What is double exponential smoothing?
Double exponential smoothing, known as Holt’s method, is a forecasting technique that tracks two things over time, the level of demand and its trend, and updates each with its own smoothing weight as new data arrives.
Exponential smoothing forecasting gives recent observations more weight than old ones; the double version adds a trend, so the forecast can rise or fall instead of staying flat.
Double exponential smoothing formula
With weights α and β between 0 and 1, each new observation y updates the level ℓ and the trend b:
- Level: ℓ(t) = α × y(t) + (1 − α) × (ℓ(t−1) + b(t−1))
- Trend: b(t) = β × (ℓ(t) − ℓ(t−1)) + (1 − β) × b(t−1)
- Forecast h periods ahead: ℓ(t) + h × b(t)
A higher α reacts faster to changes in level; a higher β reacts faster to changes in trend.
A Holt method example
Monthly demand runs 100, 108, 115 and 124. Starting from a level of 100 and a trend of 8, with α = 0.5 and β = 0.3, the level ends near 124 and the trend near 8 a month, so next month’s forecast is about 132 and three months ahead about 147. A four-month moving average would say about 112 and keep lagging behind the growth.
Holt-Winters vs Holt
Holt’s method handles level and trend but not seasonality. Holt-Winters, or triple exponential smoothing, adds a third component for repeating seasonal patterns, such as a December peak. Use Holt for series that drift steadily and Holt-Winters for anything with a yearly or weekly rhythm.
Where it is used
Smart Inventory Forecasting fits Holt’s method with seasonality per item, which is where a spreadsheet usually goes wrong by treating last month as the forecast, and it checks each forecast against held-back recent data before trusting it.
Last reviewed 2026-11-01.