STOP 15 / 31Remember the rhythm, too.
Layout
STOP 15 · Models

Remember the rhythm, too.

EXPLORE~6 min
Back to route map
  • Holt–Winters extends smoothing to level, trend, and a seasonal pattern.
  • Choose additive or multiplicative seasonality depending on whether seasonal changes are fixed amounts or proportional to the level.
Holt–Winters: seasonality: concepts and Northstar examples. Illustrative Northstar examples, not measured business results.
ConceptWhat it meansNorthstar example
Each seasonal position gets a state.

For weekly daily data, the method maintains weekday effects and updates them as new data arrive.

A weekly Holt–Winters model maintains separate weekday effects for Northstar's Monday-through-Sunday shopping rhythm.

A single period has limits.
  • Classic Holt–Winters handles one seasonal period.
  • Northstar has both weekly and annual structure, plus a moving sales event; that requires a richer strategy.

A seven-day seasonal model cannot also be assumed to know Northstar's annual holiday ramp and moving Black Friday event.

Scale is a modeling choice.
  • Multiplicative methods assume positive values and proportional effects.
  • An additive model on a transformed scale is another possibility.
  • If weekends add a similar dollar amount at every sales level, additive effects may fit.
  • If the uplift scales with Northstar's size, proportional effects deserve investigation.
FIG 15.1

Add level, trend, and weekly states

WHAT THIS SHOWS

The method updates a level and trend, then learns one seasonal state for each weekday. Add stages one at a time and compare the holdout forecast. Additive seasonality adds a fixed dollar pattern; multiplicative seasonality adds a percentage that scales with the level.

The question it answers: What does each Holt–Winters stage add to Northstar's forecast: a level, a trend, or a weekly rhythm?

How the weekly pattern scales
States in the model
Northstar daily sales with staged Holt–Winters forecasts$8k$13k$18k$23kForecast originWeek 1Week 4Week 7Today+14 days
Observed salesOne-step fitted levelHoldout forecast · 3 · Add weekly season

Learned weekday effects · additive

$-1,063Mon$-670Tue$-218Wed$56Thu$391Fri$1,123Sat$406SunDollar effect vs an ordinary weekday

3 · Add weekly season. The forecast adds the learned weekday state to the level and trend. The weekly pattern is learned only from past observations.

Weekly states active. The supplied recursions use β = .08 and γ = .25 to make each stage distinct.

Holdout MAE · last 14 days$111
Estimated trend / day$117.9
Seasonal swing$2,186
Forecast stateWeekly states active
Fine print & method

Constructed daily sales with a known upward level and a repeating weekly pattern; the last 14 days are a holdout. The recursions update level, trend, and seven seasonal states using fixed α, β, and γ by stage. No parameters are optimized; this is a teaching implementation, not a fitted ETS model. Additive and multiplicative data are generated separately, so the model form matches the data. A fixed weekly period does not automatically model annual events such as Black Friday; those need explicit event terms.

CHECK YOURSELF · CHOOSE AN ANSWER

Does a 7-day Holt–Winters model automatically capture Black Friday each year?

Your learning note

Private · just for you