skills/business-analysis/references/analysis-method.md

# The analytical engine

Apply these layers to the actual question; do not turn every short answer into a five-part report.

| Layer | Work | Evidence that matters |
| --- | --- | --- |
| 0 · Structural test | Identify the objective, mechanism, counterfactual, binding constraint, and time horizon. | An observation that distinguishes the explanation from a plausible rival. |
| 1 · Model selection | Retrieve a few models that materially change diagnosis; choose a counter-lens. | Conditions under which each model holds, fails, or becomes redundant. |
| 2 · Business architecture | Trace value, enabling capability, distribution, and financial capture. | Buyer behavior, workflow, costs, pricing, retention, and ownership of the interface. |
| 3 · Strategic system | Map relevant moats, feedback loops, power, and bottleneck changes. | Persistence under substitution; reinvestment; gatekeepers; resolved, active, and emerging constraints. |
| 4 · Decision | Compress the mechanism into an actionable conclusion. | Owner, next commitment, review trigger, and action if the thesis fails. |

For moats, test brand, switching costs, network effects, scale, data advantage, embedding, and institutional position where relevant. A claimed asset is not automatically a moat; ask what makes it persist and who captures its economic benefit.

For a flywheel, identify the input, mechanism, output, reinvestment, friction, and weakest link. Distinguish a reinforcing loop from a sequential funnel.

For a bottleneck, identify the constraint or jointly binding set for the specified objective and horizon. Explain what moves when it is relieved and which constraint may bind next. Keep a possible next bottleneck separate from an observed current one.

For organizational or agentic work, identify the accepted outcome, decision rights, exception handling, owner, and operating boundary. A model, instrument, role, mode, and artifact are different objects; a role description does not prove delivery capability.

A useful decision record contains a claim, source and date, assumption, prediction, counter-model, confidence basis, review trigger, and conditional action. Retrieve the corresponding instrument rather than making up a universal score.