Chapter 03 - Forecasting

CHAPTER 3
FORECASTING

Discussion Questions

1. Why is forecasting necessary in OSCM?

Companies need to plan for the future to ensure they are able to meet the needs of the market
efficiently and effectively. Most decisions take some time to implement. Facility decisions may
take more than a year to implement, while distribution decisions may only take a week or so.
Because of this lead time, firms cannot simply wait until a need arises to make a decision on
how to deal with that need. Instead, they need to plan ahead, and forecasts allow them to do
this based on the best information available.

2. It is a common saying that the only thing certain about a forecast is that it will be wrong. What
is meant by this?

Because forecasting involves making predictions about the future, there is no way for a forecast
to reliably be perfect. There are too many unknowns about the future, which makes the entire
planning (for the future) process in supply chain management a very challenging process.

3. From the choice of simple moving average, weighted moving average, exponential smoothing,
and linear regression analysis, which forecasting technique would you consider the most
accurate? Why?

The answer depends on the nature of the demand you are trying to forecast. If there is any
trend to the demand, a simple moving average will always lag behind the actual demand. On
the other hand, if demand is relatively flat over the long run with only random variations from
one period to the next, the latter three methods will always be adjusting the forecast in
response to random variation. In this case a simple moving average will be the best choice. If
there are seasonal or cyclical components to demand, then none of these methods will be
particularly good as a forecasting tool.

4. All forecasting methods using exponential smoothing, adaptive smoothing, and exponential
smoothing including trend require starting values to get the equations going. How would you
select the starting value for, say Ft-1?

Starting values can be simply an average of the early periods, or a guess. If the starting value is
taken some period back (as opposed to starting to use the equations on very recent data) the
equation will have a chance to adjust as it is carried forward to today.

03-1
Chapter 03 - Forecasting

5. How is a seasonal index computed from a regression line analysis?

Seasonal index is equal to the actual value (data point) divided by the value computed from the
regression line. To lessen the effects of random errors, the indices may be averaged over
several years for that same period.

6. Discuss the basic differences between the mean absolute deviation and mean absolute percent
error.

The mean absolute percent error can be used to compare forecasting accuracy when the
average demand for the items in each forecast is different. The mean absolute percent error
(MAPE) is the expected error measured as a fraction of demand, whereas the MAD is just the
average error of the forecast. The MAPE is usually a better measure.

7. What implications do forecast errors have for the search for ultra-sophisticated statistical
forecasting models?

The existence of unavoidable forecast errors seems to suggest that no matter what kind of
model one uses—simple or sophisticated—a perfect forecast is unattainable. Since forecasts
are predictions of the future based on present and past data, there is ample opportunity for
very serious forecast errors to be caused by changes in the conditions that generated the data.
This could lead to an invalid forecast or at least one that contains added error. Therefore, one
could be easily persuaded to stop searching for ways to make more accurate forecasts and look
instead for ways to quickly respond and adapt to demand changes.

8. Causal relationships are potentially useful for which component of a time series?

A time series creates an equation, such as y = a + bx, where a is the y-intercept. Therefore, if
there is a relationship between y and x, it would show up as the slope b. If there is no
relationship, b would be zero. There is some question as to the relationship being truly
“causal,” since many relationships may depend on other factors outside of the analysis.

9. Let’s say you work for a company that makes prepared breakfast cereals like corn flakes. Your
company is planning to introduce a new hot breakfast product made from whole grains that
would require some minimal preparation by the consumer. This would be a completely new
product for the company. How would you propose forecasting initial demand for this product?

Student answers will vary, but they should generally be centered on the qualitative techniques
discussed in the chapter.

10. How has the development of the Internet affected the way companies forecast in support of
their supply chain planning process?

Answers will vary, but basically this comes down to the availability of immediate and large
scale collaboration in the forecasting process thanks to the Internet. Students might draw an
analogy between this and the benefits of ERP in global supply chains.

03-2
Chapter 03 - Forecasting

11. What sorts of risks do you see in reliance on the Internet in the use of Collaborative Planning,
Forecasting, and Replenishment (CPFR)?

Students will likely come up will several ideas, but information security risks should be a
common part of their responses. Ensuring that your systems and the information they share
are secure from prying eyes and malicious hackers should be a primary concern of any I.T.
department.

03-3
Chapter 03 - Forecasting

Objective Questions

1. What is the term for forecasts used for making day-to-day decisions about meeting demand?

Tactical forecasts

2. What category of forecasting techniques uses managerial judgment in lieu of numerical data?

Qualitative techniques

3. Given the following history, use a three-quarter moving average to forecast the demand for the
third quarter of this year. Note that the 1st quarter is Jan, Feb, and Mar; 2nd quarter Apr, May,
Jun; 3rd quarter Jul, Aug, Sep; and 4th quarter Oct, Nov, Dec.

Three quarters ago (Oct, Nov, Dec last year) = 200 + 225 + 250 = 675
Two quarters ago (Jan, Feb, Mar this year) = 125 + 135 + 135 = 395
One quarter ago (Apr, May, Jun this year) = 190 + 200 + 190 = 580

For Jul, Aug, & Sep, using a three-quarter moving average the forecast would be = (675 + 395 + 580)/3 =
550.

4. Here are the data for the past 21 months for actual sales of a particular product:

Develop a forecast for the fourth quarter using a three-quarter, weighted moving average.
Weight the most recent quarter .5, the second most recent .25, and the third .25. Solve the
problem using quarters, as opposed to forecasting separate months.

Third most recent quarter: 275 + 375 +350 = 1000

Second most recent quarter: 425 +400 + 350 = 1175
Most recent quarter: 350 + 275 + 350 = 975

WMA = (.25*1000) + (.25 * 1175) + (.50 * 975) = 1031.25

03-4
Chapter 03 - Forecasting

5. The following table contains the number of complaints received in a department store for the
first 6 months of operation:

If a three-month moving average is used to smooth this series, what would have been the
forecast for May?

Use the actual demand from February through April to develop May’s forecast:

(45 + 81 + 90)/3 = 72

6. The following tabulations are actual sales of units for six months and a starting forecast in
January.

a. Calculate forecasts for the remaining five months using simple exponential smoothing
with  = 0.2.

Ft+1 = Ft + (At – Ft),  = .20

Month Demand Forecast Absolute Deviation

January 100 80 20
February 94 84 10
March 106 86 20
April 80 90 10
May 68 88 20
June 94 84 10

Total 90

b. Calculate MAD for the forecasts.

MAD = 90/6 = 15

03-5
Chapter 03 - Forecasting

03-6
Chapter 03 - Forecasting

7. The following table contains the demand from the last 10 months:

a. Calculate the single exponential smoothing forecast for these data using an  of .30
and an initial forecast (F1) of 31.

b. Calculate the exponential smoothing with trend forecast for these data using an 
of .30, a  of .30, an initial trend forecast (T1) of 1, and an initial exponentially
smoothed forecast (F1) of 30.

c. Calculate the mean absolute deviation (MAD) for each forecast. Which is best?

Exponential Absolute Absolute

Month Demand smoothing deviation Tt Ft FITt deviation
1 31 31.00 1.00 30.00 31.00
2 34 31.00 3.00 1.00 31.00 32.00 2.00
3 33 31.90 1.10 1.18 32.60 33.78 0.78
4 35 32.23 2.77 1.11 33.55 34.66 0.34
5 37 33.06 3.94 1.14 34.76 35.90 1.10
6 36 34.24 1.76 1.24 36.23 37.47 1.47
7 38 34.77 3.23 1.11 37.03 38.14 0.14
8 40 35.74 4.26 1.10 38.10 39.19 0.81
9 40 37.02 2.98 1.17 39.43 40.60 0.60
10 41 37.91 3.09 1.11 40.42 41.54 0.54

MAD 2.90 0.86

Based upon the MAD of each forecast, the exponential smoothing with trend is the better
forecasting model.

8. Actual demand for a product for the past three months was:

a. Using a simple three-month moving average, make a forecast for this month.

F (this month) = (325 + 350 + 400)/3 = 358

b. If 300 units were actually demanded this month, what would your forecast be for next
month?

03-7
Chapter 03 - Forecasting

F (next month) = (300 + 325 + 350)/3 = 325

c. Using simple exponential smoothing, what would your forecast be for this month if the
exponentially smoothed forecast for three months ago was 450 units and the
smoothing constant was 0.20?

F (two months ago) = 450 + .20(400 – 450) = 440

F (one month ago) = 440 + .20(350 – 440) = 422

F (this month) = 422 + .20(325 – 422) = 403

9. Assume an initial starting Ft of 300 units, a trend (Tt ) of 8 units, an alpha of 0.30, and a delta of
0.40. If actual demand turned out to be 288, calculate the forecast for the next period.

Ft =300T t =8
α=.30 δ=. 40
A t =288
FIT t =F t +T t =300+8=308
Ft +1=FIT t +α( A t −FIT t )=308+. 3(288−308 )=308−6=302
T t +1=T t +δ( F t+1 −FIT t )=8+.4 (302−308 )=8−2. 4=5 .6
FIT t +1=F t+1 +T t+1 =302+5. 6=307 .6

10. The number of cases of merlot wine sold by the Connor Owen winery in an eight-year period is
as follows:

Year Cases of Wine Year Cases of Wine

1 270 5 358
2 356 6 500
3 398 7 410
4 456 8 376

Using an exponential smoothing model with an alpha value of 0.20, estimate the smoothed
value calculated as of the end of 8. Use the average demand for years 1 through 3 as your
initial forecast, and then smooth the forecast forward to year 8.

Year Demand F(t)*

1 270
2 356
3 398
4 456 341
5 358 364
6 500 363
7 410 390
8 376 394
* Forecasts have been rounded to integer values, which may result in minor rounding differences.

03-8
Chapter 03 - Forecasting

F(4) is calculated as average demand from years 1-3. Later forecasts are based on a simple exponential smoothing
model with alpha = .20.

11. Not all the items in your office supply store are evenly distributed as far as demand is
concerned, so you decide to forecast demand to help plan your stock. Past data for legal-sized
yellow tablets for the month of August area.

Week 1 300 Week 3 600

Week 2 400 Week 4 700

a. Using a three-week moving average, what would you forecast the next week to be?

F5 = (700 + 600 + 400)/3 = 567

b. Using exponential smoothing with  = 0.20, if the exponential forecast for week 3 was
estimated as the average of the first two weeks [(300 + 400)/2 = 350], what would you
forecast week 5 to be?

F4 = F3 + (A3 – F3) = 350 + .20(600 – 350) = 400

F5 = F4 + (A4 – F4) = 400 + .20(700 – 400) = 460

12. Assume that your stock of sales merchandise is maintained based on the forecast demand. If
the distributor’s sales personnel call on the first day of each month, compute your forecast
sales by each of the three methods requested here.

a. Using a simple three-month moving average, what is the forecast for September?

FSeptember = (170 + 180 + 140)/3 = 163.3

b. Using a weighted moving average, what is the forecast for September with weights of .
20, .30, and .50 for June, July, and August, respectively?

FSeptember = .50(170) + .30(180) + .20(140) = 167.0

c. Using single exponential smoothing and assuming that the forecast for June had been
130, forecast sales for September with a smoothing constant  of .30.

FJuly = FJune + (AJune – FJune) = 130 + .3(140 - 130) = 133.00

FAugust = FJuly + (AJuly – FJuly) = 133.00 + .3(180 – 133.00) = 147.10

03-9
Chapter 03 - Forecasting

FSeptember = FAugust + (AAugust – FAugust) = 147.10 + .3(170 – 147.10) = 153.97

03-10
Chapter 03 - Forecasting

13. Historical demand for a product is as follows:

a. Using a simple four-month moving average, calculate a forecast for October.

FOctober = (75 + 80 + 60 + 75)/4 = 72.5

b. Using single exponential smoothing with  = 0.2 and a September forecast = 65,
calculate a forecast for October.

FOctober = FSeptember + (ASeptember – FSeptember) = 65 + .2(75 – 65) = 67.0

c. Using simple linear regression, calculate the trend line for the historical data. Say the X
axis is April = 1, May = 2, and so on, while the Y axis is demand.

y = 405/6 = 67.5
t = 21/6 = 3.5
∑ ty−nt y =1485−6 (3. 5)67 .5
b= ∑ t 2−nt̄ 2 91−6(3.5)2 = 3.86
a= y−bt = 67.5 – 3.86(3.5) = 54.00
Yt = a + bt = 54.0 + 3.86t

Using Excel, the intercept a can be found using the INTERCEPT() function. The slope b can be found with the
SLOPE() function.

d. Calculate a forecast for October using your regression formula.

FOctober = 54.00 + 3.86(7) = 81.01

03-11
Chapter 03 - Forecasting

14. Demand for stereo headphones and MP3 players for joggers has caused Nina Industries to
grow almost 50 percent over the past year. The number of joggers continues to expand, so
Nina expects demand for headsets to also expand, because, as yet, no safety laws have been
passed to prevent joggers from wearing them. Demand for the players for last year was as
follows:

a. Using linear regression analysis, what would you estimate demand to be for each
month next year? Using a spreadsheet, follow the general format in Exhibit 3.8.
Compare your results to those obtained by using the forecast spreadsheet
function.

t y ty t2 y2 Y (y-Y)2
1 4200 4200 1 17640000 3958.97 58093.360
2 4300 8600 4 18490000 4151.28 22117.028
3 4000 12000 9 16000000 4343.59 118053.912
4 4400 17600 16 19360000 4535.90 18468.113
5 5000 25000 25 25000000 4728.21 73872.452
6 4700 28200 36 22090000 4920.51 48625.904
7 5300 37100 49 28090000 5112.82 35036.160
8 4900 39200 64 24010000 5305.13 164128.863
9 5400 48600 81 29160000 5497.44 9493.754
10 5700 57000 100 32490000 5689.74 105.194
11 6300 69300 121 39690000 5882.05 174681.131
12 6000 72000 144 36000000 6074.36 5529.257
Total 78 60200 418800 650 308020000 728205.128

t = 6.5
y =
5016.667

∑ ty−nt y
2 2 192.3077
b = ∑ t −n t̄ =
a = y−bt =
3766.667

03-12
Chapter 03 - Forecasting

Month Forecast

13 6266.67
14 6458.97
15 6651.28
16 6843.59
17 7035.90
18 7228.21
19 7420.51
20 7612.82
21 7805.13
22 7997.43
23 8189.74
24 8382.05
Using Excel, the FORECAST() function comes up with the same answers as this regression model, so it seems
obvious that the function uses linear regression as its methodology.

b. To be reasonably confident of meeting demand, Nina decides to use three standard

errors of estimate for safety. How many additional units should be held to meet
this level of confidence?

S yt =
√ ∑ ( y i−Y i )2
i=1
n−2
=
√ 728205 . 128
12−2
=269 . 85

Therefore, 3 standard errors of the estimate would be 3(269.85) = 809.55 ≈ 810

15. Historical demand for a product is:

a. Using a weighted moving average with weights of 0.60, 0.30, and 0.10, find the July
forecast.

FJuly = .60(15) + .30(16) + .10(12) = 15.0

b. Using a simple three-month moving average, find the July forecast.

FJuly = (15 + 16 + 12) / 3 = 14.3

03-13
Chapter 03 - Forecasting

c. Using single exponential smoothing with  = 0.2 and a June forecast = 13, find the July
forecast. Make whatever assumptions you wish.

FJuly = FJune + (AJune – FJune) = 13 + .2(15-13) = 13.4

d. Using simple linear regression analysis, calculate the regression equation for the preceding
demand data.

t y ty t2
1 12 12 1
2 11 22 4
3 15 45 9
4 12 48 16
5 16 80 25
6 15 90 36
Total 21 81 297 91
Average 3.5 13.5

t = 3.5
y = 13.5
∑ ty−nt y =
0.77
b= ∑ t 2−nt 2
a= y−bt = 10.8

e. Using the regression equation in d, calculate the forecast for July.

Yt = a + bt = 10.8 + .77t = 10.8 + .77(7) = 16.2

16.

03-14
Chapter 03 - Forecasting

16. The tracking signals computed using past demand history for three different products are as
follows. Each product used the same forecasting technique.

Discuss the tracking signals for each and what the implications are.

TS 1:

Since there has been a rapid rise in the trend, the forecast will shortly be outside of the limits.
Therefore, the forecasting model is poor.

03-15
Chapter 03 - Forecasting

TS 2:

This is within the limits. Therefore, the forecast is acceptable.

TS 3

TS 3: This series is rising rapidly, and is outside of the limits. Consequently, the model is poor.

03-16
Chapter 03 - Forecasting

17. Here are the actual tabulated demands for an item for a nine-month period (January through
September). Your supervisor wants to test two forecasting methods to see which method was
better over this period.

a. Forecast April through September using a three-month moving average.

b. Use simple exponential smoothing with an alpha of .3 to estimate April through

September, using the average of January through March as the initial forecast for April.

c. Use MAD to decide which method produced the better forecast over the six-month
period.

For the exponential smoothing forecast we need a beginning forecast – this solution
uses the average of the first three months demand for the April forecast and the
exponential smoothing model for the remaining forecasts. Other choices will produce
different answers.

Month Demand 3-Mo. Absolute Exponential Absolute

MA Deviation Smoothing Deviation
January 110
February 130
March 150
April 170 130 40 130 40
May 160 150 10 142 18
June 180 160 20 147.4 32.6
July 140 170 30 157.18 17.18
August 130 160 30 152.03 22.03
September 140 150 10 145.42 5.42

MAD 23.3 22.5

Based upon MAD, the exponential smoothing model appears to be slightly better.

03-17
Chapter 03 - Forecasting

18. A particular forecasting model was used to forecast a six-month period. Here are the
forecasts and actual demands that resulted:

Find the tracking signal and state whether you think the model being used is giving
acceptable answers.

Sum of
Absolute absolute
Month Forecast Actual Deviation RSFE deviation deviations MAD TS
April 250 200 -50 -50 50 50 50.0 -1
May 325 250 -75 -125 75 125 62.5 -2
June 400 325 -75 -200 75 200 66.7 -3
July 350 300 -50 -250 50 250 62.5 -4
August 375 325 -50 -300 50 300 60.0 -5
September 450 400 -50 -350 50 350 58.3 -6

1 2 3 4 5 6
0

-1

-2

-3
TS

-4

-5

-6

-7
Period

For September, the MAD is 58.3 and the TS is -6. The model is performing poorly since the tracking
signal is -6 and moving in a downward direction. The model is consistently over-forecasting
demand.

03-18
Chapter 03 - Forecasting

19. Harlen Industries has a simple forecasting model: Take the actual demand for the same
month last year and divide that by the number of fractional weeks in that month. This gives
the average weekly demand for that month. This weekly average is used as the weekly
forecast for the same month this year. This technique was used to forecast eight weeks for
this year, which are shown below along with the actual demand that occurred. The
following eight weeks show the forecast (based on last year) and the demand that actually
occurred:

a. Compute the MAD of forecast errors.

b. Using the RSFE, compute the tracking signal.

c. Based on your answers to parts (a) and (b), comment on Harlen’s method of
forecasting.

Absolute Sum of Absolute

Month Forecast Actual Deviation RSFE deviation deviations MAD TS
1 140 137 -3 -3 3 3 3.00 -1.00
2 140 133 -7 -10 7 10 5.00 -2.00
3 140 150 10 0 10 20 6.67 0.00
4 140 160 20 20 20 40 10.00 2.00
5 140 180 40 60 40 80 16.00 3.75
6 150 170 20 80 20 100 16.67 4.80
7 150 185 35 115 35 135 19.29 5.96
8 150 205 55 170 55 190 23.75 7.16

4
TS

0
1 2 3 4 5 6 7 8
-2

-4

Period

a. For month 8, the MAD is 23.75

b. The tracking signal for month 8 is 7.16

c. The tracking signal is too large, so the forecast should be considered poor. It is not
effectively dealing with an apparent upward trend in demand.

03-19
Chapter 03 - Forecasting

20. In this problem, you are to test the validity of your forecasting model. Here are the
forecasts for a model you have been using and the actual demands that occurred:

Use the method stated in the text to compute MAD and tracking signal. Then decide
whether the forecasting model you have been using is giving reasonable results.

Sum of
Absolute Absolute
Week Forecast Actual Deviation RSFE deviation deviations MAD TS
1 800 900 100 100 100 100 100 1.0
2 850 1000 150 250 150 250 125 2.0
3 950 1050 100 350 100 350 117 3.0
4 950 900 -50 300 50 400 100 3.0
5 1000 900 -100 200 100 500 100 2.0
6 975 1100 125 325 125 625 104 3.1

3.5
3
2.5
2
TS

1.5
1
0.5
For week 6, the MAD is 104, and the
0
1 2 3 4 5 6
tracking signal is 3.1. This is a fairly high
Period value, which indicates the model is
unacceptable.

03-20
Chapter 03 - Forecasting

21. The following table shows predicted product demand using your particular forecasting
method along with the actual demand that occurred:

a. Compute the tracking signal using the mean absolute deviation and running sum of
forecast errors.

b. Discuss whether your forecasting method is giving good predictions.

Absolute Sum of Absolute

Period Forecast Actual Deviation RSFE deviation deviations MAD TS
1 1500 1550 50 50 50 50 50.0 1.00
2 1400 1500 100 150 100 150 75.0 2.00
3 1700 1600 -100 50 100 250 83.3 0.60
4 1750 1650 -100 -50 100 350 87.5 -0.57
5 1800 1700 -100 -150 100 450 90.0 -1.67

2.5
2
1.5
1
0.5
TS

0
-0.5 1 2 3 4 5

-1
-1.5
-2
Period

a. For period 5, the MAD = 90.00, and the TS = -1.67

b. Looking solely at the value of the TS, the model seems acceptable since the tracking signal
is only 1.67 off the mean. However, the MAD has been increasing since the first period, and
the downward trend over the last several periods in the graph is cause for concern that
there may be some bias in the model.

03-21
Chapter 03 - Forecasting

22. Your manager is trying to determine what forecasting method to use. Based upon the
following historical data, calculate the following forecast and specify what procedure you
would utilize.

a. Calculate the simple three-month moving average forecast for periods 4–12.

b. Calculate the weighted three-month moving average using weights of 0.50, 0.30,
and 0.20 for periods 4–12.

c. Calculate the single exponential smoothing forecast for periods 2–12 using an
initial forecast (F1) of 61 and an  of 0.30.

d. Calculate the exponential smoothing with trend component forecast for periods 2–
12 using an initial trend forecast (T1) of 1.8, an initial exponential smoothing
forecast (F1) of 60, an  of 0.30, and a  of 0.30.

e. Calculate the mean absolute deviation (MAD) for the forecasts made by each
technique in periods 4–12. Which forecasting method do you prefer?

Month 3-mo. Absolute 3-mo Absolute Absolute Absolute

(t) Demand MA deviation WMA deviation Ft deviation Tt Ft FITt deviation
1 62 61.00 1.80 60.00 61.80
2 65 61.30 1.82 61.86 63.68
3 67 62.41 1.94 64.07 66.01
4 68 64.67 3.33 65.40 2.60 63.79 4.21 2.03 66.31 68.33 0.33
5 71 66.67 4.33 67.10 3.90 65.05 5.95 2.00 68.23 70.23 0.77
6 73 68.67 4.33 69.30 3.70 66.84 6.16 2.07 70.46 72.53 0.47
7 76 70.67 5.33 71.40 4.60 68.68 7.32 2.11 72.67 74.78 1.22
8 78 73.33 4.67 74.10 3.90 70.88 7.12 2.22 75.14 77.36 0.64
9 78 75.67 2.33 76.40 1.60 73.02 4.98 2.28 77.55 79.83 1.83
10 80 77.33 2.67 77.60 2.40 74.51 5.49 2.11 79.28 81.39 1.39
11 84 78.67 5.33 79.00 5.00 76.16 7.84 1.99 80.98 82.96 1.04
12 85 80.67 4.33 81.60 3.40 78.51 6.49 2.08 83.27 85.35 0.35

MAD 4.07 3.46 6.17 0.89

Based upon MAD, the exponential smoothing with trend component appears to be the best
method. This should not be a surprise given the apparent upward trend.

03-22
Chapter 03 - Forecasting

23. After using your forecasting model for six months, you decide to test it using MAD and a
tracking signal. Here are the forecast and actual demands for the six-month period:

a. Find the tracking signal.

Sum of
Absolute absolute
Month Forecast Actual Deviation RSFE deviation deviations MAD TS
May 450 500 50 50 50 50 50.00 1.00
June 500 550 50 100 50 100 50.00 2.00
July 550 400 -150 -50 150 250 83.33 -0.60
August 600 500 -100 -150 100 350 87.50 -1.71
September 650 675 25 -125 25 375 75.00 -1.67
October 700 600 -100 -225 100 475 79.17 -2.84

0
1 2 3 4 5 6
TS

-1

-2

-3

-4
Period

b. Decide whether your forecasting routine is acceptable.

The TS itself is acceptable. However, you would like to see the TS going back and forth
between positive and negative. It has been headed primarily downward since June. If
this trend continues, the forecasts will be unacceptable. This forecast should be closely
monitored to see if the downward trend continues, or if this occurred by random
chance.

03-23
Chapter 03 - Forecasting

24. Zeus Computer Chips, Inc., used to have major contracts to produce the Centrino-type
chips. The market has been declining during the past three years because of the quad-core
chips, which it cannot produce, so Zeus has the unpleasant task of forecasting next year.
The task is unpleasant because the firm has not been able to find replacement chips for its
product lines. Here is demand over the past 12 quarters:

Use the decomposition technique to forecast demand for the next four quarters.

Average from Seasonal Deseasonalized t*deseasonalized

t y same quarter factor demand t2 demand
1 4800 3833.33 1.23 3902.61 1 3902.61
2 3500 2766.67 0.89 3942.77 4 7885.54
3 4300 3500.00 1.12 3829.05 9 11487.14
4 3000 2366.67 0.76 3950.70 16 15802.82
5 3500 1.23 2845.65 25 14228.26
6 2700 0.89 3041.57 36 18249.40
7 3500 1.12 3116.67 49 21816.67
8 2400 0.76 3160.56 64 25284.51
9 3200 1.23 2601.74 81 23415.65
10 2100 0.89 2365.66 100 23656.63
11 2700 1.12 2404.29 121 26447.14
12 1700 0.76 2236.84 144 26842.08
Sum 78 37400 37400.00 650 219041.20

y = 3116.67
t = 6.5

b=
∑ ty d−nt y d =−168 .24
∑ t 2−n t̄ 2
a= y d−b t=4210 .25

Forecast
Seasonal (Yt*seasonal
Period (t) Yt factor factor)
13 2023.08 1.23 2488.28
14 1854.84 0.89 1646.54
15 1686.60 1.12 1894.04
16 1518.35 0.76 1152.97

Calculations were done in Excel. Hand calculations may result in some rounding differences.

03-24
Chapter 03 - Forecasting

25. Sales data for two years are as follows. Data are aggregated with two months of sales in
each “period.”

a. a. Plot the data.

b. Fit a simple linear regression model to the sales data.

c. In addition to the regression model, determine multiplicative seasonal index

factors. A full cycle is assumed to be a full year.

03-25
Chapter 03 - Forecasting

b - c.

Average from
Period Sales same bi-monthly Seasonal Deseasonalized t*deseasonalized
t y period factor demand t2 t*y demand
1 109 112.0 0.865 125.95 1 109 125.95
2 104 108.0 0.835 124.62 4 208 249.25
3 150 154.5 1.194 125.65 9 450 376.94
4 170 176.0 1.360 125.00 16 680 500.02
5 120 123.0 0.950 126.26 25 600 631.30
6 100 103.0 0.796 125.65 36 600 753.88
7 115 0.865 132.88 49 805 930.18
8 112 0.835 134.21 64 896 1073.68
9 159 1.194 133.19 81 1431 1198.68
10 182 1.360 133.83 100 1820 1338.29
11 126 0.950 132.57 121 1386 1458.31
12 106 0.796 133.19 144 1272 1598.23
Total 78 1553 1553.00 650 10257 10234.70

Deseasonalized Simple
y = 129.4167
t = 6.5
b=
∑ ty t −n t y t =
0.9804 1.136364
∑ t 2−n t̄ 2
a=
y t −bt = 123.04 122.0303

d. Using the results from parts (b) and (c), prepare a forecast for the next year.

Simple Seasonal Seasonal

Period (t) Forecast Yt factor Forecast
13 136.80 135.79 0.865 117.5
14 137.94 136.77 0.835 114.1
15 139.08 137.75 1.194 164.4
16 140.21 138.73 1.360 188.7
17 141.35 139.71 0.950 132.8
18 142.48 140.69 0.796 112.0

03-26
Chapter 03 - Forecasting

26. The following table shows the past two years of quarterly sales information. Assume that
there are both trend and seasonal factors and that the seasonal cycle is one year. Use time
series decomposition to forecast quarterly sales for the next year.

Average from
Quarter Sales same quarterly Seasonal Deseasonalized t*deseasonalized
t y period factor demand t2 demand

1 160 187.5 1.003 159.47 1 159.47

2 195 217.5 1.164 167.54 4 335.09
3 150 177.5 0.950 157.92 9 473.77
4 140 165.0 0.883 158.56 16 634.24
5 215 1.003 214.28 25 1071.42
6 240 1.164 206.21 36 1237.24
7 205 0.950 215.83 49 1510.79
8 190 0.883 215.18 64 1721.52
Total 36 1495 1495.00 204 7143.53

y = 186.875
t = 4.5
∑ tyt −n t y t
9.91
b= ∑ t 2−nt̄ 2 =
a=
y t −bt = 142.30

Quarter Yt Seasonal factor Forecast

9 231.45 1.003 232
10 241.35 1.164 281
11 251.26 0.950 239
12 261.17 0.883 231

03-27
Chapter 03 - Forecasting

27. Tucson Machinery, Inc., manufactures numerically controlled machines, which sell for an
average price of $0.5 million each. Sales for these NCMs for the past two years were as
follows:

a. Find a line using regression in Excel.

b. Find the trend and seasonal factors.

c. Forecast sales for next year.

Within Excel, two methods can be used to compute the intercept and slope from a
regression model. We can use the Regression tool within Data Analysis in the Data
menu to perform a full-blown regression analysis, or we can use the INTERCEPT()
and SLOPE() functions to get a and b in a traditional regression model.

Using those functions we get:

a = 15.143 b = 1.024

We can now use decomposition to quantify the seasonal factors and forecast for
the next four quarters.

First procedure:

Trend Actual/
t y (a + tb) Trend Seasonal factor
1 12 16.17 0.742 0.766
2 18 17.19 1.047 1.087
3 26 18.21 1.427 1.341
4 16 19.24 0.832 0.802
5 16 20.26 0.790
6 24 21.29 1.128
7 28 22.31 1.255
8 18 23.33 0.771

Seasonal
Period Y factor Forecast
9 24.36 0.766 18.66
10 25.38 1.087 27.60
11 26.40 1.341 35.42
12 27.43 0.802 21.99

03-28
Chapter 03 - Forecasting

Second procedure: We could also deseasonalize the data first, and perform a regression on
the deseasonalized data. Full calculations for this alternate method are shown below:

Average from same Seasonal Deseasonalized t*deseasonalized

t y quarterly period factor demand t2 demand
1 12 14 0.71 16.93 1 16.93
2 18 21 1.06 16.93 4 33.86
3 26 27 1.37 19.02 9 57.06
4 16 17 0.86 18.59 16 74.35
5 16 0.71 22.57 25 112.86
6 24 1.06 22.57 36 135.43
7 28 1.37 20.48 49 143.37
8 18 0.86 20.91 64 167.29
Total = 36 158 158.00 204 741.14

y = 19.75
t = 4.5
∑ tyt −n t y t
0.718
b= ∑ t 2−nt̄ 2 =
a=
y t −bt = 16.520

Seasonal
Period Yt factor Forecast
9 22.98 0.71 16.29
10 23.70 1.06 25.20
11 24.42 1.37 33.38
12 25.13 0.86 21.63

03-29
Chapter 03 - Forecasting

28. Use regression analysis on deseasonalized demand to forecast next summer’s demand,
given the following historical demand data:

YEAR SEASON ACTUAL DEMAND

2 years ago Spring 205
Summer 140
Fall 375
Winter 575

Last year Spring 475

Summer 275
Fall 685
Winter 965

Season Demand Average from same Seasonal Deseasonalized t*deseasonalized

t y quarterly period factor demand t2 demand
1 205 340.0 0.736 278.48 1 278.48
2 140 207.5 0.449 311.63 4 623.25
3 375 530.0 1.147 326.80 9 980.40
4 575 770.0 1.667 344.91 16 1379.63
5 475 0.736 645.27 25 3226.33
6 275 0.449 612.12 36 3672.74
7 685 1.147 596.95 49 4178.66
8 965 1.667 578.84 64 4630.75
Total 36 3695 3695.00 204 18970.24

y = 461.88
t = 4.50

b=
∑ ty t −n t y t
55.78
∑ t 2−n t̄ 2 =
a=
y t −bt = 210.87

Forecast
Seasonal (Yt*seasonal
Season (t) Yt factor factor)
9 712.88 0.736 525
10 768.66 0.449 345
11 824.44 1.147 946
12 880.22 1.667 1467

03-30
Chapter 03 - Forecasting

29. Here are earnings per share for two companies by quarter from the first quarter of three
years ago through the second quarter of this year. Forecast earnings per share for the rest
of this year and next year. Use exponential smoothing to forecast the third period of this
year, and the time series decomposition method to forecast the last two quarters of this
year and all four quarters of next year. (It is much easier to solve this problem on a
computer spreadsheet so you can see what is happening.)

EARNINGS PER SHARE

QUARTER COMPANY A COMPANY B
3 years ago 1.67 0.17
II 2.35 0.24
III 1.11 0.26
IV 1.15 0.34
2 years ago I 1.56 0.25
II 2.04 0.37
III 1.14 0.36
IV 0.38 0.44
last year I 0.29 0.33
II -0.18 0.40
III -0.97 0.41
IV 0.20 0.47
this year I -1.54 0.30
II 0.38 0.47

a. For the exponential smoothing method, choose the first quarter of 3 years ago as
the beginning forecast. Make two forecasts: one with  = 0.10 and one with  =
0.30.

Company A

Forecast Absolute Forecast Absolute

Quarter EPS  = 0.10 deviation  = 0.30 deviation
3 years ago I 1.67 1.67 1.67
II 2.35 1.67 0.68 1.67 0.68
III 1.11 1.74 0.63 1.87 0.76
IV 1.15 1.68 0.53 1.64 0.49
2 years ago I 1.56 1.62 0.06 1.50 0.06
II 2.04 1.62 0.42 1.52 0.52
III 1.14 1.66 0.52 1.67 0.53
IV 0.38 1.61 1.23 1.51 1.13
last year I 0.29 1.48 1.19 1.17 0.88
II -0.18 1.36 1.54 0.91 1.09
III -0.97 1.21 2.18 0.58 1.55
IV 0.20 0.99 0.79 0.12 0.08
this year I -1.54 0.91 2.45 0.14 1.68
II 0.38 0.67 0.29 -0.36 0.74
III 0.64 -0.14
MAD 0.96 0.79

03-31
Chapter 03 - Forecasting

Company B

Forecast Absolute Forecast Absolute

Quarter EPS  = 0.10 deviation  = 0.30 deviation
3 years ago I 0.17 0.17 0.17
II 0.24 0.17 0.07 0.17 0.07
III 0.26 0.18 0.08 0.19 0.07
IV 0.34 0.19 0.15 0.21 0.13
2 years ago I 0.25 0.20 0.05 0.25 0.00
II 0.37 0.21 0.16 0.25 0.12
III 0.36 0.22 0.14 0.29 0.07
IV 0.44 0.24 0.20 0.31 0.13
last year I 0.33 0.26 0.07 0.35 0.02
II 0.40 0.26 0.14 0.34 0.06
III 0.41 0.28 0.13 0.36 0.05
IV 0.47 0.29 0.18 0.37 0.10
this year I 0.30 0.31 0.01 0.40 0.10
II 0.47 0.31 0.16 0.37 0.10
III 0.32 0.40
MAD 0.12 0.08

b. Using the MAD method of testing the forecasting model’s performance, plus actual
data from 3 years ago through the second quarter of this year, how well did the
models perform?

MAD
Company A Company B
 = 0.10 .96 .12
 = 0.30 .79 .08

Based upon MAD, an  of .30 performs better than .10.

03-32
Chapter 03 - Forecasting

c. Using the decomposition of a time series method of forecasting, forecast earnings

per share for the last two quarters of this year and all four quarters of next year. Is
there a seasonal factor in the earnings?

Company A

Average from Seasonal Deseasonalized t*deseasonalized

t y same quarter factor demand t2 demand
1 1.67 0.495 0.723 2.309 1 2.309
2 2.35 1.148 1.677 1.401 4 2.803
3 1.11 0.427 0.624 1.780 9 5.341
4 1.15 0.577 0.843 1.365 16 5.458
5 1.56 0.723 2.157 25 10.783
6 2.04 1.677 1.217 36 7.299
7 1.14 0.624 1.828 49 12.798
8 0.38 0.843 0.451 64 3.607
9 0.29 0.723 0.401 81 3.608
10 -0.18 1.677 -0.107 100 -1.073
11 -0.97 0.624 -1.556 121 -17.113
12 0.20 0.843 0.237 144 2.848
13 -1.54 0.723 -2.129 169 -27.676
14 0.38 1.677 0.227 196 3.172

Total 105 9.58 9.580 1015 14.165

y = 0.684
t = 7.5
∑ tyt −n t y t
-0.254
b= ∑ t 2−nt̄ 2 =
a=
y t −bt = 2.5867

Forecast
Seasonal (Yt*seasonal
Period (t) Yt factor factor)
15 -1.217 0.624 -0.76
16 -1.470 0.843 -1.24
17 -1.725 0.723 -1.25
18 -1.978 1.677 -3.32
19 -2.232 0.624 -1.39
20 -2.485 0.843 -2.09

03-33
Chapter 03 - Forecasting

Company B

Average from Seasonal Deseasonalized t*deseasonalized

t y same quarter factor demand t2 demand
1 0.17 0.263 0.764 0.223 1 0.223
2 0.24 0.370 1.077 0.223 4 0.446
3 0.26 0.343 0.999 0.260 9 0.781
4 0.34 0.417 1.213 0.280 16 1.121
5 0.25 0.764 0.327 25 1.636
6 0.37 1.077 0.344 36 2.061
7 0.36 0.999 0.360 49 2.522
8 0.44 1.213 0.363 64 2.902
9 0.33 0.764 0.432 81 3.887
10 0.40 1.077 0.371 100 3.714
11 0.41 0.999 0.410 121 4.513
12 0.47 1.213 0.388 144 4.651
13 0.30 0.764 0.393 169 5.104
14 0.47 1.077 0.436 196 6.110

Total 105 4.81 4.810 1015 39.672

y = 0.344
t = 7.5
∑ tyt −n t y t
0.016
b= ∑ t 2−nt̄ 2 =
a=
y t −bt = 0.225

Forecast
Seasonal (Yt*seasonal
Period (t) Yt factor factor)
15 0.462 0.999 0.46
16 0.478 1.213 0.58
17 0.494 0.764 0.38
18 0.510 1.077 0.55
19 0.525 0.999 0.53
20 0.541 1.213 0.66

There does appear to be seasonality in the data.

d. Using your forecasts, comment on each company.

The results indicate that Company A’s EPS is on a downward trend, while Company
B’s EPS is growing.

03-34
Chapter 03 - Forecasting

30. Mark Price, the new productions manager for Speakers and Company, needs to find out
which variable most affects the demand for their line of stereo speakers. He is uncertain
whether the unit price of the product or the effects of increased marketing are the main
drivers in sales and wants to use regression analysis to figure out which factor drives more
demand for its particular market. Pertinent information was collected by an extensive
marketing project that lasted over the past 12 years and was reduced to the data that
follow:
UNIT SALES PRICE PER ADVERTISING
YEAR (THOUSANDS) UNIT ($,000)
1 400 280 600
2 700 215 835
3 900 211 1100
4 1300 210 1400
5 1150 215 1200
6 1200 200 1300
7 900 225 900
8 1100 207 1100
9 980 220 700
10 1234 211 900
11 925 227 700
12 800 245 690

a. Perform a regression analysis based on these data using Excel. Answer the following
questions based on your results.

We can use the Regression tool or the LINEST() function within Excel. Answers using
LINEST function in Microsoft Excel follow.

Sales Price Advertising Fitted Values

400 280 600 451.72
700 215 835 977.21
900 211 1100 1090.98
1300 210 1400 1195.40
1150 215 1200 1095.85
1200 200 1300 1231.99
900 225 900 929.25
1100 207 1100 1118.62
980 220 700 898.79
1234 211 900 1025.98
925 227 700 850.42
800 245 690 722.80
2191.337
Constant: 4
Price: -6.9094
Advertising
: 0.3250

03-35
Chapter 03 - Forecasting

y= a+b1 x 1 +b2 x 2 =2191 .3374−6 . 9094 x 1 +. 3250 x 2

Where
a = y intercept
x1 = price
b1 = slope of price
x2 = advertising
b2 = slope of advertising

b. Which variable, price or advertising, has a larger effect on sales and how do you
know?

Price has a larger effect on sales because it slope value is much higher (-6.9094
versus .3250). Price actually has a negative effect since raising price decreases
sales.

c. Predict average yearly speaker sales for Speakers and Company based on the
regression results if the price was $300 per unit and the amount spent on
advertising (in thousands) was $900.

Sales = 2191.3374 - 6.9094 (300) + .3250 (900)

Sales = 411.04 thousand units

03-36
Chapter 03 - Forecasting

31. Sales by quarter for last year and the first three quarters of this year were as follows:

Using a procedure that you develop that captures the change in demand from last year to
this year and also the seasonality in demand, forecast expected sales for the fourth quarter
of this year.

Being left up to the student to develop a method, and given the limited amount of historical
data, there could be several good answers to this problem. An inspection of the data
makes it seem obvious that there is seasonality in demand. Also the year-to-year figures in
each quarter support an assumption of a negative long-term trend.

One simplistic solution would be to manually follow the patterns in the existing demand
data and forecast demand that is somewhat lower than fourth quarter last year, say 7,000.

Another more analytical approach would be to apply decomposition to the existing data,
but we do not have two full years’ worth of data, so computing a seasonal factor for
quarter IV is a concern. We could develop seasonal factors based solely on last year’s data,
or we could use a substitute data point to complete this year’s data. The following solution
uses last year’s quarter IV data as this year’s quarter IV demand to build the seasonal
factors. Regression is run on the deseasonalized demand for the first seven quarters.
Based on the results, this year’s quarter IV forecast is developed.

Deseasonalize
        Same Qtr Seasonal d   Seasonal
  Quarter t Demand Average Factor Demand Yt Forecast
Last Year I 1 23000 21000 1.167 19714
II 2 27000 25500 1.417 19059
III 3 18000 16500 0.917 19636
IV 4 9000 9000 0.500 18000
This Year I 5 19000 1.167 16286
II 6 24000 1.417 16941
III 7 15000 0.917 16364
  IV 8 9000   0.500   15480.3 7740
Averag
e 18000

Intercept = 20519.7 Found using INTERCEPT() function and deseasonalized data

Slope = -629.925 Found using SLOPE() function

03-37
Chapter 03 - Forecasting

32. The following are sales revenues for a large utility company for years 1 through 11.
Forecast revenue for years 12 through 15. Because we are forecasting four years into the
future, you will need to use linear regression as your forecasting method.

REVENUE REVENUE
YEAR (MILLIONS) YEAR (MILLIONS)
1 $4865.9 7 $5094.4
2 5067.4 8 5108.8
3 5515.6 9 5550.6
4 5728.8 10 5738.9
5 5497.7 11 5860.0
6 5197.7

6000
5800
5600
5400
5200
Revenue

5000
4800
4600
4400
4200
4000
1 2 3 4 5 6 7 8 9 10 11
Year

Examination of the graph of revenue over time suggests that there may be a slight upward
trend. Additionally, there may be a cyclical component, possibly 6 or 7 years. With the limited
data, it is very difficult to determine the cycle. Consequently, simple regression appears to be
the available choice for the forecast.

03-38
Chapter 03 - Forecasting

t y
1 4865.9
2 5067.4
3 5515.6
4 5728.8
5 5497.7
6 5197.7
7 5094.4
8 5108.8
9 5550.6
10 5738.9
11 5860.0

Total 66 59225.8

Using LINEST():
b= 55.62
a = 5050.444

Period Forecast
12 5718
13 5774
14 5829
15 5885

03-39
Chapter 03 - Forecasting

33. What forecasting technique makes use of written surveys or telephone interviews?

Marketing research

34. Which qualitative forecasting technique was developed to ensure the input from every
participant in the process is weighted equally?

Delphi method

35. When forecasting demand for new products, sometimes firms will use demand data from
similar existing products to help forecast demand for the new product. What technique is
this an example of?

Historical analogy

36. Often times, firms will work with their partners across the supply chain to develop
forecasts and execute production and distribution between the partners. What technique
does this describe?

Collaborative planning, forecasting, and replenishment (CPFR)

37. How many steps are there in collaborative planning, forecasting, and replenishment
(CPFR)?

Five

38. What is the first step in CPFR?

Creation of the front-end partnership agreement

03-40
Chapter 03 - Forecasting

ANALYTICS EXERCISE: Forecasting Supply Chain Demand – Starbucks Corporation

A good first step in developing a forecasting model is to create a plot of the historical demand data.

There does not appear to be an obvious trend in demand at any of the DCs, though there is quite a bit of
variation. With no evident trend, both moving average and exponential smoothing models would be
appropriate.

1, 2. Both the MA and ES models are shown on the following pages. A summary of their performance
measures is shown in the table on the next page (best measures in each column are bolded). Performance of
the two MA models is similar – neither one seems to be consistently better than the other. The same can be
said for the two ES models, though they do appear to perform slightly better than the MA models. The ES
model with an alpha of .2 appears to be the best.

3. Aggregating demand at a single DC will result in better overall forecasting performance than forecasting
for five DCs separately. The variations in demand for the separate DCs will tend to cancel each other out
somewhat when aggregating demand and forecasting for a single DC. In each performance measure, the
aggregated forecast error was about half that of the sum or average measure for the separate DCs.

Other factors to consider would be the transportation costs which might be higher serving from a central DC
rather than a regional one nearer the customers. We should also consider risk of natural disasters or other
causes that might disrupt transportation modes from the single DC and interrupt the supply chain for this
item.

03-41
Chapter 03 - Forecasting

Avg
ATL BOS CHI DAL LA of DCs
MAD 11.44 10.08 19.85 10.64 9.87 12.37
3-week
MAPE 0.288 0.243 0.420 0.223 0.214 0.278
MA
TS -0.117 1.224 -0.638 -1.222 1.992 N/A
MAD 11.17 10.63 18.45 12.37 9.58 12.44
5-week MA MAPE 0.281 0.256 0.391 0.260 0.207 0.279
TS -0.322 1.750 -0.011 -0.808 1.982 N/A
MAD 10.76 9.85 17.77 10.51 8.57 11.49
ES, MAPE 0.271 0.237 0.376 0.221 0.185 0.258
 = .2
TS -0.512 1.225 -1.63 -0.65 1.52 N/A
MAD 11.87 10.55 18.32 9.84 9.57 12.03
ES, MAPE 0.299 0.254 0.388 0.207 0.207 0.271
 = .4
TS -0.046 1.032 -0.241 0.024 1.510 N/A
Performance Measures of the Four Forecasting Models

03-42
Chapter 03 - Forecasting

Historical Demand 3-week MA

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
-5 45 62 62 42 43 254
-4 38 18 22 35 40 153
-3 30 48 72 40 54 244
-2 58 40 44 64 46 252
-1 37 35 48 43 35 198
1 33 26 44 27 32 162 41.7 41.0 54.7 49.0 45.0 231.3
2 45 35 34 42 43 199 42.7 33.7 45.3 44.7 37.7 204.0
3 37 41 22 35 54 189 38.3 32.0 42.0 37.3 36.7 186.3
4 38 40 55 40 40 213 38.3 34.0 33.3 34.7 43.0 183.3
5 55 46 48 51 46 246 40.0 38.7 37.0 39.0 45.7 200.3
6 30 48 72 64 74 288 43.3 42.3 41.7 42.0 46.7 216.0
7 18 55 62 70 40 245 41.0 44.7 58.3 51.7 53.3 249.0
8 58 18 28 65 35 204 34.3 49.7 60.7 61.7 53.3 259.7
9 47 62 27 55 45 236 35.3 40.3 54.0 66.3 49.7 245.7
10 37 44 95 43 38 257 41.0 45.0 39.0 63.3 40.0 228.3
11 23 30 35 38 48 174 47.3 41.3 50.0 54.3 39.3 232.3
12 55 45 45 47 56 248 35.7 45.3 52.3 45.3 43.7 222.3
13 40 50 47 42 50 229 38.3 39.7 58.3 42.7 47.3 226.3
13-week
39.69 41.54 47.23 47.62 46.23 222.31
Average:
Absolute Deviation RSFE

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
1 8.67 15.00 10.67 22.00 13.00 69.33 -8.7 -15.0 -10.7 -22.0 -13.0 -69.3
2 2.33 1.33 11.33 2.67 5.33 5.00 -6.3 -13.7 -22.0 -24.7 -7.7 -74.3
3 1.33 9.00 20.00 2.33 17.33 2.67 -7.7 -4.7 -42.0 -27.0 9.7 -71.7
4 0.33 6.00 21.67 5.33 3.00 29.67 -8.0 1.3 -20.3 -21.7 6.7 -42.0
5 15.00 7.33 11.00 12.00 0.33 45.67 7.0 8.7 -9.3 -9.7 7.0 3.7
6 13.33 5.67 30.33 22.00 27.33 72.00 -6.3 14.3 21.0 12.3 34.3 75.7
7 23.00 10.33 3.67 18.33 13.33 4.00 -29.3 24.7 24.7 30.7 21.0 71.7
8 23.67 31.67 32.67 3.33 18.33 55.67 -5.7 -7.0 -8.0 34.0 2.7 16.0
9 11.67 21.67 27.00 11.33 4.67 9.67 6.0 14.7 -35.0 22.7 -2.0 6.3
10 4.00 1.00 56.00 20.33 2.00 28.67 2.0 13.7 21.0 2.3 -4.0 35.0
11 24.33 11.33 15.00 16.33 8.67 58.33 -22.3 2.3 6.0 -14.0 4.7 -23.3
12 19.33 0.33 7.33 1.67 12.33 25.67 -3.0 2.0 -1.3 -12.3 17.0 2.3
13 1.67 10.33 11.33 0.67 2.67 2.67 -1.3 12.3 -12.7 -13.0 19.7 5.0
MAD: 11.44 10.08 19.85 10.64 9.87 31.46
Sum of DC MAD: 61.87
MAPE: 0.288 0.243 0.420 0.223 0.214 0.142
Average DC MAPE: 0.278
TS: -0.117 1.224 -0.638 -1.222 1.992 0.159

03-43
Chapter 03 - Forecasting

Historical Demand 5-week MA

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
-5 45 62 62 42 43 254
-4 38 18 22 35 40 153
-3 30 48 72 40 54 244
-2 58 40 44 64 46 252
-1 37 35 48 43 35 198
1 33 26 44 27 32 162 41.6 40.6 49.6 44.8 43.6 220.2
2 45 35 34 42 43 199 39.2 33.4 46.0 41.8 41.4 201.8
3 37 41 22 35 54 189 40.6 36.8 48.4 43.2 42.0 211.0
4 38 40 55 40 40 213 42.0 35.4 38.4 42.2 42.0 200.0
5 55 46 48 51 46 246 38.0 35.4 40.6 37.4 40.8 192.2
6 30 48 72 64 74 288 41.6 37.6 40.6 39.0 43.0 201.8
7 18 55 62 70 40 245 41.0 42.0 46.2 46.4 51.4 227.0
8 58 18 28 65 35 204 35.6 46.0 51.8 52.0 50.8 236.2
9 47 62 27 55 45 236 39.8 41.4 53.0 58.0 47.0 239.2
10 37 44 95 43 38 257 41.6 45.8 47.4 61.0 48.0 243.8
11 23 30 35 38 48 174 38.0 45.4 56.8 59.4 46.4 246.0
12 55 45 45 47 56 248 36.6 41.8 49.4 54.2 41.2 223.2
13 40 50 47 42 50 229 44.0 39.8 46.0 49.6 44.4 223.8
13-week
39.69 41.54 47.23 47.62 46.23 222.31
Average:
Absolute Deviation RSFE

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
1 8.60 14.60 5.60 17.80 11.60 58.20 -8.6 -14.6 -5.6 -17.8 -11.6 -58.2
2 5.80 1.60 12.00 0.20 1.60 2.80 -2.8 -13.0 -17.6 -17.6 -10.0 -61.0
3 3.60 4.20 26.40 8.20 12.00 22.00 -6.4 -8.8 -44.0 -25.8 2.0 -83.0
4 4.00 4.60 16.60 2.20 2.00 13.00 -10.4 -4.2 -27.4 -28.0 0.0 -70.0
5 17.00 10.60 7.40 13.60 5.20 53.80 6.6 6.4 -20.0 -14.4 5.2 -16.2
6 11.60 10.40 31.40 25.00 31.00 86.20 -5.0 16.8 11.4 10.6 36.2 70.0
7 23.00 13.00 15.80 23.60 11.40 18.00 -28.0 29.8 27.2 34.2 24.8 88.0
8 22.40 28.00 23.80 13.00 15.80 32.20 -5.6 1.8 3.4 47.2 9.0 55.8
9 7.20 20.60 26.00 3.00 2.00 3.20 1.6 22.4 -22.6 44.2 7.0 52.6
10 4.60 1.80 47.60 18.00 10.00 13.20 -3.0 20.6 25.0 26.2 -3.0 65.8
11 15.00 15.40 21.80 21.40 1.60 72.00 -18.0 5.2 3.2 4.8 -1.4 -6.2
12 18.40 3.20 4.40 7.20 14.80 24.80 0.4 8.4 -1.2 -2.4 13.4 18.6
13 4.00 10.20 1.00 7.60 5.60 5.20 -3.6 18.6 -0.2 -10.0 19.0 23.8
MAD: 11.17 10.63 18.45 12.37 9.58 31.12
Sum of DC MAD: 62.20
MAPE: 0.281 0.256 0.391 0.260 0.207 0.140
Average DC MAPE: 0.278
TS: -0.322 1.750 -0.011 -0.808 1.982 0.765

03-44
Chapter 03 - Forecasting

Historical Demand ES,  = 2

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
-5 45 62 62 42 43 254
-4 38 18 22 35 40 153
-3 30 48 72 40 54 244
-2 58 40 44 64 46 252
-1 37 35 48 43 35 198
1 33 26 44 27 32 162 41.7 41.0 54.7 49.0 45.0 231.3
2 45 35 34 42 43 199 39.9 38.0 52.5 44.6 42.4 217.5
3 37 41 22 35 54 189 40.9 37.4 48.8 44.1 42.5 213.8
4 38 40 55 40 40 213 40.2 38.1 43.5 42.3 44.8 208.8
5 55 46 48 51 46 246 39.7 38.5 45.8 41.8 43.9 209.7
6 30 48 72 64 74 288 42.8 40.0 46.2 43.6 44.3 216.9
7 18 55 62 70 40 245 40.2 41.6 51.4 47.7 50.2 231.1
8 58 18 28 65 35 204 35.8 44.3 53.5 52.2 48.2 233.9
9 47 62 27 55 45 236 40.2 39.0 48.4 54.7 45.5 227.9
10 37 44 95 43 38 257 41.6 43.6 44.1 54.8 45.4 229.5
11 23 30 35 38 48 174 40.7 43.7 54.3 52.4 43.9 235.0
12 55 45 45 47 56 248 37.1 41.0 50.4 49.5 44.8 222.8
13 40 50 47 42 50 229 40.7 41.8 49.3 49.0 47.0 227.9
13-week
39.69 41.54 47.23 47.62 46.23 222.31
Average:
Absolute Deviation RSFE

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
1 8.67 15.00
10.67 22.00 13.00 69.33 -8.7 -15.0 -10.7 -22.0 -13.0 -69.3
2 5.07 3.00
18.53 2.60 0.60 18.47 -3.6 -18.0 -29.2 -24.6 -12.4 -87.8
3 3.95 3.60
26.83 9.08 11.48 24.77 -7.5 -14.4 -56.0 -33.7 -0.9 -112.6
4 2.16 1.88
11.54 2.26 4.82 4.18 -9.7 -12.5 -44.5 -35.9 -5.7 -108.4
5 15.27 7.502.23 9.19 2.15 36.35 5.6 -5.0 -42.3 -26.8 -3.6 -72.0
6 12.78 8.00
25.78 20.35 29.72 71.08 -7.2 3.0 -16.5 -6.4 26.1 -1.0
7 22.22 13.40
10.63 22.28 10.23 13.86 -29.4 16.4 -5.8 15.9 15.9 12.9
8 22.22 26.28
25.50 12.82 13.18 29.91 -7.2 -9.9 -31.3 28.7 2.7 -17.0
9 6.78 22.98
21.40 0.26 0.54 8.07 -0.4 13.1 -52.7 29.0 2.2 -9.0
10 4.58 0.38
50.88 11.79 7.44 27.46 -5.0 13.5 -1.9 17.2 -5.3 18.5
11 17.66 13.69
19.29 14.43 4.05 61.03 -22.7 -0.2 -21.2 2.7 -1.2 -42.5
12 17.87 4.045.44 2.55 11.24 25.17 -4.8 3.8 -26.6 0.2 10.0 -17.4
13 0.70 8.242.35 7.04 2.99 1.14 -5.5 12.1 -28.9 -6.8 13.0 -16.2
MAD: 10.76 9.85
17.77 10.51 8.57 30.06
Sum of DC MAD: 57.47
MAPE: 0.271 0.237 0.376 0.221 0.185 0.135
Average DC MAPE: 0.258
TS: -0.512 1.225 -1.628 -0.652 1.520 -0.539

03-45
Chapter 03 - Forecasting

Historical Demand ES,  = 4

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
-5 45 62 62 42 43 254
-4 38 18 22 35 40 153
-3 30 48 72 40 54 244
-2 58 40 44 64 46 252
-1 37 35 48 43 35 198
1 33 26 44 27 32 162 41.6 40.6 49.6 44.8 43.6 220.2
2 45 35 34 42 43 199 38.2 34.8 47.4 37.7 39.0 196.9
3 37 41 22 35 54 189 40.9 34.9 42.0 39.4 40.6 197.8
4 38 40 55 40 40 213 39.3 37.3 34.0 37.6 45.9 194.3
5 55 46 48 51 46 246 38.8 38.4 42.4 38.6 43.6 201.8
6 30 48 72 64 74 288 45.3 41.4 44.6 43.6 44.5 219.5
7 18 55 62 70 40 245 39.2 44.1 55.6 51.7 56.3 246.9
8 58 18 28 65 35 204 30.7 48.4 58.2 59.0 49.8 246.1
9 47 62 27 55 45 236 41.6 36.3 46.1 61.4 43.9 229.3
10 37 44 95 43 38 257 43.8 46.6 38.5 58.9 44.3 232.0
11 23 30 35 38 48 174 41.1 45.5 61.1 52.5 41.8 242.0
12 55 45 45 47 56 248 33.8 39.3 50.6 46.7 44.3 214.8
13 40 50 47 42 50 229 42.3 41.6 48.4 46.8 49.0 228.1
13-week
39.69 41.54 47.23 47.62 46.23 222.31
Average:
Absolute Deviation RSFE

Week ATL BOS CHI DAL LA Total ATL BOS CHI DAL LA Total
1 8.60 14.605.60 17.80 11.60 58.20 -8.6 -14.6 -5.6 -17.8 -11.6 -58.2
2 6.84 0.24
13.36 4.32 4.04 2.08 -1.8 -14.4 -19.0 -13.5 -7.6 -56.1
3 3.90 6.14
20.02 4.41 13.42 8.75 -5.7 -8.2 -39.0 -17.9 5.9 -64.9
4 1.34 2.69
20.99 2.36 5.95 18.75 -7.0 -5.5 -18.0 -15.5 -0.1 -46.1
5 16.20 7.615.59 12.41 2.43 44.25 9.2 2.1 -12.4 -3.1 2.4 -1.9
6 15.28 6.57
27.36 20.45 29.46 68.55 -6.1 8.6 15.0 17.3 31.8 66.7
7 21.17 10.946.41 18.27 16.32 1.87 -27.2 19.6 21.4 35.6 15.5 64.8
8 27.30 30.44
30.15 5.96 14.79 42.12 0.1 -10.8 -8.8 41.6 0.7 22.7
9 5.38 25.74
19.09 6.42 1.12 6.73 5.4 14.9 -27.9 35.1 1.8 29.4
10 6.77 2.56
56.55 15.85 6.33 25.04 -1.3 12.3 28.7 19.3 -4.5 54.4
11 18.06 15.53
26.07 14.51 6.20 67.98 -19.4 -3.2 2.6 4.8 1.7 -13.5
12 21.16 5.685.64 0.29 11.72 33.21 1.8 2.5 -3.0 5.1 13.4 19.7
13 2.30 8.411.39 4.82 1.03 0.93 -0.5 10.9 -4.4 0.2 14.4 20.6
MAD: 11.87 10.55
18.32 9.84 9.57 29.11
Sum of DC MAD: 60.15
MAPE: 0.299 0.254 0.388 0.207 0.207 0.131
Average DC MAPE: 0.271
TS: -0.046 1.032 -0.241 0.024 1.510 0.708

03-46