Sensitivity and Scenario Analysis

Sensitivity and scenario analysis examine how a mining decision performs when its assumptions change. Sensitivity analysis examines how inputs affect results; scenario analysis evaluates combinations of conditions that form a coherent alternative future. They help identify fragile assumptions, compare responses and direct further investigation. Neither method, by itself, establishes the probability of a financial outcome. This distinction is consistent with the treatment of sensitivity and scenarios in HM Treasury's 2026 Green Book, which is UK public appraisal guidance rather than a mining valuation standard.^[1]

Establish the Decision and Reference Case

Start with a question that the analysis can answer: whether an expansion remains attractive, which development sequence is preferable, or how much additional funding a delayed start could require. Name the output being tested. For this article, peak funding requirement means the greatest cumulative cash shortfall without discounting. State opening unrestricted cash and which financing flows are included; label any shortfall after financing as a residual funding need. Net present value instead discounts net cash flows to a valuation date. An assumption can affect the two measures differently.

The reference case should have a reproducible mine plan, processing assumptions, investment schedule and cash-flow model. Record the valuation date, reporting currency, real or nominal price basis, tax treatment, financing perspective, discount convention and closure assumptions. Keep these conventions consistent across comparisons unless a convention is itself the stated test. A comparison between a nominal after-tax case and a real before-tax case cannot isolate the effect of commodity price.

Assign a version to the model and its inputs, and retain the original reference results. NASA's risk-analysis guidance emphasizes a common, configuration-controlled dataset and coordinated analysis methods when comparing alternatives. Applied to mine economics, this means that geology, scheduling, processing and financial teams must know which versions generated a sensitivity result.^[2]

Check the reference model before stressing it. Confirm that production quantities agree with the schedule, receipts follow the modeled product and payment terms, and expenditure and closure cash flows occur in their intended periods. A sensitivity chart can make an inconsistent model look orderly; it cannot validate that model.

Parameter tests do not capture every modeling uncertainty. NASA distinguishes uncertain inputs from uncertainty in model structure. For mine economics, compare credible alternative relationships where, for example, a constant unit cost may cease to represent capacity limits.^[2]

Test Individual Assumptions Deliberately

A conventional one-factor-at-a-time test changes one specified input, holds the remaining input settings fixed and recalculates dependent outputs. Raising electricity's unit price therefore changes the bill. A throughput test must specify which operating costs and production quantities respond.

The Joint Research Centre explains that fixed-reference tests miss interactions and can give different rankings elsewhere in the input space. They diagnose model responsiveness; they do not describe the complete uncertainty.^[3]

Justify ranges using testwork, quotation boundaries, market assumptions or an explicitly labeled stress condition. Record their evidence and physical bounds. Equal percentage changes compare responsiveness but do not establish equal uncertainty in the inputs.

A tornado chart ranks effects for the selected input endpoints, reference case and output. Show those endpoints and units. Its longest bar identifies the largest tested effect, without establishing an occurrence probability.

A switching value identifies where a stated decision criterion changes, such as where an alternative overtakes the reference option or its NPV reaches zero. The Green Book uses this concept in public appraisal; in a mine model, the analyst must specify the financial criterion and assumptions being held fixed. A switching price is a conditional threshold, not a commodity-price forecast or a probability of project failure. Nonlinear responses can produce several crossings or none; search only within a feasible range.^[1]

Construct Coherent Scenarios

A scenario should explain why its assumptions belong together and how they unfold over time. A temporary commissioning disruption differs from permanently lower processing capacity. Likewise, a short price downturn during construction differs from the same downturn after capital has been recovered. Specify the affected periods, any recovery trajectory and the decision responses allowed.

For example, consider a scenario involving a weaker local currency. Identify the currency of product receipts, imported equipment, consumables and local expenditure before adjusting the cash flows. Avoid assuming that every cost moves with exchange rates. Similarly, a commodity-price scenario may include a separate energy-price assumption, but their relationship needs an evidential or explicitly hypothetical basis; it is not a universal correlation.

Mining scenarios also need engineering consistency. Research on simultaneous stochastic mine planning treats extraction, blending, processing, material destinations and capital investment as connected decisions. This supports examining how changes propagate through a mining system rather than adjusting financial totals independently. The cited framework addresses mining complexes; it does not establish that every mine needs that particular optimization method.^[4]

Use a small set of distinguishable scenarios with a clear purpose. Combining every adverse endpoint can serve as an extreme stress test if the combination is physically possible, but it should be labeled accordingly. Calling it a probable downside requires separate evidence about likelihood and dependence.

Separate Exposure from Adaptation

Two questions require different calculations. Fixed-plan evaluation asks how the existing design and decision rules perform under changed assumptions. Reoptimized evaluation asks whether a different plan would perform better under those conditions. Report them separately. Otherwise, a favorable adjustment can conceal the original plan's exposure.

The mining optimization literature explicitly includes capacity, extraction access, processing flows and destination constraints. Accordingly, a revised economic case should retain the relevant physical restrictions: an attractive spreadsheet result is insufficient if its production sequence cannot be mined or its feed cannot be processed.^[4]

State when new information becomes available and which decisions remain reversible. A plan optimized separately with complete knowledge of every future outcome can provide a comparison benchmark, but it overstates an implementable response if the operator would not possess that information. An actual adaptation may require development time, extra capital, approvals, transition losses or contractual changes. Include those consequences before attributing value to flexibility.

Qualitative Example: A Harder Ore Domain

Consider a hypothetical mine where additional testwork indicates that an upcoming ore domain requires more grinding energy per tonne to reach the reference grind. No project data or numerical outcome is assumed here. The example illustrates how to structure the analysis, not how any particular plant would respond.

First, isolate the assumption. A controlled test might change the energy requirement per tonne while retaining the reference ore sequence. The model must state whether it assumes that the installed mill can maintain throughput and product quality. If that assumption is unverified, the result describes a conditional cost exposure rather than a demonstrated operating case.

Next, build a technically reviewed scenario. With available mill power constrained, the processing team might identify lower achievable throughput or a different operating point. Recovery and product quality require reassessment at that point; they should not be held at values that depend on an incompatible grind. The planner then checks feed availability, blending, stockpiles and the timing of receipts. Fixed site expenditure may continue while less product is sold.

Finally, compare responses such as changing the feed sequence or modifying processing capacity, where feasible. Each response needs its own schedule, investment and operating consequences. If production extends, revisit sustaining expenditure and the timing and scope of closure obligations. The conclusion should identify which assumption drives the decision and which additional test could distinguish between the alternatives.

Distinguish Sensitivity from Probabilistic Risk

Probabilistic analysis adds a representation of likelihood. Monte Carlo simulation repeatedly samples uncertain inputs and propagates them through the model. NASA's handbook explains why connected outcomes should be generated within a common analysis framework so that dependencies are preserved. It also distinguishes variability from uncertainty caused by incomplete knowledge, a useful distinction when deciding whether further investigation can reduce uncertainty.^[2]

For a mining application, justify the distributions, their bounds, dependencies and time behavior before reporting percentiles. Independently sampling annual prices, ore grades and recoveries may produce combinations that contradict the chosen geological or processing model. A statistical fit alone does not establish that the relationships will persist throughout mine life.

Global methods explore sensitivities across the input space; some quantify interaction effects. Results depend on the chosen ranges and distributions.^[3] Simulated downside probabilities consequently remain conditional on the model and its evidence. If credible likelihoods cannot be assigned, use transparent scenarios.

Keep an inventory of where risk has already entered the evaluation. A cost allowance, explicit delay scenario, adjusted recovery assumption and changed discount rate may partly address the same concern. Check their rationale together before combining them. Conversely, removing an allowance is justified only when the revised model actually represents the exposure that allowance covered.

Turn Results into Decisions and Follow-up

Report the assumptions changed, their rationale, the engineering response and the resulting financial effects together. Retain both absolute outputs and differences from the reference case. For timing risks, show period cash flows and funding needs alongside NPV; a small change in total value can accompany a difficult temporary funding gap.

Distinguish three conclusions: a decision remains acceptable across the tested conditions; its preference changes near an identifiable threshold; or unresolved information prevents a defensible comparison. These are statements about the tests performed, not guarantees about every possible future.

Link the conclusion to an action. A sensitive recovery assumption may justify representative testwork; an expensive delay exposure may justify investigating construction interfaces; a scenario that defeats every alternative may require a different project scope. Set a review trigger tied to the relevant evidence, such as a changed schedule, new ore characterization or revised commercial terms. Preserve the previous model so that later revisions can be explained and the original decision assumptions checked against experience.