ISQA 459/559

Problem Set 1
1. FORECASTING COMPARISON
Use the spreadsheet program (template provided on Mellie Pullmans website) to
compare forecasting using MA, WMA, exponential smoothing, and linear regression. A
demand pattern for 10 periods for a certain product is provided below. Forecast the
demand for period 11 using each of the following methods: a 3-month moving average; a
3-month weighted moving average using weights .2, .3., and .5; exponential smoothing
with a smoothing constant or = .3; and linear regression.
Note: Using the 3 month moving average forecast for week 4 as your starting point Ft-1
for exponential smoothing.
a) Compute the MAD value for each technique (using the forecasts for periods 4-10) to
determine which method would be preferable under the circumstances. What model
gives the best MAD?
b) Plot the forecast data for the four methods relative to the actual demand. Explain what
you see there.
Please use Excel Functions such as AVERAGE, INTERCEPT, SLOPE, Chart making
(scatter Plot) with all 4 methods shown on one chart.
2. COMPLEX FORECAST
The goal of this exercise is to forecast demand for week 31. You will find the historic
data for the project on the second worksheet of file Problem Set 1 Template.xls on the
Course website. You will also find some supporting powerpoints to locate the appropriate
functions in Excel 07.
You have 30 weeks of historical data in the form of unit sales for each week.
1) Determine the appropriate Alpha () using exponential smoothing. You should
provide your best Alpha in 1 decimal place. Extra credit for 2-decimal places.
a. Calculate the error using MAD. Use the smoothing approach with phi ()
=.15 to calculate the different error measurements.
2) Please use the first weeks actual demand for your initial forecast (F1).
3) Go through Part 1 through Part 4 of the Excel tools (next pages). This will cover
Simple Exponential Smoothing, Forecasts with Different Alpha Values,
Developing a Forecast Table, and Best Forecast Automatically.

NOTE: If your spreadsheet is not calculating the performance table, check under
Tools/Options/Calculations to see if it is set to automatically calculate.
For part b) on the new worksheet, go through parts 5 through 7 (next pages). This will
cover Exponential Smoothing with Trend Forecasting.
3. SEASONAL DECOMPOSITION
Management of the Pearl Floral Shop feels that its sales are seasonal in nature with a
monthly seasonally pattern and a trend. The demand data for the past three years are
given in the third worksheet of the Problem Set 1 spreadsheet.
1) Calculate the Seasonal Ratio for each of the 3 years, the average seasonal ratio,
and deseasonalize the data (divide the data by its appropriate average seasonal
ratio.
2) Determine the slope and intercept of the deseasonalized data and forecast for the
next year (12 months).
3) Reseasonalize the forecast data (multiply by the appropriate average seasonal
ratio).

Using Excel for Forecasting

Part 1: Simple Exponential Smoothing
Lets first develop simple exponential smoothing forecasts for a set of
sales data. Begin by entering the label Alpha in cell a1 and the
value 0.3 in cell b1. Next, name cell b1 Alpha (highlight cells a1:b1,
select the Insert, Name, Create, Left Column menu choices). Then
enter the remainder of the following data:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

A
Alpha

B
0.3

C
#Periods

Period Demand Forecast

1
=b3
1891
2
=alpha*b3
+(1alpha)*c3
1693
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

D
30
Round

E
F
Best
=offset(f2
Values ,$d$1,0)
:
Error
MADt

=round( =d4c4,0)
b4

=abs(e4)
=.15*abs(
e5)+.85*f
4

1696
1632
1596
1565
1565
1653
1461
1603
1768
1817
1749
1678
1773
1774
1705
1690
1814
1757
1802
1666
1855
1662
1675
1886
1688
1622
1759
1944

A. Copy the formula in cell C4 into cells C5-C33.

B. Copy the formula in cell D4 into cells D5-D33.
C. Copy the formula in cell E4 into cells E5-E32.

D. Copy the formula in cell F5 into cells F6-F32.

Our basic model is now complete. To forecast using a different value
of alpha, simply put the desired value of alpha in cell B1.

Using Excel for Forecasting

Part 2: Developing Forecasts with Different
Alpha Values
Part 2 assumes that you have completed Part 1 and have a spreadsheet set
up as described in Part 1. In Part 2, we see how we can use Excels Table
function to quickly create forecasts (using different alpha values) to graph
simultaneously.

1. Enter the data as shown below.

J

M
N
O
P
1
Forecasts
2 Period Demand Alpha=. Alpha=.5 Alpha=.
1
9
3 =A3
=B3
.1
4
=D4

2.
3.
4.
5.

.5

.9

Copy the formula in cell J3 into cells J4 to J32.

Copy the formula in cell K3 into cells K4 to K32.
Copy the formula in cell O4 into cells O5 to O33.
Highlight cells O3:R33. Select the Data and Table menu choices. Under
the Row Input Cell, enter either B1 or alpha and click OK (leave the
Column Input Cell blank). In EXCEL 2007: Data > What if
analysis>Data Table)
6. Excel will quickly fill in cells P4:R33 which are the forecasts for under
the different alpha values we specified in cells P3:R3.
7. Highlight the values in cells P4:R33, click the copy button, move your
cursor to cell L4 and the use the Edit and Paste Special menu choices.
Select Values under Paste and click OK. Please make sure that cells L3,
M3, and N3 are emptyyou dont have any forecasts for week 1.
8. Highlight the data in cells J2 to N33 and create a line chart (make sure to
specify that the first column is to be used for the X-axis labels).
a) Highlight the data in cells J2:N33.
b) Select the Insert and Chart menu choices.
c) In step 1 of the chart wizard, select a Line chart. Select the chart
that has lines connecting the dots (middle row, first column).
d) In step 2 of the chart wizard, the data to be plotted should be
$J$2:$N$33 (the data you had highlighted) and the data series in
columns.
e) In step 3 of the chart wizard, add the chart title Sales Forecasts, add
the x-axis title Week and the y-title Sales.
f) In step 4 of the chart wizard, select to add the chart As new sheet.
g) Admire your chart!

Using Excel for Forecasting

Part 3: Developing a Performance Table
Part 3 assumes that you have completed Part 1 and have set up a
spreadsheet as described in Part 1.
1. Add the following to your spreadsheet:
1
2
3
4
5
6
7
8
9
10
11
12
13

S
Alpha Value

T
MAD
=F1

0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1.0

2. Highlight cells S2:T13. Select the Data and Table menu choices. Under
the Column Input Cell, enter either B1 or alpha and click OK (leave
the Row Input Cell blank). In EXCEL 2007: Data > What if
analysis>Data Table)

3. Excel will quickly fill in cells T3:T13 which are the most recent (latest)

MAD values under the different alpha values we specified in cells S3:S13.
4. Thats ityouve just created a performance table. Now just select the
alpha that yields the best performance.

Part 4: Getting Excel to Find the Best Forecast

AUTOMATICALLY
Part 4 assumes that you have completed Parts 1 and 3 and have set up
a spreadsheet as described in Parts 1 and 3.
1. Add the following to your spreadsheet:
T

1
2
3

=IF(T3=T$14,S3,"
")

(rows 4-13 not shown)

14 =MIN(T3:T13)

=MIN(U3:U13)

2. Copy the formula in cell U3 into cells U4:U13. You will now see the best
value of alpha (of the initial values) in cell U14.
3. Next, we build a second performance table that narrows the range of
alpha.

17
18
19
20
30

S
Alpha Value

T
MAD
=F1

=MIN(.90,
MAX(U14-.05,
0))
=S19+.01

=IF(T19=T$30,
S19,"")

(rows 21-29 not shown)

=MIN(T19:T
29)

=MIN(U19:U2
9)

4. Copy the formula in cell S20 into cells S21:S29. Copy the formula in cell
U19 into cells U20:U29.
9. Highlight cells S18:T29. Select the Data and Table menu choices (In
EXCEL 2007: Data > What if analysis>Data Table)
5. . Under the Column Input Cell, enter either B1 or alpha and click OK
(leave the Row Input Cell blank). Cell U30 now has the best value of
alpha to the nearest 0.01.
6. Finally, in cell B1, put the formula "=U30". This takes the best value of
alpha that we found and plugs it back into our model (so that our graphs
are based on the most accurate forecast).

Using Excel for Forecasting

Part 5: Setting Up the Exponential Smoothing
With Trend

In setting up the exponential smoothing with trend model, we need

the initial estimates of the average (At) and trend (Tt). Heres one
way to do it.
Start by putting the label Alpha in cell a1 and Beta in cell c1 and
the values 0.3 in cells b1 and d1. Now names cells b1 Alpha and d1
Beta using the Insert, Name, Create menu choices (see part 1 if you
forgot how to do this). Next, enter the following:
1
2
3

A
B
C
Alpha
0.3
Beta
Period Demand
Average
1
1891
=average(B3:B7)-2*D3

1693

3
Etc

1693
Etc

D
0.3

=alpha*b4+(1alpha)*(c3+d3)

Trend
=(average(B8:B12)average(B3:B7))/5
=beta*(c4-c3)+(1beta)*d3

etc (copy down)

etc (copy down)

E
Periods
Forecast

=round(c3+d3,
0)
etc (copy down)

1. The formulas for the average (in cell C4), trend (in cell D4), and
forecast (in cell E4) are just the formulas for the ES/WT model.
2. The initial trend (cell D3) is found by taking the difference of the
average demand in the second five periods (i.e., periods 6-10) and
the average demand for the first five periods (i.e., periods 1-5), and
then dividing the difference by five.
3. The initial average is found by taking the average demand in the
first five periods (i.e., periods 1-5) and then subtracting 2 times the
initial trend. This gives a pretty good estimate of the initial
average.
4. Error and MAD formulas are entered similarly to what we did in
Part 1.

Using Excel for Forecasting

Part 6: Creating A Performance Table for the
Exponential Smoothing with Trend Model

Part 6 assumes that you have completed Part 5. Add the following to
your spreadsheet:
F
1
2
3
4
5
6
7
8

G
=offset(g2,$f$1,0)
MADt

30
Error
=E4-B4

=abs(F4)
=.15*abs(F5)+.85*
G4

etc (copy down)

Etc (copy down)

Next, add the following to your spreadsheet:

K
1
2
3
4
5
6
7
8
9
10
11
12
13

Beta

L
=g1
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1.0

M
Alpha
0

.1

.2

.3

.4

.5

.6

.7

.8

.9

1.0

Note that in cell L2 weve put a reference to the most recent MAD
value (based on the number of periods we've specified), which is cell
G1.
Now highlight the data in cells L2:W13 and then select the Data and
Table menu choices. In EXCEL 2007: Data > What if analysis>Data
Table)

For the Row Input Cell, specify B1 (or Alpha) and for the Column
Input Cell, specify D1 (or beta) then click OK. Voila! Excel fills in
the table for all combinations of alpha and beta.

Using Excel for Forecasting

Part 7: AUTOMATICALLY Finding the Best
Values of
Alpha and Beta
Part 7 assumes that you have completed Parts 5 and 6.
1. Add the following to your spreadsheet:
L
1
2
3

M
Alpha
0

=g1
0

=MIN(M3:W
3)

=MIN(Y3:Y13)
=IF(X3=X$14,L3
,"")

[Note: rows 4-13 and columns N-W not shown]

14
15

=MIN(M15:W1
5)

=MIN(M3:M13)
=IF(M14=$X14,M2
,"")

2. Copy:

a) The formula in cell M14 into cells N14:X14.

b) The formula in cell M15 into cells N15:W15.
c) The formula in cell X3 into cells X4:X13.
d) The formula in cell Y3 into cells Y4:Y13.
The best initial values of alpha is given in cell L15 and the best
initial value of beta is given in cell Y2.
3. Now we narrow the search to the nearest 0.01 for alpha and beta.
L
18
19

=g1

20

=MIN(.90,
MAX(Y2-.05,0))

21

=L20+.01

M
Alpha
=MIN(.9,
MAX(L15-.05,0))

=M19+.
01

Y
=MIN(Y20:Y30)

=MIN(
M20:W2
0)

=IF(X20=X$31,
L20,"")

[Rows 22-30 and columns O-W not shown]

31
32

=MIN(M32:W
32)

=MIN(M20:M30)
=IF(M31=$X31,M1
9,"")

4. Copy:

a) The formula in cell N19 into cells O19:W19.

b) The formula in cell L21 into cells L22:L30.
c) The formula in cell M31 into cells N31:X31.
d) The formula in cell M32 into cells N32:W32.
e) The formula in cell X20 into cells X21:X30.
f) The formula in cell Y20 into cells Y21:Y30.
5. Highlight the data in cells L19:W30 and then select the Data and
Table menu choices. For the Row Input Cell, specify B1 (or
Alpha) and for the Column Input Cell, specify D1 (or beta) then
click OK. Voila! Excel fills in the table for all combinations of

10

alpha and beta. In EXCEL 2007: Data > What if analysis>Data

Table)
6. Cells L32 and Y19 give the best values of alpha and beta,

respectively. We can place the best values of Alpha and Beta back
into the spreadsheet by putting the formula "=L32" into cell B1
and "=Y19" into cell D1.

11