Category 27
Quantitative Business: Evidence, Experiments, and Optimization
Proving what works — then optimizing it.
- 7 modules
- 42 lessons
- ~9h total
What this category covers
Every other category teaches you what to measure. This one teaches you whether the measurement means anything. It runs in sequence: read evidence without being fooled by it, design an experiment that could actually detect the effect you are looking for, recover a causal answer when you cannot run one, decide under uncertainty, optimize the thing, and only then let a model decide on its own. This is the first statistics anywhere in the library, and it is written for someone with a profit-and-loss statement rather than a degree — every method arrives attached to a decision it changes. Fair warning: about half of it exists to tell you that your last clear win was noise.
Reading Evidence
- Sampling Error and How Wrong a Small Sample Can Be
- Statistical Significance and What a P-value Does not Mean
- Confidence Intervals and the Range You Should Have Quoted
- Base Rates and the Prosecutor's Fallacy in Business Data
- Regression to the Mean and the Illusion of a Turnaround
- Correlation, Confounding, and the Causal Question Underneath
Running Experiments
- Designing an A/B Test That Can Actually Detect the Effect
- Statistical Power and the Minimum Detectable Effect
- The Peeking Problem and Why You Cannot Watch a Test Run
- Multiple Comparisons and the False-discovery Rate
- Holdouts and Geo Experiments When You Cannot Randomize Users
- Switchback Tests and Interference Between Treated Units
Causal Inference Without an Experiment
- Difference-in-differences and the Parallel-trends Assumption
- Instrumental Variables and the Exclusion Restriction
- Regression Discontinuity and the Local Average Treatment Effect
- Synthetic Control and the Constructed Counterfactual
- Propensity-score Matching and the Selection-on-observables Bet
- Uplift Modeling and Heterogeneous Treatment Effects
Deciding Under Uncertainty
- Bayesian Updating and the Posterior You Should Have Carried
- Expected Utility and Why a Risk-neutral Firm Is a Fiction
- The Value of Information and What a Test Is Worth Before You Run It
- Multi-criteria Decision Analysis and Weighting Things That Do not Compare
- Robust Optimization and the Price of the Worst Case
- Stochastic Programming and Decisions You Make in Two Stages
Optimization and Operations Research
- Linear Programming and Reading the Shadow Price
- Integer Programming and Why Scheduling Is Hard
- Network Flow and the Vehicle Routing Problem
- The Newsvendor Problem and the Cost of Ordering Wrong
- Queueing Theory and Why Utilization Above 85 Percent Breaks
- The Theory of Constraints and Managing the Bottleneck
Models That Predict
- Supervised Learning and the Bias-variance Tradeoff
- Overfitting, Cross-validation, and the Honest Holdout
- Forecasting Demand and the Limits of a Time Series
- Multi-armed Bandits and the Explore-exploit Tradeoff
- Reinforcement Learning for Sequential Business Decisions
- Model Risk and When a Model Should not Be Allowed to Decide
Pricing and Revenue Management
- Revenue Management and Where Dynamic Pricing Came From
- Estimating a Demand Curve From Your Own Price History
- Protection Levels and Littlewood's Rule for Perishable Capacity
- Surge and Peak Pricing and the Fairness Constraint
- Bandit-based Price Testing Without Burning the Market
- Personalized Pricing and the Legal and Reputational Limits