# Multiplicative Holt-Winters Exponential Smoothing

*Multiplicative Holt-Winters Smoothing* is the logical extension of Holt's Trend-Corrected Smoothing. It accounts for a level, a trend, and the need to adjust the demand up or down on a regular basis due to seasonal fluctuations. Note that the seasonal fluctuation needn't be every 12 months like in this example. In the case of MailChimp, we have periodic demand fluctuations every Thursday (people seem to think Thursday is a good day to send marketing e-mail). Using Holt-Winters, we could account for this 7-day cycle.

Now, in most situations you can't just add or subtract a fixed amount of seasonal demand to adjust the forecast. If your business grows from selling 200 to 2,000 swords each month, you wouldn't adjust the Christmas demand in both those contexts by adding 20 swords. No, seasonal adjustments usually need to be multipliers. Instead of adding 20 swords maybe it's *multiplying* the forecast by 120 percent....