Simple Explanation of Exponential Smoothing Forecast Method

1. Formula
   Observation

   The worksheet shows the general formula $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Audio
   Observation

   The speaker explains $F(t)$ as the forecast for the time you are looking for and $F(t-1)$ as the forecast from the previous period.

$F(t)$

Forecast value at time t, here the forecast for a given week.

Discrete time index t corresponding to weeks in the example; numeric forecast values.

1. Formula
   Observation

   The displayed formula contains $A(t-1)$.

2. Audio
   Observation

   The speaker refers to the actual from the previous week when explaining the term multiplied by alpha.

$A(t-1)$

Actual observed value at the previous time point, here last week's unit sales.

Discrete time index t-1; numeric actual sales values.

1. Formula
   Observation

   The displayed formula uses alpha as the multiplier.

2. Diagram
   Observation

   Cell B2 is labeled Smoothing Constant and contains 0.2.

3. Audio
   Observation

   The speaker says alpha is the exponential smoothing constant and that in this case it will be 0.2.

alpha

Exponential smoothing constant controlling how much the forecast adjusts toward the previous forecast error.

In this example alpha = 0.2, stored in cell B2.

1. Formula
   Observation

   The formula uses t and t-1 in $F(t)$, $F(t-1)$, and $A(t-1)$.

2. Audio
   Observation

   The speaker maps the time periods to weeks while walking through the example.

t

Current time index; in the example it denotes the week being forecast.

Integer week index, with the worked example covering Week 1 through Week 11.

1. Diagram
   Observation

   The cursor selects cell C4 in the Forecast column for Week 1.

2. Audio
   Observation

   The speaker says that if there is no previous week, we simply use the actual from that week as our forecast.

C4

Excel cell containing the Week 1 forecast entry.

1. Diagram
   Observation

   The user types a formula into cell C5 for Week 2.

2. Audio
   Observation

   The speaker says Week 2 is when we start using the formula.

C5

Excel cell where the recursive exponential smoothing formula is first entered for Week 2.

1. Diagram
   Observation

   The typed Excel formula references B4 as the actual sales value for Week 1.

B4

Excel cell containing the Week 1 actual unit sales value.

1. Diagram
   Observation

   The typed Excel formula references C4 as the previous forecast value.

C4

Excel cell containing the Week 1 forecast value used as the previous-period forecast in the Week 2 formula.

1. Diagram
   Observation

   The formula bar shows $B$2 inside the Week 2 formula.

2. Audio
   Observation

   The speaker says alpha is a constant and they lock it in place so it stays consistent through the drag of the formula.

$B$2

Absolute reference to the smoothing constant cell B2 in Excel.

1. Diagram
   Observation

   The table shows Week 1 with Unit Sales 1103 and Forecast 1103.

2. Audio
   Observation

   The speaker says 1103 becomes our forecast for time one.

Week 1: $A(1)=1103$, $F(1)=1103$

Initial condition for the worked example: the first forecast equals the first actual value.

1. Formula
   Observation

   The formula in cell C5 uses the absolute reference $B$4, which corresponds to the 'Smoothing Constant'.

2. Diagram
   Observation

   Cell B4 contains the value 0.2 and is labeled 'Smoothing Constant'.

$\alpha$

Smoothing constant, a parameter between 0 and 1 that determines the weight given to the most recent observation.

(0, 1)

1. Formula
   Observation

   The text overlay shows $F(t)$ as the left-hand side of the equation.

2. Formula
   Observation

   The formula in cell C5 calculates the forecast for the current row.

$F(t)$

Forecast for period t.

Real numbers

1. Formula
   Observation

   The worksheet displays $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Audio
   Observation

   The speaker verbally defines each part of the formula as previous forecast plus alpha times previous actual minus previous forecast.

The video presents the single-parameter exponential smoothing update rule. The new forecast equals the prior forecast plus a fraction alpha of the prior forecast error, where the error is measured as actual minus forecast from the previous period.

1. Used recursively over discrete time periods.

2. Requires a starting forecast value.

3. In the example, alpha is supplied as 0.2.

1. Audio
   Observation

   The speaker says that if we do not have a previous week to come up with a forecast, we just simply use the actual from that week as our forecast.

2. Diagram
   Observation

   Cell C4 is filled with 1103, matching the Week 1 actual sales value in B4.

Because the recursive formula needs a previous forecast, the video demonstrates a simple initialization method: set the first forecast equal to the first actual observation. In the example this gives $F(1)=A(1)=1103$.

1. Applies when no earlier forecast is available.

2. Shown specifically for Week 1 in the example.

1. Exponential smoothing forecast formula

1. Audio
   Observation

   The speaker identifies alpha as the exponential smoothing constant and says it is typically given in the problem.

2. Diagram
   Observation

   Cell B2 is labeled Smoothing Constant and contains 0.2.

Alpha is defined in the video as the exponential smoothing constant. It multiplies the previous forecast error and determines the size of the adjustment added to the previous forecast. In this worked example alpha is fixed at 0.2.

1. Treated as a given constant in the problem.

2. Held fixed across all periods in the example.

1. Diagram
   Observation

   The user enters =C4+$B$$2*(B4-C4)$ into cell C5.

2. Audio
   Observation

   The speaker walks through the formula as previous forecast plus alpha times previous actual minus previous forecast and mentions locking alpha for dragging.

The video translates the mathematical recurrence into an Excel formula. For Week 2, the current forecast is computed from the Week 1 forecast in C4, the Week 1 actual in B4, and the smoothing constant in B2. The constant is made absolute with $B$2 so the same alpha is reused when the formula is copied downward.

1. Requires a table with actuals in one column and forecasts in another.

2. Uses relative references for prior row values and an absolute reference for alpha.

1. Exponential smoothing forecast formula
2. Smoothing constant alpha

1. Formula
   Observation

   Text overlay displays the equation $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Formula
   Observation

   Excel formula bar shows =C4+$B$$4*(B4-C4)$.

The exponential smoothing forecast for the next period is calculated by taking the previous period's forecast and adding a fraction (determined by the smoothing constant alpha) of the error from the previous period (the difference between the actual value and the forecast).

1. Requires an initial forecast value.

2. Alpha must be between 0 and 1.

1. Formula
   Observation

   The formula in cell C5 is =C4+$B$$4*(B4-C4)$.

2. Animation
   Observation

   The user double-clicks the fill handle of cell C5, auto-filling the formula down column C.

In Excel, the formula is implemented using relative references for the previous forecast (C4) and actual sales (B4), and an absolute reference ($B$4) for the smoothing constant. The formula can then be copied down the column to calculate forecasts for subsequent periods.

1. The smoothing constant must be stored in a fixed cell and referenced absolutely.

1. Exponential Smoothing Formula

1. Formula
   Observation

   The displayed equation is $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Audio
   Observation

   The speaker states that the forecast for the time you are looking for equals the forecast from the previous period plus alpha times the actual from the previous week minus the forecast from the previous week.

For each period t, the forecast is updated by adding alpha times the previous period's forecast error to the previous forecast.

1. A previous forecast $F(t-1)$ exists.

2. A previous actual $A(t-1)$ exists.

3. A smoothing constant alpha is specified.

For the discrete time index t in the forecasting sequence.

1. Audio
   Observation

   The speaker says that without a previous week, the actual from that week is used as the forecast.

2. Diagram
   Observation

   The Week 1 forecast cell is filled with 1103, identical to the Week 1 actual.

When no earlier forecast is available, the first forecast can be initialized as the first actual value, so in this example $F(1)=1103$.

1. This is the first period in the data series.

2. No prior forecast is available.

Applied to Week 1 in the demonstrated dataset.

1. Formula
   Observation

   The general formula on screen is $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Diagram
   Observation

   The Excel formula entered in C5 is =C4+$B$$2*(B4-C4)$.

3. Audio
   Observation

   The speaker explains that for Week 2 we use the forecast from Week 1 plus alpha times the actual from Week 1 minus the forecast from Week 1.

1. Expression

   $$F(t) = F(t-1) + \alpha \cdot (A(t-1) - F(t-1))$$
   Explanation

   Start from the general exponential smoothing recurrence shown on the worksheet.

   Justification

   Directly observed formula in the video.

   Shown in the video
2. Expression

   $$F(2) = F(1) + \alpha \cdot (A(1) - F(1))$$
   Explanation

   Specialize the recurrence to the second week, so the current forecast is Week 2 and the previous values are from Week 1.

   Justification

   Substitution $t=2$ into the displayed recurrence.

   Derived from the video
3. Expression

   $$\alpha = 0.2$$
   Explanation

   Use the smoothing constant provided in the example.

   Justification

   Observed in cell B2 and stated aloud by the speaker.

   Shown in the video
4. Expression=C4+$B$$2*(B4-C4)$
   Explanation

   Translate the symbolic expression into Excel references: C4 is $F(1)$, B4 is $A(1)$, and $B$2 is alpha.

   Justification

   Observed formula entry in cell C5, with audio explanation of each referenced quantity.

   Shown in the video

The Week 2 forecast formula in Excel is the direct cell-reference implementation of the general exponential smoothing recurrence with alpha fixed by an absolute reference.

1. Formula
   Observation

   Text overlay: $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

2. Formula
   Observation

   Excel formula: =C4+$B$$4*(B4-C4)$.

1. Expression

   $$F(t) \rightarrow \text{Cell C5}$$
   Explanation

   The current forecast being calculated corresponds to the active cell where the formula is entered.

   Justification

   Standard Excel formula entry.

   Derived from the video
2. Expression

   $$F(t-1) \rightarrow \text{Cell C4}$$
   Explanation

   The previous forecast corresponds to the cell directly above the current one in the Forecast column.

   Justification

   Relative referencing in Excel.

   Derived from the video
3. Explanation

   The smoothing constant is stored in cell B4. The dollar signs make it an absolute reference so it doesn't change when the formula is copied down.

   Justification

   Absolute referencing in Excel.

   Derived from the video
4. Expression

   $$A(t-1) \rightarrow \text{Cell B4}$$
   Explanation

   The previous actual sales correspond to the cell in the Unit Sales column on the same row as the previous forecast.

   Justification

   Relative referencing in Excel.

   Derived from the video

The Excel formula =C4+$B$$4*(B4-C4)$ is a direct translation of the mathematical formula $F(t) = F(t-1) + \alpha \cdot (A(t-1) - F(t-1))$.

1. Diagram
   Observation

   The table lists Week 1 through Week 11 with Unit Sales and Forecast columns; Week 1 actual is 1103 and Week 2 actual is 1057.

2. Audio
   Observation

   The speaker says the example uses actual sales from week 1 to week 10 and asks for the week 11 forecast using exponential smoothing.

3. Diagram
   Observation

   The user enters 1103 in the Week 1 forecast cell and then enters =C4+$B$$2*(B4-C4)$ in the Week 2 forecast cell.

1. The final Week 11 forecast value is not reached within this clip.

2. Some later weekly actual values are visible but not all are discussed in detail before the clip ends.

Given weekly unit sales for a mock company and a smoothing constant of 0.2, compute exponential smoothing forecasts recursively and ultimately determine the Week 11 forecast.

1. Weekly actual sales are listed for Week 1 through Week 10.

2. Smoothing Constant = 0.2 in cell B2.

3. Week 1 actual sales = 1103.

4. Week 2 actual sales = 1057.

5. The general formula shown is $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

Set up the forecast sequence using exponential smoothing and prepare to extend it through Week 11.

1. Expression

   $$F(1) = A(1) = 1103$$
   Explanation

   Initialize the first forecast by copying the first actual because there is no previous week.

   Justification

   Stated verbally and shown by entering 1103 into the Week 1 forecast cell.

   Shown in the video
2. Expression

   $$F(2) = F(1) + \alpha \cdot (A(1) - F(1))$$
   Explanation

   Apply the recurrence for Week 2 using Week 1 values.

   Justification

   Direct specialization of the displayed formula.

   Derived from the video
3. Expression=C4+$B$$2*(B4-C4)$
   Explanation

   Enter the Week 2 forecast as an Excel formula referencing the Week 1 forecast, Week 1 actual, and the absolute alpha cell.

   Justification

   Observed formula entry in cell C5.

   Shown in the video
4. Expression

   $$F(3) = F(2) + \alpha \cdot (A(2) - F(2))$$
   Explanation

   The speaker explains that dragging the formula down makes Week 3 use the Week 2 forecast and Week 2 actual.

   Justification

   Audio explanation near the end of the clip.

   Shown in the video

Within this clip, the setup is completed through the Week 2 formula and the recursion pattern for later weeks is explained; the numerical Week 11 forecast is not yet computed on screen.

The example can be checked by confirming that each new forecast uses the immediately preceding forecast and actual, with alpha held constant at 0.2 through the absolute reference $B$2.

1. Diagram
   Observation

   The spreadsheet contains columns for Week, Unit Sales, and Forecast with numerical data.

2. Animation
   Observation

   The chart updates to show the forecast line after the formula is filled down.

Given historical weekly unit sales and a smoothing constant of 0.2, calculate the forecast for each subsequent week.

1. Initial forecast for Week 1 is 1103.

2. Smoothing constant $\alpha = 0.2$.

3. Historical unit sales data for Weeks 1 through 10.

Compute the exponential smoothing forecast for Weeks 2 through 11.

1. Expression

   $$\text{Week 2 Forecast} = 1103 + 0.2 \times (1103 - 1103) = 1103$$
   Explanation

   Using the formula with Week 1 actuals and forecast.

   Justification

   Exponential smoothing formula.

   Derived from the video
2. Expression

   $$\text{Week 3 Forecast} = 1103 + 0.2 \times (1057 - 1103) = 1094$$
   Explanation

   Using the formula with Week 2 actuals and the newly calculated Week 2 forecast.

   Justification

   Exponential smoothing formula.

   Derived from the video
3. Expression

   $$\text{Auto-fill remaining weeks}$$
   Explanation

   The user double-clicks the fill handle in Excel to apply the formula to all remaining rows automatically.

   Justification

   Excel functionality.

   Shown in the video

The forecast values for Weeks 2 through 11 are calculated and displayed in column C, and plotted as a red line on the chart.

The resulting forecast line on the chart visually smooths out the erratic fluctuations of the actual sales data (blue line).

1. Diagram
   Observation

   An Excel worksheet titled around exponential smoothing is visible with a data table, a formula line, and a chart.

2. Audio
   Observation

   The speaker introduces the exponential smoothing forecast method using a mock company example.

1. Data table with Week, Unit Sales, and Forecast columns.

2. Formula text $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$).

3. Line chart titled Exponential Smoothing Forecast Method (example).

4. Cursor highlight circle.

1. The cursor moves among the title area, table, and formula region.

2. The speaker frames the problem as forecasting Week 11 from Weeks 1 through 10.

1. The worksheet remains on the same example throughout the clip.

2. The smoothing constant shown is 0.2.

The visual layout establishes the three components needed for the lesson: historical data, the recurrence formula, and a chart for comparing actuals with forecasts.

1. Diagram
   Observation

   The chart legend labels Unit Sales in blue and Forecast in red.

2. Audio
   Observation

   The speaker says that as forecasts are added, they will populate in red on the graph.

1. Blue line for Unit Sales.

2. Red line for Forecast.

3. Chart legend at the bottom.

1. The explanation links future formula entries in the Forecast column to the red chart series.

1. Unit Sales remain the blue series.

2. Forecast remains the red series.

The color coding visually separates observed sales from model-generated forecasts so the smoothing effect can be compared directly on the same axes.

1. Diagram
   Observation

   Cell C4 is selected and filled with 1103.

2. Audio
   Observation

   The speaker says that with no previous week, the actual from that week is used as the forecast.

1. Week 1 row.

2. Cell C4 in the Forecast column.

3. Value 1103.

1. The previously empty Week 1 forecast cell becomes 1103.

1. Week 1 actual remains 1103.

2. The smoothing constant remains 0.2.

This step supplies the seed value required before the recursive formula can begin at Week 2.

1. Diagram
   Observation

   The user types =C4+$B$$2*(B4-C4)$ into cell C5.

2. Audio
   Observation

   The speaker narrates the formula as previous forecast plus alpha times previous actual minus previous forecast and notes locking alpha for dragging.

1. Cell C5.

2. Formula bar.

3. References C4, $B$2, and B4.

1. The Week 2 forecast cell changes from empty to an entered formula.

2. The alpha reference is made absolute as $B$2.

1. The underlying recurrence remains $F(t)=F(t-1)$+alpha($A(t-1)-F(t-1)$).

2. The Week 1 seed value remains 1103.

The screen recording demonstrates how the abstract recurrence is mapped to spreadsheet cell references, with relative references moving down rows and the constant pinned in place.

1. Animation
   Observation

   The user double-clicks the bottom-right corner of cell C5, causing the formula to fill down column C.

2. Animation
   Observation

   Simultaneously, a red line representing the 'Forecast' appears and extends across the chart.

1. Excel spreadsheet

2. Line chart

3. Blue line (Unit Sales)

4. Red line (Forecast)

1. Column C is populated with forecast values from row 5 to row 15.

2. A red line appears on the chart, plotting the forecast values against the weeks.

1. The blue line representing actual unit sales remains unchanged.

2. The smoothing constant in cell B4 remains 0.2.

This visual event demonstrates how applying the exponential smoothing formula to the entire dataset generates a smoothed trend line that contrasts with the volatile actual data.

1. Audio
   Observation

   The speaker says the formula looks very complicated and that going through an example should clear everything up.

Learners may assume the exponential smoothing formula is difficult once it is written symbolically.

The video addresses this by immediately moving from the symbolic formula to a concrete Excel example, showing that the recurrence is applied mechanically from one row to the next.

1. Audio
   Observation

   The speaker explicitly handles the case where there is no previous week by using the actual from that week as the forecast.

A learner might try to apply the recurrence at the first period even though $F(t-1)$ does not exist yet.

The video shows a separate initialization step: set $F(1)=A(1)$ before using the recurrence from Week 2 onward.

1. Formula
   Observation

   The general formula is displayed above the table.

2. Diagram
   Observation

   The Excel formula =C4+$B$$2*(B4-C4)$ is entered in the Forecast column.

The Excel entry is a direct application of the mathematical recurrence to spreadsheet cells.

1. Diagram
   Observation

   Cell B2 contains the smoothing constant 0.2.

2. Formula
   Observation

   The recurrence multiplies the error term by alpha.

The recurrence cannot be evaluated numerically without first specifying the smoothing constant alpha.

1. Audio
   Observation

   The speaker initializes Week 1 before saying Week 2 is when the formula starts being used.

2. Diagram
   Observation

   C4 is filled before C5 receives the recursive formula.

The recursive formula requires a previous forecast, so the initialization rule supplies the starting value needed to begin the recursion.

1. Diagram
   Observation

   The chart legend distinguishes Unit Sales and Forecast.

2. Audio
   Observation

   The speaker says forecasts will populate in red on the graph as they are added.

The computed forecast column feeds the red chart series, allowing the recursive results to be compared visually with the blue actual-sales series.

1. Formula
   Observation

   Both the mathematical formula and the Excel formula are shown simultaneously.

The Excel implementation is a direct application of the mathematical exponential smoothing formula.

1. Formula
   Observation

   The formula $F(t) = F(t-1)$ + alpha * ($A(t-1) - F(t-1)$) is shown on screen.

2. Audio
   Observation

   The speaker explains each term in words.

1. Exponential smoothing forecast formula
2. Smoothing constant alpha

1. Audio
   Observation

   The speaker says to use the actual from that week as the forecast when there is no previous week.

2. Diagram
   Observation

   Cell C4 is filled with 1103.

1. Initialization rule for the first forecast

1. Diagram
   Observation

   The formula =C4+$B$$2*(B4-C4)$ is typed into cell C5.

2. Audio
   Observation

   The speaker explains locking alpha so it stays consistent when dragging the formula.

1. Implementing exponential smoothing in Excel
2. Exponential smoothing forecast formula

1. Audio
   Observation

   The speaker calls alpha the exponential smoothing constant.

2. Diagram
   Observation

   Cell B2 is labeled Smoothing Constant and contains 0.2.

1. Smoothing constant alpha

1. Audio
   Observation

   The speaker states the goal is to find the Week 11 forecast.

2. Diagram
   Observation

   Only the Week 1 seed and Week 2 formula setup are completed before the clip ends.

1. The clip ends before the Week 11 numerical result is shown.

1. Sales forecasting example with alpha = 0.2

1. Diagram
   Observation

   The chart legend labels Unit Sales and Forecast.

2. Audio
   Observation

   The speaker says forecasts will appear in red on the graph.

1. Chart legend and color coding

1. Formula
   Observation

   The video explicitly shows the Excel formula and the auto-fill action.

1. Implementing Exponential Smoothing in Excel
2. Mapping the mathematical formula to the Excel syntax

1. Formula
   Observation

   The mathematical formula is displayed as text on the screen.

1. Exponential Smoothing Formula