Upstream Asset Evaluation & Stochastic Economic Modeling Framework

This post reconstructs the evaluation, financial modeling, and decision analytics framework used when leading an upstream (oil & gas) analytics team (circa 2015–2018). It blends technical reservoir & production modeling with fiscal, stochastic, real‑options, and portfolio layers plus emerging carbon governance.

1. Checklist (Top-Level Components)

• Scope definition
• Technical (subsurface & production) models
• Commercial & fiscal models
• Market & price modeling
• Cost & economic models
• Real options layer
• Stochastic engine & correlations
• Portfolio aggregation
• Risk & sensitivity
• Carbon / ESG integration
• Data architecture & governance
• Validation & model risk management
• Implementation blueprint

2. Scope & Objectives

Asset lifecycle: exploration → appraisal → development planning → execution → ramp-up → plateau → decline → abandonment.
Decisions supported: license bidding, sanction (FID), phasing, drilling sequence, facility sizing, hedging, M&A, divestment, suspension, expansion, abandonment timing.
Outputs: NPV (pre/post tax), IRR, payback, PI, EMV / ENPV, free cash flow profiles, value at risk (P10/P50/P90), option-adjusted value, carbon-adjusted value, capital efficiency, portfolio efficient frontier.

3. Technical Modeling

• Reservoir / production: decline curves (Arps variants), reservoir simulation inputs (pressure, permeability, recovery factor).
• Development schedule: drilling calendar, well productivity distributions (IP, decline b-factors), workover frequency, facility uptime (availability model).
• Flow assurance: constraints (pipeline diameter, backpressure, compression limits).
• Operational losses: planned maintenance, stochastic unplanned downtime.
• Recovery uplift levers: EOR phases as contingent scenarios with trigger thresholds.

4. Commercial & Fiscal Regimes

• Fiscal: royalties (sliding), cost recovery ceiling, profit oil splits, ring-fencing, depreciation schedules, uplift factors, corporate & withholding tax.
• PSC: cost pool accumulation, unrecovered cost carry-forward, R-factor / IRR tranche splits.
• Transfer pricing & tariffs: pipeline tariffs, processing fees, storage.
• Local content / penalties: schedule or cost multipliers.
• Cash waterfall: gross revenue → royalties → allowable recoverable costs → profit split → tax → net cash flow.

5. Price & Market Modeling

• Commodity prices: mean-reverting jump diffusion (oil), two-factor (spot + long) gas model.
• Basis differentials: correlated regional spreads (Brent–WTI, Henry Hub vs contract).
• Inflation: multi-bucket cost indices (labor, steel, services) with correlation matrix.
• FX: GBM or mean-reverting with correlations to commodity indices.

6. Cost & Economic Drivers

• CAPEX: drilling (probabilistic cost per well), facilities (modular capacity curve), subsea/pipeline, expansion options.
• OPEX: fixed + variable (per BOE), escalation indices, logistics, emissions compliance.
• Decommissioning: probabilistic tail cost; timing via economic limit test (ELT).
• Working capital: receivables/payables days; inventory (condensate/LNG).
• Insurance & carbon: scenario carbon price paths (base / accelerated transition).

7. Real Options Layer

• Option types: timing (delay sanction), staging (phase 2 expansion), switching (reinjection vs export), abandonment (economic cutoff), suspension (shut-in), expansion (debottleneck), contraction (mothball partial train).
• Triggers: price thresholds, volatility bands, utilization %, reserve reclassification (2C→2P), regulatory approvals.
• Methods: binomial lattice, Monte Carlo decision rules, Longstaff–Schwartz regression, compound option valuation.
• Integration: option-adjusted NPV = static ENPV + incremental option premium.

8. Stochastic Engine & Correlations

• Simulation: Monte Carlo (10k–50k) + variance reduction (Sobol, antithetic samples).
• Random variables: prices, well IP, decline params, drill duration, CAPEX overrun %, OPEX variance, uptime, FX, inflation, carbon price.
• Correlation: Cholesky of calibrated matrix (price–FX–inflation, intra-pad well correlation, CAPEX component linkages).
• Scenario overlay: structural transition / geopolitical shocks layered deterministically.

9. Production & Revenue Mechanics

• Well-level: IP distribution (lognormal), decline (hyperbolic → exponential tail), downtime factor, choke constraints.
• Facility capacity: throughput constraint; shadow pricing for debottleneck identification.
• Stream splits: GOR evolution; per-stream pricing.
• Regulatory gas handling: flaring limits trigger gas infrastructure CAPEX or curtailment.

10. Portfolio Aggregation

• Consolidate asset cash flows; allocate overhead; tax consolidation & loss utilization.
• Capital rationing: optimize ENPV under CAPEX budget, emissions ceiling, CVaR risk constraint.
• Techniques: mixed-integer stochastic programming or heuristic efficient frontier sampling.

11. Risk & Sensitivity

• Deterministic sensitivities: tornado (±10–30%) for oil price, CAPEX unit cost, IP, decline b, uptime.
• Probabilistic outputs: P10/P50/P90 NPV & IRR distribution, downside risk (5th percentile), breakeven distributions.
• Attribution: Shapley / variance decomposition on value drivers.
• Stress: low price crash, inflation spike, regulatory delay, carbon acceleration.

12. Carbon / ESG Integration

• Emissions: scope 1 (fuel gas, flaring, venting, methane leaks) with intensity trajectory & abatement project levers (electrification, CCS).
• Carbon pricing: stochastic or scenario-based (IEA SDS, APS).
• Carbon-adjusted NPV: subtract discounted compliance & internal shadow pricing.
• ESG risk: probability-modified schedules for permitting / social license.

13. Data Architecture & Governance

• Layers: Raw → Curated → Analytical Mart → Results Store.
• Parameter governance: metadata (source, calibration date, owner, confidence level).
• Reproducibility: hash of input bundle + model version persisted with outputs.
• Audit lineage: reservoir model revision → production curve change traceability.

14. Calibration & Validation

• Historical fitting: decline (RMSE, MAPE), price process (MLE), cost curve regression.
• Backtesting: forecast distribution vs realized production / costs / prices.
• Benchmarking: analog fields for IP & decline.
• Statistical tests: Kolmogorov–Smirnov (distribution fit), Ljung–Box (autocorrelation).

15. Model Risk Management

• Documentation: scope, assumptions, limitations, parameter ranges, validation pack.
• Change control: semantic versioning; material change triggers re-validation.
• Independent review: separation developer vs validator; periodic cycle.
• Limitations ledger: tail risks outside modeled domain.

16. Implementation Blueprint

• Stack: Python (NumPy, Pandas, SciPy), probabilistic (PyMC/custom), optimization (Pyomo / OR-Tools), parallel (Dask / Spark large scenario sets), dashboard (Streamlit / Power BI).
• Modular services: pricing, production, fiscal, real options, portfolio optimizer.
• API contract: input bundle (JSON/YAML) → simulation config → results object (summary + distributions).
• Performance: vectorized decline simulation, cache deterministic transforms, scenario slicing.
• Quality gates: unit tests (cash waterfall, tax), property tests (option monotonicity), regression (NPV tolerance stability).

17. Key Formulas

NPV = Σ_t ( CF_t / (1 + r)^t )
Economic limit: stop when (revenue_t - variable_cost_t - incremental fixed share) < threshold
Hyperbolic decline: q_t = q_i / (1 + b * D_i * t)^(1/b)
Mean-reverting price: dP = k(μ - P) dt + σ dW + J dN
Real option (binomial): Option = discount( p * V_up + (1-p) * V_down )
ENPV = (1/N) * Σ_i NPV_i
CVaR_α = E[ NPV | NPV ≤ VaR_α ]

18. Prioritization Roadmap

19. Common Pitfalls

• Double counting optionality (in production profiles and option layer)
• Ignoring correlation (tail risk underestimation)
• Using single P50 case for sanction without downside assessment
• Overfitting early decline data; missing exponential tail transition
• Static cost escalators during volatile inflation regime

20. Success Metrics

• Cycle time: new parameter bundle → refreshed valuation < 30 min
• Backtest error vs realized 1-year production < ±8%
• Audit reproducibility: 100% outputs traceable to parameter hash
• Simulation convergence: NPV (P90–P10) band variation <2% with +20% runs
• Option value attribution: explicit delta vs static ENPV

Historical reconstruction for knowledge retention & reference; not investment advice.