forecasting-ch04-smoothinglisted

# Chapter 4: The Smoother ## Scope and intake Use when regular demand, workload or revenue series whose recent level, trend and repeating seasonality may forecast the next periods. Do not use for a new launch with no comparable history, intermittent demand dominated by zeros, or an intervention-effect question. Ask only for unresolved material inputs: What is the observation frequency and decision horizon? Is demand censored by stockouts? Which seasonal period is plausible? Is the business changing enough that old cycles are misleading? ## Input contract and additional evidence CSV: timestamp (unique ISO period), target (finite number in consistent units). Config: frequency, horizon, season, as_of, units and source. At least two complete seasonal cycles are needed even to attempt seasonal estimation; reserve further data for evaluation. The rolling comparison needs 2 seasons + 4 horizons of history (72 monthly points for a 12-month horizon, 48 for six months, 36 for three); with less, the tool returns a provisional persistence baseline and says how many points are missing. Zero sales are observations; absent months are missing. Minimal **format illustration**, not sufficient training data: ```csv timestamp,target 2024-01-01,100 2024-02-01,110 2024-03-01,105 ``` ## Executable interface Exact CLI columns: `timestamp,target`. Supported method controls: `as_of, conformal, country, criterion, frequency, future_regressors, horizon, origins, per_horizon_buckets, periods, po