Decision Support, Analytics, and Quantitative Methods
Decision support and analytics test quantitative reasoning for office and operations decisions. Expect NPV, expected value, decision trees, forecasting, linear programming, queuing, sensitivity analysis, and the descriptive-predictive-prescriptive spectrum.
Core rules
- Discount cash flows to a common date. NPV = present value of inflows minus outflows; accept independent projects with positive NPV at the required rate.
- Expected monetary value weights outcomes by probability. Decision trees sequence decisions, chance events, and payoffs.
- Forecasting: moving averages smooth noise; exponential smoothing weights recent data; measure error with MAD or MAPE. Use holdout periods.
- Optimization requires an objective, decision variables, and constraints. Linear programming assumes linearity and divisibility; sensitivity analysis shows how results change with inputs.
- Queuing and Little’s Law link arrivals, waiting, and service. Variability, not just average utilization, drives queues.
Worked example
A project costs 100,000 now and returns 40,000 per year for four years. At a 10 percent discount rate, the annuity factor is (1 − 1.10^−4) / 0.10 = 3.1699. PV of inflows = 40,000 × 3.1699 = 126,796. NPV = 126,796 − 100,000 = 26,796, so accept if the risk-adjusted rate is truly 10 percent. The reasoning uses the annuity factor for a level stream and compares PV with the initial outlay; adding undiscounted cash flows would overstate value.
Traps
- Discounting year 1 at time zero or mixing nominal cash flows with a real discount rate.
- Using expected value when the decision maker is risk-averse or outcomes are catastrophic.
- Overfitting a forecast to noise and reporting a single point estimate without error bounds.
- Confusing correlation with causation in analytics.