🎓 Lesson 13
D5
Beyond Best-Case Scenarios: Modeling Uncertainty in ROI
ROI uncertainty modeling means using math and data to predict how much profit a mining or blasting investment might actually make—not just in perfect conditions, but when things like rock hardness, weather, or equipment delays change the outcome.
🎯 Learning Objectives
- ✓ Calculate probabilistic ROI ranges using Monte Carlo simulation inputs
- ✓ Analyze sensitivity of ROI to key uncertain variables (e.g., fragmentation cost, dilution rate, copper price)
- ✓ Design a risk-adjusted investment decision matrix incorporating confidence intervals and downside thresholds
- ✓ Explain how uncertainty distributions (e.g., lognormal ore grade, triangular delay time) affect NPV and payback period
📖 Why This Matters
In mining and blasting projects, over-optimistic ROI forecasts lead to undercapitalized operations, cost overruns, and stranded assets. A 2022 CIM Economic Committee report found that 68% of underperforming open-pit expansions traced back to unmodeled uncertainty in fragmentation efficiency and ore recovery—proving that 'best-case' ROI is not a plan; it’s a liability. This lesson equips you to turn ambiguity into actionable risk intelligence.
📘 Core Principles
Uncertainty in ROI arises from epistemic (knowledge gaps, e.g., unknown joint orientation) and aleatory (inherent randomness, e.g., explosive performance variation) sources. Effective modeling requires: (1) identifying high-impact, high-uncertainty inputs using tornado diagrams; (2) assigning empirically grounded probability distributions (not guesses); and (3) propagating uncertainty through integrated technical-economic models—linking blast design outputs (e.g., P80 size distribution) directly to downstream processing costs and metal recovery. Industry best practice follows the AACE International Total Cost Management Framework, which mandates probabilistic analysis for Class 3+ capital estimates.
📐 Probabilistic ROI Sensitivity Index
This index quantifies how much ROI variance is driven by a single uncertain input. It enables prioritization of data collection and mitigation efforts before final investment approval.
Sensitivity Index (SI)
SI_i = (Var(ROI) − Var(ROI|X_i = μ_i)) / Var(ROI)Measures fractional contribution of uncertain variable X_i to total ROI variance.
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| SI_i | Sensitivity Index for input i | dimensionless | Proportion of total ROI variance explained by uncertainty in input i |
| Var(ROI) | Variance of ROI across full simulation | USD² × 10⁶² | Total dispersion of ROI outcomes from Monte Carlo sampling |
| Var(ROI|X_i = μ_i) | Conditional variance of ROI with X_i fixed at mean | USD² × 10⁶² | Residual ROI variance when input i is held constant |
Typical Ranges:
Highly variable ore bodies (e.g., porphyry copper): 0.35 – 0.75
Stable hard-rock gold deposits: 0.10 – 0.30
💡 Worked Example
Problem: A proposed drill-and-blast upgrade has ROI modeled across 5,000 Monte Carlo iterations. Variance of ROI is 0.042 (USD²M²). When powder factor is fixed at its median value (0.35 kg/m³), ROI variance drops to 0.018. Calculate SI for powder factor.
1.
Step 1: Compute reduction in variance: 0.042 − 0.018 = 0.024
2.
Step 2: Divide by original variance: 0.024 / 0.042 = 0.571
3.
Step 3: Interpret: 57.1% of total ROI variance is attributable to powder factor uncertainty
Answer:
The Sensitivity Index is 0.571, indicating powder factor is the dominant uncertainty driver—warranting targeted geotechnical testing and blast trials before CAPEX approval.
🏗️ Real-World Application
At Newmont’s Boddington Mine (Western Australia), a $120M blasting optimization program initially projected 14.2% ROI under deterministic assumptions. When uncertainty in rock mass rating (RMR), explosive water resistance, and mill throughput variability were modeled using site-specific historical data (12 years of blast logs + metallurgical reports), the 90% confidence interval for ROI narrowed to 5.3%–11.8%. This revised range triggered redesign of initiation sequencing and buffer zone controls—reducing predicted dilution by 1.7% and lifting the P50 ROI to 9.1%, aligning with corporate hurdle rate requirements.