MODEL EVIDENCE · 19 OCTOBER 2026

Cost, vaccine loss and value.

English guide to the frozen V3 cost and weather evidence. All scenarios are synthetic. Prices, annual frequency and payer allocations contain explicit estimates or assumptions. This guide does not change the audited source results.

What the calculations support

Within declared demand, roads, household rules and prepared resources, the model identifies planned dose obligations retained and remaining gaps. Independent audits check the plans. The cost layer then prices the model’s actual activity quantities using INR_ESTIMATE.

No comparative reduction in real vaccine discard, procurement payments, disease burden or observed child coverage has been established. Verified savings are not deducted by default. This accounting choice does not prove that actual savings are exactly zero. Full immunization funding needs remain unknown.

Annual cash and service

Central prices; one standalone deployment; 48 illustrative rounds: 36 normal, six delayed and six road closures. The normal reference is frozen before the same disruptions are applied. The fixed policy cancels infeasible routes and retains feasible ones; it is not verified local emergency practice. These are separate case deployments, not budgets to add together.

Repeated planned obligations are not unique children or completed vaccinations. Run cash, first-year cash, subsequent steady-year cash and economic cost are distinct. Economic costing includes existing staff time and depreciation; cash includes actual purchase allowances. Later steady years precede future replacement or rebuild costs.

Platform sharing and comparison choices

The central development allowance is 160 hours × INR500 = INR80,000. Annual maintenance and hosting total INR27,600. Sharing scenarios allocate 80% of development, 50% of maintenance and all hosting across three or ten deployments; local adaptation remains. These are assumptions, not contracts.

C1 and C2 show no extra planned annual service over the independent simple rule in this frozen comparison. C3 shows 300 additional annual obligations over that rule, versus 660 over the fixed plan. Use the denominator for the selected reference. Negative cash differences within a shared-platform comparison are model estimates, not observed savings.

Do not confuse unmet demand with vaccine loss

• Undelivered stock may remain available at source.
• Unplanned or uncompleted service is not automatically vaccine discard.
• A temperature or protection-time exception requires quality assessment; it does not automatically authorize disposal or release.
• Confirmed discard requires an explicit authorized event; repeated events must not deduct stock twice.
• Unopened returns and open-vial remnants need product, opening, attendance and disposition records.

The household 60-minute target and 90-minute acceptance limit are travel-policy assumptions, not vaccine-stability limits. Product-specific handling and retention requirements need verified equipment and product evidence.

Inventory value is separate from purchasing cash

All loss-rate controls are sensitivities, not predictions or achieved effects. The reference pools repeat the same normal product-specific capacity over 48 rounds: C1 3,840 nominal doses, C2 2,400, C3 12,480. A five-percentage-point reduction from an assumed 15% baseline means 15% to 10%.

At an assumed INR90/dose, C3’s five-point sensitivity preserves 624 dose equivalents valued at INR56,160. Cash credit requires all of that stock to replace genuinely avoidable paid procurement. Stock used to expand potential service receives no purchasing cash credit and cannot also be added to service already counted by the model. Whole-vial rounding, future purchasing caps and confirmed order reductions matter.

Preserved quantity = reference doses × loss-rate reduction. Inventory value = preserved quantity × assumed product price. Purchasing credit = value × share used to reduce purchases × avoidable payment fraction, subject to procurement limits. Scoped net cash = added cash − eligible purchasing credit + separately priced clinical increments. Unknown full clinical and procurement costs remain unknown.

Weather and resilience: same tasks, existing resources

These are single-run results. Heat configurations are declared synthetic stress tests, not field qualification at 45–48°C. Changing cooling packs also changes weight, usable vaccine capacity, preparation demand and costs. Protection time is an explicit assumption rather than an extrapolated performance curve.

C3 heat + delay retains 150 / 110 / 130 under fixed, rule and bounded optimization. Keep the audited 150-task fixed plan available; no optimization improvement or global-optimum claim is made. Zero net cold supply or a shared supply failure produces zero planned service with unmet demand retained. Missing equipment or net-supply evidence instead returns unassessed service, not zero. Every case currently has one known source; no independent second-source comparison has been established.

The separate recovery example accounts for an additional 9km and 77 driver minutes, with a known incremental cash lower bound of about INR278. Full rescue cost is unknown. The original loading clock is not reset, stock is not reissued, and accepting a recovery commitment does not record completed vaccination.

Price references and remaining assumptions

Equipment price references do not establish a technical match to the simulated kit. Training references are historical, while inflation, wages, mileage services, maintenance and hosting are budgeting assumptions. International procurement prices are not local delivered quotations. The modeled DEMO products lack a verified SKU mapping. Tariff eligibility, payer commitments, field acceptance and local operating responsibility still need verification.

Download all price bands, English assumptions and source links ↓

Calculation and audit scope

The core solver keeps its DEMO_UNITS budget and accessibility/fairness objectives. The separate cost module performs post-solve INR_ESTIMATE accounting for compare, improve and recover. It does not optimize routes under a rupee budget. Annual workload assumptions do not automatically update from the custom-demand or temperature controls.

The original V3 delivery has its own algorithm and cost verification record. This website release separately verifies source bindings, unchanged Python bytes, interface behavior and public assets. Some C3 searches remain incomplete. Synthetic success does not establish field effectiveness, clinical benefit, realized loss reduction or a universally acceptable investment threshold.

Annual results CSV ↓ · Loss/value sensitivities CSV ↓ · Interface and publication scope →