You are on page 1of 39

Independent Versus Dependent Demand

Only independent demand needs to be forecasted.


Dependent demand should not be forecasted; it should
be calculated.
Sources of Demand

Forecasts
Customer orders
Replenishment orders from distribution centers
Interplant transfers
Other sources of demand
Demand Patterns: Trend

Increasing
Decreasing
Level
Demand

Quarters
Demand Patterns: Seasonal Demand
Demand

Quarters
Demand Patterns: Random
5

4
Demand

Quarters
Stable Versus Dynamic Demand

Stable demand retains same general shape over time.


Dynamic demand tends to be erratic.
Dynamic
Stable
Stable versus
dynamic
demand

Average demand
Introduction: Forecasting

Purposes and uses of the forecast


Principles of forecasting
Principles of data collection and preparation
How Forecasting Supports Planning

Planning level Forecast Horizon (up to)

Sales volume ($); new


Business planning market and supply 2 to 10 years
chain initiatives

Sales and operations Physical units of


production at the 1 to 3 years
planning product family level

Physical units of
Master scheduling production at the end 3 to 18 months
item level
Principles of Forecasting

Forecasts
– are rarely 100 percent accurate over time
– should include an estimate of error
– are more accurate for product groups and families
– are more accurate for nearer periods of time.
Data Collection and Preparation

Record data in terms needed for the forecast.


Record circumstances relating to the data.
Record demand separately for different customer
groups.
Data Collection and Preparation Example

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

A 6000 6000

B 500 500 500 500 500 500 500 500 500 500 500 500

Average
1500 1500 1500 1500 1500 1500 1500 1500 1500 1500 1500 1500
forecast

Customer A’s annual demand: 12,000


Customer B’s annual demand: 6,000
Total: 18,000
Average over 12 months: 1,500 per month
Forecasting Techniques

Forecasting
techniques

Qualitative Quantitative

Judgment Mathematics

Intrinsic Extrinsic
(time series) (causal)
Qualitative techniques

are based on intuition and informed opinion


tend to be subjective
are used for business planning and forecasting for new
products
are used for medium-term to long-term forecasting.
Quantitative Techniques: Extrinsic

Extrinsic
is based on correlation and causality
relies on external indicators
is useful in forecasting total company demand or
demand for families of products
has two types of leading indicators
– economic
– demographic.
Quantitative Techniques: Intrinsic

Intrinsic is based on several assumptions:


– The past helps you understand the future.
– Time series are available.
– The past pattern of demand predicts the future pattern of
demand.
Examples:
– moving averages
– exponential smoothing
Moving Averages: Principles

Moving averages are best used when demand is


stable, there is little trend or seasonality, and
demand variations are random.
When past demand shows random variation
– do not second-guess what the effect of random variation will
be
– it is better to forecast based on average demand.
Moving Average Forecast Example

Assume it is the end of December;


forecast demand for the next month, January.

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan
Month Mont Month
1 h2 3

92 83 66 74 75 84 84 81 75 63 91 84 ?
Moving Average Forecast Logic
Moving average forecast = average demand of past periods

Moving average forecast for month 4


Σ demand for months 1 - 3 288
= = = 96 units
number of months 3

Month Demand Three-month total Forecast

1 102
2 91
3 95 288
4 96

Key:  = sum
Month 4 forecast
Three-month
Month Demand Forecast
total
1 102

2 91

3 95 288

4 105 291 96

5 94 294 97

6 101 300 98

7 100
Three-Month Moving-Average Forecast

Month Demand Three-month total Forecast


1 89
2 89
3 94 272
4 91 274 91
5 95 280 91
6 104 290 93
7 106 305 97
8 110 320 102
9 107
Six-Month Moving-Average Forecast
Month Demand Six-month total Forecast
1 89
2 89
3 94
4 91
5 95
6 104 562
7 106 579 94
8 110 600 97
9 100
Moving Averages: Lessons Learned

The moving average forecast will make the development


of a rising or falling trend lag.
The farther back the moving average forecast reaches
for data, the greater the lag.
The three-month moving-average forecast may have
overreacted if the demand surge had abated.
The moving-average forecast works best when demand
is stable with random variation; it will filter out random
variation.
Exponential Smoothing Logic

Take the old forecast and the actual demand for the
latest or most current period.
Assign a weighting factor or smoothing constant
(α, alpha) to the latest period demand versus the old
forecast.
Calculate the weighted average of the old forecast and
the latest demand.

New forecast = (α) (latest demand) + (1 – α) ( old forecast)


Smoothing Constant (α, Alpha)

New forecast = (α) (latest demand) + (1 – α) (previous forecast)

A low smoothing constant gives more weight to the old


forecast. Here are some examples:
– α = .2 for latest demand (for example, period X)
– 1 – α = .8 for old forecast (for example, period X)
This is appropriate if demand is stable, not rising or
falling.
Run simulations with different α values to see which one
best fits the historical demand pattern.
New forecast = (α) (latest demand) + (1 – α) (previous forecast)

A. Prepare an exponential smoothing forecast for June.


May data: actual demand = 220; forecast = 200.

Calculate the forecast for June using a smoothing constant (α) of .20.

B. Prepare an exponential smoothing forecast for July.


June data: actual demand = 240

Calculate the forecast for July also using a smoothing constant (α) of .20.
New forecast = (α) (latest demand) + (1 – α) (previous forecast)

A. Prepare an exponential smoothing forecast for June.


= (.2) 220 + (.8) 200 =
= 44 + 160 = 204

B. Prepare an exponential smoothing forecast for July.


= (.2) 240 + (.8) 204 =
= 48 + 163 = 211
Seasonal Demand

Average demand
for all periods
Demand (units)

Seasonal demand

Time (quarters)
Seasonal Forecast Process

Develop a seasonal forecast for each


period of the year being forecast.
3

Develop a deseasonalized demand forecast


spanning all periods.
2

Calculate a seasonal index of demand for


each period to establish seasonality.
1
Seasonal Demand Indexes (Step 1)
Demand history
Year Quarter Total
1 2 3 4
1 122 108 81 90 401
2 130 100 73 96 399
3 132 98 71 99 400
Average 128 102 75 95 400

Average demand for all quarters = 400 = 100 units.


4

Quarter Average quarterly demand/100 Seasonal index


1 128/100 = 1.28
2 102/100 = 1.02
3 75/100 = 0.75
4 95/100 = 0.95
Total = 4.00
Deseasonalized Forecast (Step 2)

Make the forecast for the next year.


Deseasonalize the forecast—distribute it evenly across
the four quarters.

Deseasonalized demand annual forecast


(average demand/period)
=
number of periods

420
= = 105 units
4
Seasonal Forecast (Step 3)
Calculation

= (seasonal index) 
Expected quarter demand (deseasonalized forecast
demand)
Expected first quarter demand = 1.28 X 105 = 134 units
Expected second quarter =
1.02 X 105 = 107 units
demand
Expected third quarter demand = .75 X 105 = 79 units
Expected fourth quarter =
.95 X 105 = 100 units
demand
Total forecast demand = 420 units
Tracking the Forecast

Forecasts are rarely 100 percent correct over time.


Why track the forecast?
– To understand why demand differs from the forecast
– To plan around error in the future
– To improve forecasting methods
Bias Versus Random Variation
Bias Random variation
Cumulative demand may not be the Demand will vary plus and minus
same as forecast about the average
Month Forecast Actual Variation Forecast Actual Variation

1 100 90 -10 100 105 +5


2 100 125 +25 100 94 -6
3 100 120 +20 100 98 -2
4 100 125 +25 100 104 +4
5 100 120 +20 100 103 +3
6 100 110 +10 100 96 -4
Cumulative
total 600 690 +90 600 600 0

Bias exists since cumulative There is no bias since


variation is not zero. cumulative variation is zero.
Forecast Error Data

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total

Forecast 500 500 500 500 500 500 500 500 500 500 500 500 -

Actual 460 520 530 490 460 500 530 490 530 480 490 520 -

Absolute
40 20 30 10 40 0 30 10 30 20 10 20 260
deviation
Mean Absolute Deviation (MAD)

 | ||A - F|
MAD = n
n

Σ absolute errors 260


MAD = = = 22 units
number of periods 12

Key:  = sum; I I = absolute value


MAD Analysis: Normal Distribution

1% 4% 15% 30% 30% 15% 4% 1%

-3 -2 -1 0 1 2 3 MAD

-66 -44 -22 22 44 66 Units


Uses of Forecast Measurement

Identify changes and trends in demand.


Identify and adjust for forecast error that results from
random events.
Adjust the period forecast so that it is close to the true
forecast average demand to minimize bias.
Making decisions on safety stock and service levels
based on the degree of random variation (forecast error).
Supply Chain Management Implications
Deal with demand uncertainty through process improvements.

Decrease reliance on long-term forecasts and increase


ability to react quickly to demand.
Collaborate with customers and suppliers, especially in
sharing demand information.
Increase manufacturing flexibility internally and
operations integration externally with customers and
suppliers.

You might also like