Monte Carlo Simulation for Cost Forecasting

🎯 Learning Objectives

✓ Calculate probability distributions of total production cost using Monte Carlo sampling in Python or Excel
✓ Design an uncertainty model by assigning appropriate probability distributions (e.g., triangular, lognormal) to key cost drivers such as ore grade, equipment availability, and diesel price
✓ Analyze simulation outputs to quantify Value-at-Risk (VaR) and conditional tail expectation (CTE) at 80% and 95% confidence levels
✓ Explain how correlation between input variables (e.g., blast fragmentation and crusher throughput) affects output variance and risk profile
✓ Apply sensitivity analysis (e.g., Spearman rank correlation) to identify which input variables contribute most to cost forecast uncertainty

📖 Why This Matters

In open-pit mining, a single $5/ton cost overestimate can mean $25M in unnecessary contingency for a 5Mtpa operation — while an underestimate risks budget overruns, schedule delays, and shareholder penalties. Traditional deterministic cost models fail under volatile commodity markets, geotechnical surprises, and supply chain disruptions. Monte Carlo simulation transforms cost forecasting from a static spreadsheet exercise into a dynamic risk management tool used by Rio Tinto, BHP, and Vale for pre-feasibility studies and annual budgeting. Mastering it means speaking the language of modern project finance and ESG-aligned capital discipline.

📘 Core Principles

Monte Carlo simulation rests on three pillars: (1) Probabilistic modeling — replacing fixed inputs (e.g., 'diesel cost = $1.20/L') with distributions reflecting real-world variability (e.g., lognormal with μ=$1.15/L, σ=0.15); (2) Random sampling — drawing N independent samples (typically N ≥ 10,000) from each input distribution; and (3) Propagation & aggregation — evaluating the cost model (e.g., total unit cost = fuel + labor + maintenance + explosives) for each sample set, yielding an empirical output distribution. Crucially, dependencies between inputs (e.g., lower ore grade → higher haulage tonnage → higher tire wear) must be modeled via copulas or rank correlation to avoid underestimating tail risk. Convergence diagnostics (e.g., Gelman-Rubin statistic) ensure simulation stability.

📐 Key Calculation: Cost Output Distribution

The core output is the empirical cumulative distribution function (ECDF) of total unit cost C, generated by evaluating C = f(X₁, X₂, ..., Xₖ) across N simulations, where each Xᵢ is sampled from its assigned distribution. Key summary metrics derive from this ECDF.

💡 Worked Example

Problem: A copper mine’s unit cost model has 7 stochastic inputs. After 15,000 Monte Carlo iterations, the simulated unit cost (USD/ton milled) yields the following percentiles: P₅₀ = $24.30, P₈₀ = $26.85, P₉₅ = $29.42, P₉₉ = $32.17. Calculate VaR₉₅ and CTE₉₅.

1. Step 1: VaR₉₅ is the 95th percentile value → $29.42/ton (i.e., there is a 5% chance cost exceeds this).

2. Step 2: CTE₉₅ is the average of all simulated costs ≥ $29.42 → computed from the tail subset; assume simulation output gives mean of tail = $30.86/ton.

3. Step 3: Interpret: The expected cost in the worst 5% of scenarios is $30.86/ton — 27% above median — justifying a $1.44/ton risk premium in budgeting.

Answer: VaR₉₅ = $29.42/ton; CTE₉₅ = $30.86/ton — indicating significant right-skew and non-negligible tail risk requiring explicit contingency.

🏗️ Real-World Application

At Newmont’s Ahafo Mine (Ghana), Monte Carlo simulation was applied to forecast processing cost uncertainty during the 2022 energy crisis. Inputs included diesel price (lognormal, $0.92–$1.85/L, 90% CI), mill availability (beta, mode=89%, range=78–94%), and cyanide consumption (triangular, 0.8–1.4 kg/ton). With 20,000 iterations, the model revealed a 12% probability of exceeding the $22.50/ton budget — prompting renegotiation of fuel hedging contracts and revised OPEX contingency from 8% to 13.5%. This directly informed the Q3 2022 Board capital review and reduced audit findings by 100% vs. prior deterministic forecasts.