How do you calculate exponential smoothing forecasts step-by-step?

Conditions

• The smoothing constant $\alpha$ is fixed across all periods.
• An initial forecast $F_1$ is established from $A_1$.
• Each new forecast depends only on the immediately preceding actual demand and forecast.

Reasoning, step by step

1. Initialize $F_1 = A_1$.
2. For period 2, calculate $F_2 = \alpha A_1 + (1-\alpha)F_1$.
3. For period 3, calculate $F_3 = \alpha A_2 + (1-\alpha)F_2$.
4. Continue recursively: for period $t+1$, calculate $F_{t+1} = \alpha A_t + (1-\alpha)F_t$.
5. Round the final decimal results to the nearest whole number if recording discrete demand units in a table.

Example

The presenter calculates Year 5 using $A_4=415$ and $F_4=357$: $F_5 = 0.4(415) + 0.6(357) = 380.2$, rounded to 380. Then Year 6 uses $A_5=450$ and $F_5=380$: $F_6 = 0.4(450) + 0.6(380) = 408$. Finally, Year 7 uses $A_6=465$ and $F_6=408$: $F_7 = 0.4(465) + 0.6(408) = 430.8$, rounded to 431.

Common misconceptions

• Trying to calculate all future forecasts simultaneously without using the recursive output of the previous step.
• Forgetting to update the actual demand $A_t$ to the next period's value when moving from one step to the next.

Watch the explanation

Explore next

• What does the smoothing constant alpha represent?
• Why is the Year 4 forecast written as 357 instead of 357.2?