Make Better Capital Decisions Under Deep Uncertainty
Challenge:
"Should we delay, upgrade, or replace? We don't know."
Solution:
We model multi-stage decisions over decades, accounting for evolving risks and learning.
Case Study: Chemicals Manufacturer – Process Unit Revamp
Built a dynamic decision model to evaluate phased investment options under uncertainty in feedstock prices, regulatory changes, and technology obsolescence. The model prescribed adaptive policy paths rather than static recommendations, allowing the client to respond intelligently as conditions evolved.
Behind the Scenes: Technical Approach
Method
Stochastic Dynamic Programming (DP)
Tools
Wolfram Mathematica for symbolic computation, custom Markov models
Horizon
5+ year forecast with probabilistic outcomes
Outcome
Prescribed adaptive policy paths, not static recommendations
Why This Approach Works
Multi-Stage Decision Trees
We model not a single "invest or don't" decision, but sequences of decisions over time — where each choice opens or closes future options under evolving uncertainty.
Adaptive Policy Design
Instead of prescribing "build this now," we deliver decision rules: "if feedstock prices exceed X and regulatory clarity improves, then upgrade; otherwise delay and monitor."
Learning Over Time
Using Bayesian updating and Markov models, we account for the value of waiting to learn — sometimes the best decision today is to preserve future flexibility.