013 interactive stages
Commerce & Marketplaces
Assortments, substitution, learning from customer choices, and decisions under preference shift.
View the live caseMenu
Examples
M2W Decision Labs
From a model to a decision—and then toward the world. Each example starts with a concrete decision, exposes the mechanics, compares policies on common inputs, and names the evidence still required for real-world use.
Applications
Explore by domain.
Applications give each decision its context. Methods show how M2W moves it from a modeled world toward reliable use.
02Prototype-demo coming soon
Finance & Risk
Portfolios, valuation, insurance, rare events, and performance under changing markets.
03Prototype-demo coming soon
Operations & Health
Capacity, scheduling, patient flow, and service-system decisions under uncertainty.
04Prototype-demo coming soon
Autonomy & Engineering
Control, routing, reliability, and fast decisions under operating-regime shift.
Methods across M2W
Different lenses on the same transfer problem.
- SimulationExplore plausible, rare, and consequential worlds.
- LearningEstimate mechanisms and decisions from partial observations.
- UQ & Model RiskMeasure uncertainty and model-to-world discrepancy.
- RobustnessProtect performance when the modeled world changes.
- Causal & CounterfactualReason about outcomes that cannot be jointly observed.
- OptimizationSearch for strong actions under constraints.
Live case study · Commerce & Marketplaces
Assortment decisions
One retail decision, followed through three increasingly realistic worlds: start with a known choice model, learn from logged behavior, then protect the policy against preference shift.
3 interactive stagesSynthetic evidenceReproducible
One decision, three stages.
- Decision Lab 01Known model
Choose the shelf
When less shelf space makes more revenue.
Compare every feasible assortment when customer substitution is known.
- Choice modeling
- Exact evaluation
- Optimization
- Decision Lab 02Logged data
Learn the shelf
More data cannot reveal what you never offered.
Compare incumbent logs with designed exploration, then test the learned decision on a fresh holdout.
- Offline learning
- Pessimism
- Validation
- Decision Lab 03Preference shift
Protect the shelf
Pay a little now. Protect the decision later.
Trade a small amount of nominal value for a stronger protected floor under coherent preference shift.
- Robustness
- Model risk
- Stress testing
Prototype-demos coming soon
Where M2W goes next.
Each prototype-demo will use synthetic or approved materials and state the evidence required for transfer.
Operations & HealthPrototype-demo coming soon
Hospital Operations
How should capacity and patient-flow decisions respond to uncertain demand and operational constraints?
Commerce & MarketplacesPrototype-demo coming soon
Marketplace Dynamics
Which policies improve service and participation without destabilizing a two-sided system?
Reading the examples
Build maturity and evidence answer different questions.
- Build maturity
- Concept → Demonstration build → Interactive prototype → Evaluated demonstration → Transferred
- Evidence basis
- Synthetic → Public data → Partner or shadow evidence → Real-world evidence