CORE MRP II

1
CORE MRP II
MANUFACTURING RESOURCE PLANNING
2
FORECASTING

• Time-series (intrinsic) models
• Time series components
• Base-level forecasting models
• Evaluating forecasting models Mad versus Bias
• Adding trend and seasonality
• Selecting a technique

3
TIME-SERIES (INTRINSIC) FORECASTING MODELS

• Make no assumption about what causes sales
• Identify patterns in past data
• Project those patterns into the future
• Also called autoprojection, intrinsic techniques

4
TIME SERIES COMPONENTS

• Predictable components
• Base level
• Trend
• Seasonality
• Unpredictable components
• Randomness
• All time series techniques work by
• Smoothing randomness out of the past data
•  Projecting the leftover pattern implied by these
  predictable components into the future

5
SHORT-TERM FORECASTING TECHNIQUES

• Assume absence of trend and seasonality
• Separate base-level (average level) from the
  randomness
• Some useful models
• Moving Averages
• Exponential Smoothing

6
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods

7
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

8
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

9
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

10
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

11
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

12
SIMPLE MOVING AVERAGE

• Smoothes out randomness by averaging positive and
  negative random elements over several periods
• n -- number of periods
• n -- 3

13
FORECAST EVALUATION

• Need to compare forecast to actual sales
• BIAS
• Too high or to low (on average)?
• Mean Absolute Deviation (MAD)
• Off by how much, per period (on average)?

14
FORECAST EVALUATION

• Need to introduce terminology

15
FORECAST EVALUATION
16
FORECAST EVALUATION
17
WEIGHTED MOVING AVERAGE

• Same idea as SMA, but less smoothing more
  weight on recent sales data
• n -- number of periods
• ai weight applied to period t-i1

18
WEIGHTED MOVING AVERAGE

• Same idea as SMA, but less smoothing more
  weight on recent sales data
• n -- number of periods
• n -- 3
• ai weight applied to period t-i1

19
WEIGHTED MOVING AVERAGE

• Same idea as SMA, but less smoothing more
  weight on recent sales data
• n -- number of periods
• n -- 3
• ai weight applied to period t-i1

20
WEIGHTED MOVING AVERAGE

• Same idea as SMA, but less smoothing more
  weight on recent sales data
• n -- number of periods
• n -- 3
• ai weight applied to period t-i1

21
EXPONENTIAL SMOOTHING

• Simpler equation, equivalent to WMA
• a exponential smoothing parameter (0lt alt1)
• Higher alpha more reactive, less smoothing
• Lower alpha less reactive, more smoothing

22
EXPONENTIAL SMOOTHING

• Simpler equation, equivalent to WMA
• a exponential smoothing parameter (0lt alt1)
• Higher alpha more reactive, less smoothing
• Lower alpha less reactive, more smoothing

23
EXPONENTIAL SMOOTHING

• Simpler equation, equivalent to WMA
• a exponential smoothing parameter (0lt alt1)
• Higher alpha more reactive, less smoothing
• Lower alpha less reactive, more smoothing

24
EXPONENTIAL SMOOTHING

• Simpler equation, equivalent to WMA
• a exponential smoothing parameter (0lt alt1)
• Higher alpha more reactive, less smoothing
• Lower alpha less reactive, more smoothing

25
EXPONENTIAL SMOOTHING

• Simpler equation, equivalent to WMA
• a exponential smoothing parameter (0lt alt1)
• Higher alpha more reactive, less smoothing
• Lower alpha less reactive, more smoothing

26
EXPONENTIAL SMOOTHING

• A higher smoothing parameter means less smoothing
  and a more reactive forecast

27
ADDING TREND SEASONALITY

• Can use estimates of trend and seasonality to
  extend basic smoothing equation
• a -- Smoothing parameter (0 lt a lt 1) 
• T -- Estimate of trend component
• Increase/decrease of sales per year
• L -- number of seasons per year (4)

28
ADDING TREND SEASONALITY

• Can use estimates of trend and seasonality to
  extend basic smoothing equation
• a -- Smoothing parameter (0 lt a lt 1) 
• T -- Estimate of trend component
• Increase/decrease of sales per year
• L -- number of seasons per year (4)

29
ADDING TREND SEASONALITY

• Can use estimates of trend and seasonality to
  extend basic smoothing equation
• a -- Smoothing parameter (0 lt a lt 1) 
• T -- Estimate of trend component
• Increase/decrease of sales per year
• L -- number of seasons per year (4)

30
ADDING TREND SEASONALITY

• Can use estimates of trend and seasonality to
  extend basic smoothing equation
• a -- Smoothing parameter (0 lt a lt 1) 
• T -- Estimate of trend component
• Increase/decrease of sales per year
• L -- number of seasons per year (4)

31
ADDING TREND SEASONALITY

• Can use estimates of trend and seasonality to
  extend basic smoothing equation
• a -- Smoothing parameter (0 lt a lt 1) 
• T -- Estimate of trend component
• Increase/decrease of sales per year
• L -- number of seasons per year (4)

32
ADDING TREND SEASONALITY

• When forecasting mgt1 periods ahead, use more
  general formula

33
SELECTING A TECHNIQUE

• Ockham's razor -- use the simplest possible model
  or theory (William of Ockham, 1300-1349, England)
• 1) Determine type of technique which is
  appropriate (i.E., Base-level, trend, etc.)
• 2) Select a group of competing techniques which
  satisfy condition (1)
• 3) Select a set of data as a test set
• 4) Simulate forecasts for this set of data using
  all techniques from (2)
• 5) Pick the technique with the best combination
  of MAD/MAPE and Bias