Forecasting

FORECASTING

Forecasting is an estimation of an event that will happen in the future.

• The event may be the demand for a product, rainfall at a particular place, population
of the country or growth of technology.
• Estimation of demand for a particular product, inventory needs, personnel requirements
and other business variables
• The primary goal of OM is to match supply to demand. Forecasting demand is essential
for how much capacity or supply is needed to meet demand.
• Non-profit sectors also forecast: libraries, blood banks, police, fire depart., etc. base
their yearly plan on forecasts of needed services and expected revenues.
• 80 percent of forecasting is done with quantitative methods
• In industry, forecasting is the first level of decision activity.
• Forecasting today is becoming increasingly important as firms focus on increasing
customer satisfaction while reducing the cost of providing products or services.

• Business decisions always depend on forecasts.

• Every functional area of business makes use of some type of forecast.

1. Accountants rely on forecasting costs and revenues in tax planning.

2. Financial experts must forecast cash flows
3. Production managers rely on forecasting to determine raw-materials needs and the desired
inventory of finished products.
4. Marketing managers use sales forecasting to establish promotion budgets.
5. The personnel depart. depends on forecasting as it plans to recruit new employees
and other changes in the workforce.
Two professional forecasting organizations offer programs, and conferences. Summits
to increase the skills and abilities of business professionals who find forecasting
an important part of job responsibilities.
• The International Institute of Forecasters
• The Institute of Business Forecasting

Journals:
• The international journal of Applied Forecasting
• Journal of Business Forecasting

Forecasting has two aspects:

• The expected level of demand – the structural variation, trend, sessional

• The degree of accuracy that can be assigned to a forecasting

Time horizon:

• Short range forecasting- an hour, day, week, or up to a month –ongoing operations:

job scheduling, workforce levels, job assignment, production level
• Medium range forecasting- time span one to three years: sales planning,
production planning, cash budgeting
• Long range forecasting-three years to five years or more years: the life of the product,
new product development, new facility, facility location, research, and development

Forecasting Process:

1. Determine what to forecast: decide whether to forecast unit sales, $ sales, total sales,
sales by production line, domestic sales, sales in the region
2. Identify time dimensions: length and period of forecasting- annual, quarterly, month,
weekly, daily basis

3. Data Considerations: Quantity and the type of data that are available for forecasting.
- Internally, Externally
- Some data are not retained for use in forecasting. How frequently data are kept:
annual basis, quarterly basis, monthly basis, and daily basis.

4. Model selection: Depend on the number of criteria

(i) The demand pattern
(ii) The quantity of historical data available
(iii) The length of the forecasting horizon
Forecasting method, data pattern, the quantity of data, forecast horizon
5. Model evaluation: the result fit with historical data, error calculation, and accuracy.
When sufficient data is available, use a “ holdout” period to evaluate forecast accuracy.
If the model did not an acceptable level of accuracy, go to step no 5 and select an alternate
model.

6. Forecast Preparation: Some methods or sets of methods are used for developing the forecast.
Prepare a worst-case, a best-case, and most likely forecast. A combination of all forecasting
result for getting a better forecast.

7. Forecast presentation: Presentation of forecast results

8.Tracking result: Continuous tracking of how well forecast compares with actual values,
deviation from forecast and actual events should discuss and understand the errors that
occurred.
 Time Series Data

• The data that is used most often in forecasting are time series.
• A time series is a set of numerical values of some variable obtained at a regular period
over time.
• The series is usually tabulated or graphed and understand the behavior of the variable.

Example: Sales data by month from January 2017 to December 2021

No. of visitors to a national park last year
Stock price daily basis

Objectives of Time Series Analysis:

• Some underline patterns exist in historical data.
• Objective to identify the pattern and influencing factors for the prediction of future
planning and control.
 Time Series Patterns

Actual value of the variable at time t = mean value at time t + random deviation from mean value

Y = pattern(mean) + ɛ (noise)

Demand Patterns

Historical Pattern:
This exists where there is the trend in data when the mean value does not change over time.

Trend: Time series display either increasing or decreasing in the average values
of the forecasting variable over time. Sales of products, stock price.
Cycles:
Wavelike upward and downward movement of the data about the trend time over a
period of time.
Price of metals, gross national products

Seasonal:
It is a special case of a cycle in which fluctuations are repeated usually within a year.
Sales of soft drinks, sales of refrigerators, sales of wool items

Irregular:
Random fluctuations and most difficult to capture in the forecasting model.
 Forecasting Models
The forecasting techniques can be classified into quantitative techniques and
qualitative techniques.

Quantitative techniques --- Historical data

Qualitative techniques --- Subjective approaches

 Quantitative Forecasting Techniques

• Simple moving average

• Weighted moving average
• Simple exponential smoothing
• Exponential smoothing with linear trend
• Simple regression
• Seasonal index model
 Qualitative Techniques

• Delphi method
• Market Research
 Simple Moving Average:
The moving averages which serve as an estimate of the next period’s value of a variable
given a period of length n.
OR
SMA is a method of computing the average of a specified number of the most recent data
value in a series.

Moving average

Mat = {Dt – (n-1)+Dt –(n-2)+….+ Dt-2+Dt-1 + Dt}/n

Where Mt – Simple moving average at the end of period t

(It is to be used as a forecasting for period t+1)
Dt – Actual demand in period t
n- Number of periods
Ft+1- forecast for period t+1
• A manufacturer of television has the following data pertaining the actual demand for television
(in thousands) during in a year. Use the three-time moving average method to forecast the
demand of next year month of January.
Time Demand Moving Forecast Ft
Month (t) Month (t) Average Mt

1 95
2 100
3 87
4 123
5 90
6 96
7 75
8 78
9 106
10 104
11 89
12 83

Forecast for period of 13 month.

 Weighted Moving Average
•Equal weights were assigned to all period in a simple moving average.
•In some demand value weighted given more than others.

Problem
In the table for a three months weighted moving average with a weight of 0.50 assigned
to the most recent demand value, 0.30 assigned to the next most recent value and 0.20
assigned to the oldest of the demand value. Forecast the demand for period 7.

Time (t) Demand MA Forecast

1 120
2 130
3 110 118
4 140 129 118
5 110 119 129
6 130 126 119
Weight MA3 = (0.2x120 + 0.3x 130 + 0.5x 110)/(0.2+0.3+0.5)
=118

Weight Mai = ∑wiDi/ ∑wi

 Exponential Smoothing Method

•It is sophisticated weighted moving average

•ESM use for forecasting for a large number of items.

•The forecasting horizon is relatively short i.e. daily, weekly or monthly demand.

•ESM model is applicable when there is no trend or seasonality component in the data.

•Demand fluctuation around an average demand which is called “base”.

•It keep on running average demand and adjust it for each period in proportion to the
difference between the latest actual demand figure and the latest value of the average.
 Exponential Smoothing Method
Ft = Ft-1 +α (Dt-1 – Ft-1)

Ft = αDt-1 + (1- α)Ft-1

Where Ft= current period forecast

Ft-1= last period forecast
α = a weight called smoothing constant
(0 < α <1)
Dt-1= last period actual demand

•If α = 0, Ft = Ft-1
The new base is just same as previous base (average)
•If α = 1, Ft= Dt-1
The new base is the previous period demand.
If the value of α is small are smooth and fluctuation is less.
 Problem

A firm uses simple exponential smoothing with α = 0.2 to forecast demand. The
forecast for the first week of January was 400 units, where as the actual demand
turned out to be 450 units:

a) Forecast the demand for the second week of January.

b) Assume that the actual demand during the second week of January turned out to
be 460 units. Forecast the demand up to February fourth week, assuming the
sequence demand as 465,434, 420, 498 and 462 units.
Forecast of second week of January:

Ft= αDt-1 + (1-α) Ft-1

F2 = 0.2x450 + 0.8 x 400

= 410 units
Week Demand old forecast new forecast
Dt-1 Ft-1 Ft= αDt-1 + (1-α)Ft-1

Jan 1 450 400 410

Jan 2 460 410 420
Jan 3 465 420 429
Jan 4 434 429 430

Feb 1 420 430 428

Feb 2 498 428 442
Feb 3 462 442 446
 Home Assignment
A food processor uses ESM with α= 0.10 to forecast the next months demand. Past(actual)
demand in units and the simple exponential forecast up to month 51 are shown in the following table:
Month Actual Old forecast
Demand
43 105 100.00
44 106 100.50
45 110 101.05
46 110 101.95
47 114 102.46
48 121 103.61
49 130 105.35
50 128 107.82
51 137 109.84
a) Using simple exponential smoothing, forecast the demand for month 52.
Answer: a) 112.56
 Forecasting Error
•Demand forecast influences most of the decisions in all functions.
•It must be estimated with the highest level of precision.
•Accuracy and control of forecasting is a vital aspect of forecasting.
•Decision makers will want to include accuracy as a factor.
•The measure may be an aggregate error (deviation) of the forecasting values from the
actual demand.

Error = actual - forecast

Et = Dt – Ft
t= time period

MAD= mean absolute deviation

MSE= mean squared error
MAPE= mean absolute percent error
MAD=∑|actual t – forecast t | /n

MSE= ∑(|actual t – forecast t | )2 /(n-1)

MAPE=∑{ (|actual t – forecast t |)/ actual t x 100}/n

Compute MAD, MSE, MAPE for the following data.

Period Actual Forecast

1 217 215
2 213 216
3 216 215
4 210 214
5 213 211
6 219 214
7 216 217
8 212 216
Period Actual Forecast Error |Error| Error*Error
1 217 215 2 2 4
2 213 216 -3 3 9
3 216 215 1 1 1
4 210 214 -4 4 16
5 213 211 2 2 4
6 219 214 5 5 25
7 216 217 -1 1 1
8 212 216 -4 4 16
Total -2 22 76
 Problem 2
Month Demand Forecast
1 150 165
2 160 165
3 165 170
4 175 180
5 180 165
6 182 174
7 176 182
8 181 167
9 164 170
10 155 163
11 162 168
12 170 174

Find MAD, MSE, MAPE?

 Exponential Smoothing with Linear Trend Model
•The Simple ESM forecast is a smoothing average position on the current period.

•In reality, trend exits in the demand pattern of many business.

•Adjusted ESM is projects the next period forecast by adding a trend component to the
current period forecast Ft

•Ft+1 =αDt + (1-α) (Ft + Tt)

•Tt+1 = β(Ft+1 –Ft )+ (1-β) Tt
• New forecasting, Ft+1 = Ft+1 + T t+1
α and β are smoothing constants
(0 < α <1) smoothing constant for level,
(0 < β <1) smoothing constant for the trend estimate
Example:

Compute the linear or trend ESM forecast for the first week of March for a firm with
the following data. Assume forecast for the first week of January(F 0) as 600 and
Corresponding initial trend (T0) is 0. Let α =0.1 and β=0.2

Week Demand
Week 1 650
Week 2 600
Week 3 550
Week 4 650
Week 5 665
Week 6 675

Forecast for next Week.

Week Previous Actual Smoothed Smoothe Next period
Average Demand Average d Trend Projection
(Ft-1 ) (Dt-1 ) (Ft ) (Tt ) (Ft+1 )

Week 1 600.00 650.00 605.00 1.00 606.00

Week 2 605.00 600.00 605.40 0.880 606.28

Week 3 605.40 550.00 600.65 -0.246 600.40

Week 4 600.65 650.00 605.36 0.742 606.10

Week 5 605.36 625.00 607.99 1.120 609.11

Week 6 607.99 675.00 615.70 2.440 618.11
 Linear Regression :

The relationship between a dependent y and only one independent variable x, then
such model is called a simple regression model.

Y = a+ bX

Y= dependent variable
X= independent variable
a= y- intercept
b= slope (trend)

XY – nXY
a = Y – bX b=
X 2 – n(X) 2
A firm believes that its annual profit depends on its expenditures for research. The information for the
preceding six years is given below:

Year Expenditure for research (x)Lac Annual profit (y)Lac

1 2 20
2 3 25
3 5 34
4 4 30
5 11 40
6 5 31

Forecast the profit for the 7th year when the expenditure is 6 lac.

Appling the formula you will get the value of b=2 and a= 20

Y= 20 + 2x
= 20 + 2 (6) = 32 lac
The profit in the 7th year when the expenditure is 6 lac will be 32 lac.
The advertisement expenditure and the actual sales (in thousand) for five months
is given below. Using regression analysis forecast the sales for advertisement
expenditure is 2.3 thousand $.

Sales Advertising
Month (000 units) (000 $)

1 264 2.5
2 116 1.3
3 165 1.4
4 101 1.0
5 209 2.0
 Sales Advertising
Month (000 units) (000 $)
1 264 2.5
2 116 1.3
3 165 1.4
4 101 1.0
5 209 2.0

XY – nXY
a = Y – bX b=
X 2 – nX 2
 Sales, Y Advertising, X
Month (000 units) (000 $) XY X2 Y2
1 264 2.5 660.0 6.25 69,696
2 116 1.3 150.8 1.69 13,456
3 165 1.4 231.0 1.96 27,225
4 101 1.0 101.0 1.00 10,201
5 209 2.0 418.0 4.00 43,681

XY – nX Y
a = Y – bX b=
X 2 – n(X) 2
 Sales, Y Advertising, X
Month (000 units) (000 $) XY X2 Y2
1 264 2.5 660.0 6.25 69,696
2 116 1.3 150.8 1.69 13,456
3 165 1.4 231.0 1.96 27,225
4 101 1.0 101.0 1.00 10,201
5 209 2.0 418.0 4.00 43,681
Total 855 8.2 1560.8 14.90 164,259
Y = 171 X = 1.64

XY – nXY
a = Y – bX b=
X 2 – n(X) 2
 Sales, Y Advertising, X
Month (000 units) (000 $) XY X2 Y2
1 264 2.5 660.0 6.25 69,696
2 116 1.3 150.8 1.69 13,456
3 165 1.4 231.0 1.96 27,225
4 101 1.0 101.0 1.00 10,201
5 209 2.0 418.0 4.00 43,681
Total 855 8.2 1560.8 14.90 164,259
Y = 171 X = 1.64

a = Y – bX b = 109.229
 Sales, Y Advertising, X
Month (000 units) (000 $) XY X2 Y2
1 264 2.5 660.0 6.25 69,696
2 116 1.3 150.8 1.69 13,456
3 165 1.4 231.0 1.96 27,225
4 101 1.0 101.0 1.00 10,201
5 209 2.0 418.0 4.00 43,681
Total 855 8.2 1560.8 14.90 164,259
Y = 171 X = 1.64

a = – 8.136 b = 109.229
 Sales, Y Advertising, X
Month (000 units) (000 $) XY X2 Y2
1 264 2.5 660.0 6.25 69,696
2 116 1.3 150.8 1.69 13,456
3 165 1.4 231.0 1.96 27,225
4 101 1.0 101.0 1.00 10,201
5 209 2.0 418.0 4.00 43,681
Total 855 8.2 1560.8 14.90 164,259
Y = 171 X = 1.64

a = – 8.136 b = 109.229
Y = – 8.136 + 109.229(X)
Problem:

Alpha company has the following sales pattern. Compute the sales forecasting for
the year 10.
Year 1 2 3 4 5 6 7 8 9
Sales 6 8 11 23 29 34 40 45 56
(in lacs)
Home Assignment:

Alpha company has the following sales pattern. Compute the sales forecasting for
the year 10.
Year 1 2 3 4 5 6 7 8 9
Sales 6 8 11 23 29 34 40 45 56
(in lacs)
 Home Assignment
A company that manufactures steel observed the production of steel(in metric tones)
represented by the time series

Year Production of Steel

1996 60
1997 72
1998 75
1999 65
2000 80
2001 85
2002 95

a) Find the liner equation that describe the trend in the production of steel by the company.
b) Estimate the production of steel in 2003.
Answer: a) y= 76+4.857x b) 95.428 metric tons.
 Seasonal Models

Consider the following data for the demand for school uniforms from a garment store.
The data in the table shows the demand (in hundreds) for four quarters of the year.
Fit seasonal model for the data.

Year I Year II Year III

April- June 53 58 62
July- September 22 25 27
October- December 37 40 44
January- March 45 50 56
The data in table shows the demand of a product for four quarters of the year.
Fit seasonal model for the data.

Year I Year II Year III

April- June 520 590 650

July- September 730 810 900

October- December 820 900 1000

January- March 530 600 650

 Delphi Method
•It is used as long term prediction.
•The objective of Delphi method is to probe into the future in hopes of
new products and processes on the rapid change of environment of today culture and
economy.
•In the short range, Delphi method can also be used to estimate market size and timing.

•Examine the development of computer languages- prediction of fifth generation

•Several knowledge persons are asked to provide subjective estimates of demand or

forecasting of possible advances of technology.
•The experts may provide several opinions. Based on the opinions arrived at the demand
of product/advance tech.
•Panel members must be unknown to each other.
•The initial questionnaire should be unambiguous and it should explain every matter.

•After getting the opinions from the panel members they are to be compared for
similarity.
•If the variation among the opinion is too much, the summary is to be circulated again
among the members and again provide for opinions.
•Generally 50% of the estimate is treated as the basis for comparison.
•The panel members whose opinions differ significantly from the middle of 50% of the
estimate will asked for the reconsider their opinions.
•Still, if they want to stick to their original opinions they will be asked to provide rationale
for the same.

•Delphi method is an iterative process, until the panel converges on the specific value
or range of values as defined by the required accuracy or arrives at point under
considerations.
 Marketing Research

 Primary Research

 Surveys
• Interviews
• Observations
• Experiment, field project

 Secondary Research – From existing data