Operational economics 1.0.0

WorldSpec.cost_scenarios[] is optional. The generated output schema is schemas/operational-economics/1.0.0.json in the repository. A runnable example is fixtures/worlds/rural-clinic-economics.json; follow the resource cost guide.

Scenario inputs

Field	Meaning
id, title	Unique scenario ID and readable title. Maximum 20 scenarios per world.
metric_id	A declared registered alive.metric.count metric. Unit is events.
perspective	Whose resource costs the scenario considers; does not establish a payment mechanism.
resource_basis	Why each counted event maps to the assumed resource cost, including exclusions.
rate	value, low, high, unit USD/event, provenance, source_pointer, note. Requires finite nonnegative values and low <= value <= high.
additional_costs[]	Exactly one entry per world variant, with variant, integration, maintenance, alive_run.

Each additional cost is either null (unknown) or {value, provenance, source_pointer, note}. Its value is USD over the full run horizon; integration is included once, and maintenance must already cover that horizon. There is no automatic annualization, amortization or currency conversion. Rates and costs are bounded at 1e12. Source pointers and notes must be nonempty.

Provenance uses measured, public_evidence, expert_assumption, synthetic_assumption. A supplied estimate should use an assumption label. Provenance and pointers are supplied inputs, not an independent verification of the cost source. No default commercial rate is inferred.

Run outputs

GET /v1/runs/{run_id}/results returns operational_economics[] (empty for worlds without scenarios and for historical runs without this field). Each report includes the complete input scenario, horizon_days, warmup_days, currency: USD, metric_unit: events, the metric's evidence_grade, type: modeled_outcome, status: completed | partial, formula, sensitivity[], and limitations[]. Warmup events are excluded by the count metric; additional costs still cover the full horizon.

Each sensitivity point carries case: low | base | high, rate_usd_per_event, per-variant resource_costs and total_costs, and paired resource_cost_comparisons and total_cost_comparisons. Summaries use the same mean, SD, undefined count and Student-t interval fields as run metrics. Comparisons are treatment minus baseline in USD, with matched replication IDs. A positive difference means higher modeled cost. Fewer than two pairs have no interval.

For every replication:

resource cost = counted events × rate (USD/event)
total cost = resource cost + integration + maintenance + Alive run cost
difference = treatment cost − baseline cost

If any additional cost is unknown, that variant's total and its total comparisons are undefined. Unknown counts stay undefined even at a zero rate. Explicit zero costs are retained as inputs.

The ZIP includes evaluation/operational-economics.json when scenarios exist; its checksum is in manifest.json and checksums.sha256. It can be reproduced with the exported world and runtime.

Applicability and limitations

This version values counted operational events only. It does not price mean waits, ratios or staff utilization; a total staff-time adapter is still pending. Sensitivity varies the rate alone, with fixed additional costs. Intervals cover simulation noise, not rate uncertainty or evidence that the model represents a real organization. Mock effects remain assumptions. No realized savings, revenue, ROI, reimbursement or viability score is emitted.

See WorldSpec validation for malformed inputs and business assessment for buyer/value hypotheses that work without a world.