146 PA RT 1 | I N T R O D UC T I O N T O OP ERATI ONS MANAGEMENT

SOLVED PROBLEM 4.7 SALES, y ADVERTISING, x x2 xy

Cengiz Haksever runs an Istanbul high-end jewelry shop. He
11 5 25 55
advertises weekly in local Turkish newspapers and is thinking
6 3 9 18
of increasing his ad budget. Before doing so, he decides to eval-
uate the past effectiveness of these ads. Five weeks are sampled, 10 7 49 70
and the data are shown in the table below: 6 2 4 12
12 8 64 96
AD BUDGET g y = 45 g x = 25 g x 2 = 151 g xy = 251
SALES THAT WEEK
($1,000s) ($100s) 45 25
y = = 9 x = = 5
5 5
11 5
6 3
gxy - nx y 251 - (5)(5)(9)
10 7 b = =
gx - nx
2 2
151 - (5)(52)
6 2
12 8 251 - 225 26
= = = 1
151 - 125 26
Develop a regression model to help Cengiz evaluate his
a = y - bx = 9 - (1)(5) = 4
advertising.
So the regression model is yn = 4 + 1x, or
SOLUTION Sales (in $1,000s) = 4 + 1 (Ad budget in $100s)
We apply the least-squares regression model as we did in This means that for each 1-unit increase in x (or $100 in ads),
Example 12. sales increase by 1 unit (or $1,000).

SOLVED PROBLEM 4.8 ngxy - gx gy

Using the data in Solved Problem 4.7, find the coefficient of r =
2[ngx - ( gx)2][ngy2 - ( gy)2]
2
determination, r2, for the model.
5(251) - (25)(45)
SOLUTION =
2[5(151) - (25)2][5(437) - (45)2]
To find r2, we need to also compute gy2.
1,255 - 1,125 130 130
gy2 = 112 + 62 + 102 + 62 + 122 = = =
2(130)(160) 220,800 144.22
= 121 + 36 + 100 + 36 + 144 = 437
= .9014
The next step is to find the coefficient of correlation, r:
Thus, r2 = (.9014)2 = .8125, meaning that about 81% of the
variability in sales can be explained by the regression model
with advertising as the independent variable.

Problems Note: PX means the problem may be solved with POM for Windows and/or Excel OM.

Problems 4.1–4.42 relate to Time-Series Forecasting b) Use a 3-week weighted moving average, with weights of .1, .3,
and .6, using .6 for the most recent week. Forecast demand for
• 4.1 The following gives the number of pints of type B the week of October 12.
blood used at Woodlawn Hospital in the past 6 weeks: c) Compute the forecast for the week of October 12 using exponential
WEEK OF PINTS USED smoothing with a forecast for August 31 of 360 and a = .2. PX
August 31 360 • • 4.2
September 7 389 YEAR 1 2 3 4 5 6 7 8 9 10 11
September 14 410 DEMAND 7 9 5 9 13 8 12 13 9 11 7
September 21 381
September 28 368 a) Plot the above data on a graph. Do you observe any trend,
cycles, or random variations?
October 5 374
b) Starting in year 4 and going to year 12, forecast demand using
a) Forecast the demand for the week of October 12 using a a 3-year moving average. Plot your forecast on the same graph
3-week moving average. as the original data.

M04_HEIZ0422_12_SE_C04.indd 146 14/12/15 9:53 am

 CHAP T ER 4 | FORE CAST ING 147

c) Starting in year 4 and going to year 12, forecast demand using a) Plot the monthly sales data.
a 3-year moving average with weights of .1, .3, and .6, using .6 b) Forecast January sales using each of the following:
for the most recent year. Plot this forecast on the same graph. i) Naive method.
d) As you compare forecasts with the original data, which seems ii) A 3-month moving average.
to give the better results? PX iii) A 6-month weighted average using .1, .1, .1, .2, .2, and .3,
with the heaviest weights applied to the most recent months.
• • 4.3 Refer to Problem 4.2. Develop a forecast for years 2
iv) Exponential smoothing using an a = .3 and a September
through 12 using exponential smoothing with a = .4 and a fore-
forecast of 18.
cast for year 1 of 6. Plot your new forecast on a graph with the
v) A trend projection.
actual data and the naive forecast. Based on a visual inspection,
c) With the data given, which method would allow you to fore-
which forecast is better? PX
cast next March’s sales? PX
• 4.4 A check-processing center uses exponential smooth-
• • 4.7 The actual demand for the patients at Omaha
ing to forecast the number of incoming checks each month. The
Emergency Medical Clinic for the first 6 weeks of this year fol-
number of checks received in June was 40 million, while the fore-
lows:
cast was 42 million. A smoothing constant of .2 is used.
a) What is the forecast for July? ACTUAL NO. OF
b) If the center received 45 million checks in July, what would be WEEK PATIENTS
the forecast for August? 1 65
c) Why might this be an inappropriate forecasting method for
this situation? PX 2 62
3 70
• • 4.5 The Carbondale Hospital is considering the purchase
of a new ambulance. The decision will rest partly on the antici- 4 48
pated mileage to be driven next year. The miles driven during the 5 63
past 5 years are as follows: 6 52
YEAR MILEAGE
1 3,000
Clinic administrator Marc Schniederjans wants you to forecast
patient demand at the clinic for week 7 by using this data. You
2 4,000
decide to use a weighted moving average method to find this fore-
3 3,400 cast. Your method uses four actual demand levels, with weights
4 3,800 of 0.333 on the present period, 0.25 one period ago, 0.25 two peri-
5 3,700 ods ago, and 0.167 three periods ago.
a) What is the value of your forecast? PX
a) Forecast the mileage for next year (6th year) using a 2-year b) If instead the weights were 20, 15, 15, and 10, respectively, how
moving average. would the forecast change? Explain why.
b) Find the MAD based on the 2-year moving average. (Hint: c) What if the weights were 0.40, 0.30, 0.20, and 0.10, respec-
You will have only 3 years of matched data.) tively? Now what is the forecast for week 7?
c) Use a weighted 2-year moving average with weights of .4
• 4.8 Daily high temperatures in St. Louis for the last week
and .6 to forecast next year’s mileage. (The weight of .6 is
were as follows: 93, 94, 93, 95, 96, 88, 90 (yesterday).
for the most recent year.) What MAD results from using this
a) Forecast the high temperature today, using a 3-day moving
approach to forecasting? (Hint: You will have only 3 years of
average.
matched data.)
b) Forecast the high temperature today, using a 2-day moving
d) Compute the forecast for year 6 using exponential smoothing,
average.
an initial forecast for year 1 of 3,000 miles, and a = .5. PX
c) Calculate the mean absolute deviation based on a 2-day mov-
• • 4.6 The monthly sales for Yazici Batteries, Inc., were as ing average.
follows: d) Compute the mean squared error for the 2-day moving average.
e) Calculate the mean absolute percent error for the 2-day mov-
MONTH SALES ing average. PX
January 20
• • • 4.9 Lenovo uses the ZX-81 chip in some of its laptop
February 21 computers. The prices for the chip during the past 12 months
March 15 were as follows:
April 14
May 13 PRICE PER PRICE PER
MONTH CHIP MONTH CHIP
June 16
January $1.80 July 1.80
July 17
February 1.67 August 1.83
August 18
September 20 March 1.70 September 1.70
October 20 April 1.85 October 1.65
November 21 May 1.90 November 1.70
December 23 June 1.87 December 1.75

M04_HEIZ0422_12_SE_C04.indd 147 14/12/15 9:53 am

 148 PA RT 1 | I N T R O D UC T I O N T O OP ERATI ONS MANAGEMENT

a) Use a 2-month moving average on all the data and plot the
averages and the prices.
b) Use a 3-month moving average and add the 3-month plot to
the graph created in part (a).
c) Which is better (using the mean absolute deviation): the
2-month average or the 3-month average?
d) Compute the forecasts for each month using exponential
smoothing, with an initial forecast for January of $1.80. Use a
= .1, then a = .3, and finally a = .5. Using MAD, which a

Dmitry Kalinovsky/Shutterstock
is the best? PX
• • 4.10 Data collected on the yearly registrations for a Six
Sigma seminar at the Quality College are shown in the following
table:

YEAR 1 2 3 4 5 6 7 8 9 10 11
REGISTRATIONS (000) 4 6 4 5 10 8 7 9 12 14 15

a) Develop a 3-year moving average to forecast registrations a) Use exponential smoothing, first with a smoothing constant
from year 4 to year 12. of .6 and then with one of .9, to develop forecasts for years 2
b) Estimate demand again for years 4 to 12 with a 3-year weighted through 6.
moving average in which registrations in the most recent year b) Use a 3-year moving average to forecast demand in years 4, 5,
are given a weight of 2, and registrations in the other 2 years and 6.
are each given a weight of 1. c) Use the trend-projection method to forecast demand in years 1
c) Graph the original data and the two forecasts. Which of the through 6.
two forecasting methods seems better? PX d) With MAD as the criterion, which of the four forecasting
methods is best? PX
• 4.11 Use exponential smoothing with a smoothing con-
stant of 0.3 to forecast the registrations at the seminar given in • • 4.14 Following are two weekly forecasts made by two dif-
Problem 4.10. To begin the procedure, assume that the forecast ferent methods for the number of gallons of gasoline, in thou-
for year 1 was 5,000 people signing up. sands, demanded at a local gasoline station. Also shown are
a) What is the MAD? PX actual demand levels, in thousands of gallons.
b) What is the MSE?
FORECASTS
• • 4.12 Consider the following actual and forecast WEEK METHOD 1 METHOD 2 ACTUAL DEMAND
demand levels for Big Mac hamburgers at a local McDonald’s
1 0.90 0.80 0.70
restaurant:
2 1.05 1.20 1.00
DAY ACTUAL DEMAND FORECAST DEMAND 3 0.95 0.90 1.00
Monday 88 88 4 1.20 1.11 1.00
Tuesday 72 88
What are the MAD and MSE for each method?
Wednesday 68 84
Thursday 48 80
• 4.15 Refer to Solved Problem 4.1 on page 144.
a) Use a 3-year moving average to forecast the sales of
Friday Volkswagen Beetles in Nevada through year 6.
b) What is the MAD? PX
The forecast for Monday was derived by observing Monday’s c) What is the MSE?
demand level and setting Monday’s forecast level equal to this • 4.16 Refer to Solved Problem 4.1 on page 144.
demand level. Subsequent forecasts were derived by using expo- a) Using the trend projection (regression) method, develop a
nential smoothing with a smoothing constant of 0.25. Using this forecast for the sales of Volkswagen Beetles in Nevada through
exponential smoothing method, what is the forecast for Big Mac year 6.
demand for Friday? PX b) What is the MAD? PX
• • • 4.13 As you can see in the following table, demand for c) What is the MSE?
heart transplant surgery at Washington General Hospital has • 4.17 Refer to Solved Problem 4.1 on page 144. Using
increased steadily in the past few years: smoothing constants of .6 and .9, develop forecasts for the sales
of VW Beetles. What effect did the smoothing constant have on
YEAR 1 2 3 4 5 6 the forecast? Use MAD to determine which of the three smooth-
HEART TRANSPLANTS 45 50 52 56 58 ? ing constants (.3, .6, or .9) gives the most accurate forecast. PX
• • • • 4.18 Consider the following actual (At) and forecast (Ft)
The director of medical services predicted 6 years ago that demand levels for a commercial multiline telephone at Office
demand in year 1 would be 41 surgeries. Max:

M04_HEIZ0422_12_SE_C04.indd 148 14/12/15 9:53 am

 CHAP T ER 4 | FORE CAST ING 149

TIME ACTUAL FORECAST a) Compute MAD and MAPE for management’s technique.
PERIOD, t DEMAND, At DEMAND, Ft b) Do management’s results outperform (i.e., have smaller MAD
1 50 50 and MAPE than) a naive forecast?
2 42 50 c) Which forecast do you recommend, based on lower forecast
3 56 48
error? PX
4 46 50 • 4.24 The following gives the number of accidents
5 that occurred on Florida State Highway 101 during the past 4
months:
The first forecast, F1, was derived by observing A1 and setting F1
equal to A1. Subsequent forecast averages were derived by expo- MONTH NUMBER OF ACCIDENTS
nential smoothing. Using the exponential smoothing method, January 30
find the forecast for time period 5. (Hint: You need to first find February 40
the smoothing constant, a.)
March 60
• • • 4.19 Income at the architectural firm Spraggins and Yunes
April 90
for the period February to July was as follows:
Forecast the number of accidents that will occur in May, using
MONTH FEBRUARY MARCH APRIL MAY JUNE JULY
least-squares regression to derive a trend equation.PX
Income
70.0 68.5 64.8 71.7 71.3 72.8 • 4.25 In the past, Peter Kelle’s tire dealership in Baton
(in $ thousand)
Rouge sold an average of 1,000 radials each year. In the past
Use trend-adjusted exponential smoothing to forecast the firm’s 2 years, 200 and 250, respectively, were sold in fall, 350 and 300
August income. Assume that the initial forecast average for in winter, 150 and 165 in spring, and 300 and 285 in summer.
February is $65,000 and the initial trend adjustment is 0. The With a major expansion planned, Kelle projects sales next year
smoothing constants selected are a = .1 and b = .2. PX to increase to 1,200 radials. What will be the demand during each
season?
• • • 4.20 Resolve Problem 4.19 with a = .1 and b = .8. Using
MSE, determine which smoothing constants provide a better • • 4.26 George Kyparisis owns a company that manufactures
forecast. PX sailboats. Actual demand for George’s sailboats during each of
the past four seasons was as follows:
• 4.21 Refer to the trend-adjusted exponential smoothing
illustration in Example 7 on pages 122–123. Using a = .2 and YEAR
b = .4, we forecast sales for 9 months, showing the detailed cal-
SEASON 1 2 3 4
culations for months 2 and 3. In Solved Problem 4.2, we contin-
ued the process for month 4. Winter 1,400 1,200 1,000 900
In this problem, show your calculations for months 5 and 6 for Spring 1,500 1,400 1,600 1,500
Ft, Tt, and FITt. PX Summer 1,000 2,100 2,000 1,900
• 4.22 Refer to Problem 4.21. Complete the trend-adjusted Fall 600 750 650 500
exponential-smoothing forecast computations for periods 7, 8,
and 9. Confirm that your numbers for Ft, Tt, and FITt match
George has forecasted that annual demand for his sailboats
those in Table 4.2 (p. 123). PX
in year 5 will equal 5,600 sailboats. Based on this data and the
• • 4.23 Sales of quilt covers at Bud Banis’s department store multiplicative seasonal model, what will the demand level be for
in Carbondale over the past year are shown below. Management George’s sailboats in the spring of year 5?
prepared a forecast using a combination of exponential smooth- • • 4.27 Attendance at Orlando’s newest Disneylike attrac-
ing and its collective judgment for the 4 months (March, April, tion, Lego World, has been as follows:
May, and June):
GUESTS (IN GUESTS
MONTH UNIT SALES MANAGEMENT’S FORECAST QUARTER THOUSANDS) QUARTER (IN THOUSANDS)
July 100 Winter Year 1 73 Summer Year 2 124
August 93
Spring Year 1 104 Fall Year 2 52
September 96
Summer Year 1 168 Winter Year 3 89
October 110
November 124 Fall Year 1 74 Spring Year 3 146
December 119 Winter Year 2 65 Summer Year 3 205
January 92 Spring Year 2 82 Fall Year 3 98
February 83
March 101 120 Compute seasonal indices using all of the data. PX
April 96 114
• 4.28 North Dakota Electric Company estimates its demand
May 89 110 trend line (in millions of kilowatt hours) to be:
June 108 108
D = 77 + 0.43Q

M04_HEIZ0422_12_SE_C04.indd 149 14/12/15 9:54 am

 150 PA RT 1 | I N T R O D UC T I O N T O OP ERATI ONS MANAGEMENT

where Q refers to the sequential quarter number and Q 5 1 for • • • 4.32 Using the 911 call data in Problem 4.31, forecast
winter of Year 1. In addition, the multiplicative seasonal factors calls for weeks 2 through 25 with a trend-adjusted exponential
are as follows: smoothing model. Assume an initial forecast for 50 calls for week
1 and an initial trend of zero. Use smoothing constants of a = .3
QUARTER FACTOR (INDEX) and b = .2. Is this model better than that of Problem 4.31? What
Winter .8 adjustment might be useful for further improvement? (Again,
Spring 1.1 assume that actual calls in week 25 were 85.) PX
Summer 1.4 • • • 4.33 Storrs Cycles has just started selling the new Cyclone
mountain bike, with monthly sales as shown in the table. First,
Fall .7
co-owner Bob Day wants to forecast by exponential smoothing
Forecast energy use for the four quarters of year 26 (namely quar- by initially setting February’s forecast equal to January’s sales
ters 101 to 104), beginning with winter. with a = .1. Co-owner Sherry Snyder wants to use a three-period
moving average.
• 4.29 The number of disk drives (in millions) made at a
plant in Taiwan during the past 5 years follows:
BOB’S SHERRY’S
SALES BOB SHERRY ERROR ERROR
YEAR DISK DRIVES
1 140
JANUARY 400 —

2 160 FEBRUARY 380 400

3 190 MARCH 410
4 200 APRIL 375
5 210 MAY

a) Forecast the number of disk drives to be made next year, using

a) Is there a strong linear trend in sales over time?
linear regression.
b) Fill in the table with what Bob and Sherry each forecast for
b) Compute the mean squared error (MSE) when using linear
May and the earlier months, as relevant.
regression.
c) Assume that May’s actual sales figure turns out to be 405.
c) Compute the mean absolute percent error (MAPE). PX
Complete the table’s columns and then calculate the mean
• • 4.30 Dr. Lillian Fok, a New Orleans psychologist, spe- absolute deviation for both Bob’s and Sherry’s methods.
cializes in treating patients who are agoraphobic (i.e., afraid d) Based on these calculations, which method seems more
to leave their homes). The following table indicates how many accurate? PX
patients Dr. Fok has seen each year for the past 10 years. It also
• • • • 4.34 Boulanger Savings and Loan is proud of its long tra-
indicates what the robbery rate was in New Orleans during the
dition in Winter Park, Florida. Begun by Michelle Boulanger
same year:
22 years after World War II, the S&L has bucked the trend of
YEAR 1 2 3 4 5 6 7 8 9 10 financial and liquidity problems that has repeatedly plagued
the industry. Deposits have increased slowly but surely over the
NUMBER
OF PATIENTS
36 33 40 41 40 55 60 54 58 61 years, despite recessions in 1983, 1988, 1991, 2001, and 2010. Ms.
Boulanger believes it is necessary to have a long-range strategic
ROBBERY RATE
PER 1,000 58.3 61.1 73.4 75.7 81.1 89.0 101.1 94.8 103.3 116.2 plan for her firm, including a 1-year forecast and preferably even
POPULATION a 5-year forecast of deposits. She examines the past deposit data
and also peruses Florida’s gross state product (GSP) over the same
Using trend (linear regression) analysis, predict the number of 44 years. (GSP is analogous to gross national product [GNP] but
patients Dr. Fok will see in years 11 and 12 as a function of time. on the state level.) The resulting data are in the following table.
How well does the model fit the data? PX
YEAR DEPOSITSa GSPb YEAR DEPOSITSa GSPb
• • • 4.31 Emergency calls to the 911 system of Durham,
North Carolina, for the past 24 weeks are shown in the following 1 .25 .4 13 .50 1.2
table: 2 .24 .4 14 .95 1.2
3 .24 .5 15 1.70 1.2
WEEK 1 2 3 4 5 6 7 8 9 10 11 12
4 .26 .7 16 2.3 1.6
CALLS 50 35 25 40 45 35 20 30 35 20 15 40
5 .25 .9 17 2.8 1.5
WEEK 13 14 15 16 17 18 19 20 21 22 23 24
6 .30 1.0 18 2.8 1.6
CALLS 55 35 25 55 55 40 35 60 75 50 40 65
7 .31 1.4 19 2.7 1.7
a) Compute the exponentially smoothed forecast of calls for each 8 .32 1.7 20 3.9 1.9
week. Assume an initial forecast of 50 calls in the first week, 9 .24 1.3 21 4.9 1.9
and use a = .2. What is the forecast for week 25? 10 .26 1.2 22 5.3 2.3
b) Reforecast each period using a = .6.
11 .25 1.1 23 6.2 2.5
c) Actual calls during week 25 were 85. Which smoothing con-
stant provides a superior forecast? Explain and justify the 12 .33 .9 24 4.1 2.8
measure of error you used. PX (continued)

M04_HEIZ0422_12_SE_C04.indd 150 14/12/15 9:54 am