Sales Forecasting Methods

We apply four forecasting methods to the given sales series (Feb–Sep). For each, we compute forecasts,
then compare to actuals (Mar–Sep) to get errors and accuracy metrics (MAD, MSE, MAPE). The standard
formulas are:
- MAD = (Σ|Actual–Forecast|)/n 1 ,
- MSE = (Σ(Actual–Forecast)²)/n 2 ,
- MAPE = (100%/n) Σ|Actual–Forecast|/Actual 3 .

Naïve Method
By definition, the naive forecast for each period is simply the actual value of the previous period 4 . Thus:
- F_Mar = Actual_Feb = 450
- F_Apr = Actual_Mar = 30
- F_May = 385, F_Jun = 740, F_Jul = 1000, F_Aug = 950,
F_Sep = 1400 (each equals the prior month’s actual).

The table below shows actual vs forecast, with errors. We then compute MAD, MSE, and MAPE using all 7
points (Mar–Sep):

Month Actual Forecast Abs Error Squared Error % Error

Mar 30 450 420 176400 1400.00%

Apr 385 30 355 126025 92.21%

May 740 385 355 126025 47.97%

Jun 1000 740 260 67600 26.00%

Jul 950 1000 50 2500 5.26%

Aug 1400 950 450 202500 32.14%

Sep 1800 1400 400 160000 22.22%

From these errors:

- MAD = (420+355+355+260+50+450+400)/7 = 327.14.
- MSE = (176400+126025+…+160000)/7 = 123007.14.
- MAPE = (1/7)Σ(|Error|/Actual×100%) = 232.26%.

Exponential Smoothing (α = 0.3)

In simple exponential smoothing, each forecast is a weighted average of the last actual and the previous
forecast 5 . We take F_Mar = Actual_Feb = 450 as the initial forecast. Then:

1
- F_Apr = 0.3·Actual_Mar + 0.7·F_Mar = 0.3·30 + 0.7·450 = 324.00.
- F_May = 0.3·385 + 0.7·324.00 = 342.30.
- F_Jun = 0.3·740 + 0.7·342.30 = 461.61.
- F_Jul = 0.3·1000 + 0.7·461.61 = 623.13.
- F_Aug = 0.3·950 + 0.7·623.13 = 721.19.
- F_Sep = 0.3·1400 + 0.7·721.19 = 924.83.

Comparing to actuals:

Month Actual Forecast Abs Error Squared Error % Error

Mar 30 450.00 420.00 176400.000 1400.00%

Apr 385 324.00 61.00 3721.000 15.84%

May 740 342.30 397.70 158165.290 53.74%

Jun 1000 461.61 538.39 289863.792 53.84%

Jul 950 623.13 326.87 106845.958 34.41%

Aug 1400 721.19 678.81 460784.509 48.49%

Sep 1800 924.83 875.17 765918.626 48.62%

Calculating the metrics:

- MAD = (420.00+61.00+397.70+538.39+326.87+678.81+875.17)/7 = 471.13.
- MSE = (176400+3721+158165.29+...+765918.63)/7 = 280242.74.
- MAPE = (1/7)(1400.00+15.84+53.74+53.84+34.41+48.49+48.62) = 236.42%.

3-Month Moving Average

A 3-month moving average forecast is the simple average of the last three actual values 6 . (No forecast is
available for Mar/Apr because we lack 3 prior months.) The first forecast is for May using Feb–Apr:
- F_May = (450 + 30 + 385)/3 = 285.00.
- F_Jun = (30 + 385 + 740)/3 = 385.00.
- F_Jul = (385 + 740 + 1000)/3 = 708.33.
- F_Aug = (740 + 1000 + 950)/3 = 896.67.
- F_Sep = (1000 + 950 + 1400)/3 = 1116.67.

Month Actual Forecast Abs Error Squared Error % Error

Mar 30 – – – –

Apr 385 – – – –

May 740 285.00 455.00 207025.00 61.49%

Jun 1000 385.00 615.00 378225.00 61.50%

2
 Month Actual Forecast Abs Error Squared Error % Error

Jul 950 708.33 241.67 58402.78 25.44%

Aug 1400 896.67 503.33 253344.44 35.95%

Sep 1800 1116.67 683.33 466944.44 37.96%

(We omit Mar/Apr errors since forecasts begin in May.) Using May–Sep (5 points):
- MAD = (455.00+615.00+241.67+503.33+683.33)/5 = 499.67.
- MSE = (207025+378225+58402.78+253344.44+466944.44)/5 = 272788.33.
- MAPE = (1/5)(61.49+61.50+25.44+35.95+37.96)% = 44.47%.

Weighted Moving Average (weights 0.5, 0.3, 0.2)

Here the forecast uses a weighted average of the last three actuals 7 (weights 0.5 on most recent, 0.3 on
next, 0.2 on third). Again, first forecast is May:
- F_May = 0.5·Actual_Apr + 0.3·Actual_Mar + 0.2·Actual_Feb
= 0.5·385 + 0.3·30 + 0.2·450 = 291.50.
- F_Jun = 0.5·740 + 0.3·385 + 0.2·30 = 491.50.
- F_Jul = 0.5·1000 + 0.3·740 + 0.2·385 = 799.00.
- F_Aug = 0.5·950 + 0.3·1000 + 0.2·740 = 923.00.
- F_Sep = 0.5·1400 + 0.3·950 + 0.2·1000 = 1185.00.

Month Actual Forecast Abs Error Squared Error % Error

Mar 30 – – – –

Apr 385 – – – –

May 740 291.50 448.50 201152.25 60.61%

Jun 1000 491.50 508.50 258572.25 50.85%

Jul 950 799.00 151.00 22801.00 15.89%

Aug 1400 923.00 477.00 227529.00 34.07%

Sep 1800 1185.00 615.00 378225.00 34.17%

Using May–Sep forecasts (5 points):

- MAD = (448.50+508.50+151.00+477.00+615.00)/5 = 440.00.
- MSE = (201152.25+258572.25+22801+227529+378225)/5 = 217655.90.
- MAPE = (1/5)(60.61+50.85+15.89+34.07+34.17)% = 39.12%.

Final Forecasts (Oct & Nov)

Using each method, we project two steps ahead. For October we use the last known actuals (Sep and
earlier). For November, with no actual Oct available, we use the method’s rules (often plugging in the Oct
forecast). The resulting forecasts are:

3
 Method Oct Forecast Nov Forecast

Naïve 1800.00 1800.00

Exponential Smoothing (α=0.3) 1187.38 1187.38

3-Month Moving Average 1383.33 1527.78

Weighted MA (0.5,0.3,0.2) 1510.00 1575.00

• Naïve: Oct = actual Sep = 1800; Nov assumed = 1800 (carry forward) 4 .
• Exp. Smoothing: Oct = 0.3·1800 + 0.7·924.83 ≈ 1187.38; Nov forecast stays at 1187.38 (flat) 8 .
• 3-MA: Oct = average(Jul,Aug,Sep) = (950+1400+1800)/3 = 1383.33; Nov =
average(Aug,Sep,Oct_forecast) = (1400+1800+1383.33)/3 = 1527.78.
• Weighted MA: Oct = 0.5·1800 + 0.3·1400 + 0.2·950 = 1510.00; Nov = 0.5·Oct_forecast + 0.3·1800 +
0.2·1400 = 1575.00.

These tables and calculations give a complete view of each method’s forecasts and accuracy on the
historical data.

Sources: Definitions and formulas for naive forecasting 4 , moving averages 6 7 , and error metrics 1

2 3 were used.

1 2 3 Use Excel to Calculate MAD, MSE, RMSE & MAPE - Dawn Wright, Ph.D.
[Link]

4 Naive Forecasting in Excel: Step-by-Step Example

[Link]

5 8 Chapter 8 Exponential smoothing | Notes for “Forecasting: Principles and Practice, 3rd edition”
[Link]

6 7 Time Series Forecasting | GoodData

[Link]