Procedural Architecture

Design a procedure around decisions, state and evidence rather than a list of topics.

Workflow

1. Establish the goal, audience constraints, starting state and observable completion. For a simple procedure, record only audience details that change the steps.
2. Identify each decision and the information/authority needed before it. Identify stage entry conditions, exit artifacts and responsibility for handoffs.
3. Separate the dependency graph from control flow. Prerequisites must be satisfiable; retry/rework loops are valid when guarded, bounded or convergent, with a clear failure/exit path.
4. Check branch coverage, reachable completion, dead ends and invariants across failure/cancel/retry paths. A branch should represent a real decision, not a cosmetic alternative.
5. Set granularity to the audience and stakes. Split stages when they have different readiness/ownership; merge stages that add no independent outcome.
6. Where contention or arrivals matter, model resources and validate assumptions using queueing or simulation. Do not infer capacity from a tidy flow diagram.
7. Review against concrete counterexamples. A critic may agree after checking; disagreement and finding counts are not success criteria.

Evidence

Deliver a flow or short stage table with inputs, decisions, outputs, owner and failure/exit conditions. Include tested walkthroughs, unresolved assumptions and any capacity calculation's units/model/parameters. Use larger models only when they clarify real concurrency or state.

Optional references