis is a hypothesis unless supported by scoped evidence.
- Write the decision in plain language: who must choose what, by when, against which success measure and constraints? Include the status quo or least-change option. Keep an objective distinct from a candidate solution.
- Bound the problem: segment, geography, process, metric definition, period, exclusions and dependencies. Define a baseline and target only if supplied or explicitly agreed. Record metric changes that could create a false improvement.
- Build a decision-relevant issue tree. Use a driver identity when possible, such as revenue minus costs. Check overlap and missing causes; do not declare MECE merely because there are three branches. Keep causes distinct from interventions.
- Form competing hypotheses. For each, record supporting evidence, disconfirming evidence, a test, data/definition needed and the decision it could change. No evidence is not evidence of no problem.
- Prioritize work by decision impact, uncertainty and time/cost to test. Name an accountable role, a due date and a stop/escalation criterion. If the available time cannot resolve the question, narrow the decision or state the remaining exposure.
- Return a concise brief for agreement. Distinguish supplied facts, calculations, assumptions, hypotheses and unresolved choices. A drafted brief is not sponsor sign-off.
Output and Handoff
| Deliverable | Minimum useful content |
|---|---|
| Decision brief | Decision question, owner, deadline, baseline/target, constraints, options, in/out of scope |
| Issue tree | Explicit driver logic and treatment of overlap or omissions |
| Hypothesis plan | Rival explanation, evidence for/against, test, owner, deadline, decision consequence |
| Handoff | Agreed versus provisional scope, metric definitions, source IDs, priority evidence gaps |
The handoff can be pasted into an interview brief, market analysis, diligence sprint or decision memo. It does not depend on another installed skill.
Guardrails
- Do not promote a requested tool, restructuring or price change into the proven answer.
- Do not substitute generic frameworks for the actual business mechanism.
- Do not fabricate targets, causes, savings, sponsor agreement or decision authority.
- Keep material counterevidence visible and state what would reverse the leading hypothesis.
- Distinguish a feasible test from a causal conclusion; correlation and expert opinion alone do not prove causation.
- Do not collect private data, contact stakeholders or change systems without authorization.