Back to all methods
DATUM ALGORITHMIC FAMILY

Method: Statistical Time-Series & Demand Sensing

Autoregressive baselines, CUSUM drift detection, Fourier multi-seasonality decomposition, and optimal hierarchical reconciliation (MinT).

#002
Forecast Diagnostics

Forecasts are consistently too high or too low

Are we repeatedly leaning in one direction?

#002 CUSUM Bias Detection+23 Over-forecast
Actual Demand
Forecast Baseline
CUSUM OFFSET:-0 u
SYSTEM BIAS:+23 Shift
💡Small errors in one direction create compounding operational waste.
Blueprint→