Forecast Simulation

Document Sample
Forecast Simulation Powered By Docstoc
					   Forecast
 Simulation
User Group Presentation
              Table of Contents
• Overview
• Limitations
• Setting up Forecast Simulation
• Processing Forecast Simulation
• Summary
• Demonstration of calculating forecasts for four
  patterns of demand (cyclical, seasonal, trend and
  horizontal)
• Demonstration of factors used to manipulate
  forecast results (combine/replace, base, scale
  and trend)
                 Overview


Forecast Simulation provides a means to create a
forecast by extrapolating sales history (cph_hist)
and applying one of several predefined statistical
methods and one or more weighing factors.
                   Process

TEMPLATE
CALCULATE
    Sales History (cph_hist)
    Statistical Method (01 - 06)
    Weighing Factors (alpha and trend)
    Forecast
ADJUST
COPY/COMBINE
    More Weighing Factors (base, scale, trend)
SIMULATION TO SUMMARIZED
                             Limitations
•   Cannot use data other than sales history in the
    calculation

•   Cannot forecast configured products
    Components should be forecasted.


•   Cannot forecast a base process
    A base process is a process resulting in the production of one or more co-
    products and by-products.



•   Cannot forecast by sales channel
    This is an optional code that identifies where the sale originated.
                        Limitations
•Cannot  generate forecast based on dollar value of
sales history


•Memo  items and drop shipments are excluded
from any forecast calculation.
A memo item is a non-inventory item (for example, setup charges,
service charges). Memo items are not found in the item master file
and have no effect on MRP.
A drop shipment is a distribution arrangement in which the seller
serves as a selling agent by collecting orders but does not maintain
inventories. Upon receipt of the order, the manufacturer ships
directly to the customer.
              Setting up Forecast Simulation




This is the main menu of the Forecast Simulation Module.
The first step is to create a Forecast Template in 22.7.1 Simulation Criteria
Maintenance. The template, together with sales history will be used latter in
22.7.5 Simulation Forecast Calculation to create a forecast.
Forecast ID: Forecast ID, year and item identify forecast simulation results.
 It cannot be blank and must be unique. It accepts alphanumeric characters.
Years of History: Years of history upon which to base the forecast calculation.
The criteria template may be stored with years of history set to zero. At the
time of calculation you are prompted for a number between 1 and 5.
Forecast Year: The fiscal year to which this forecast applies.

Ending: This is the last year of sales history that the forecast calculation
analyzes. Ending Year and Years of History determine the amount of sales
history used by the forecast calculation. Forecast criteria may be stored
with Ending Year blank. At the time of the calculation there is a prompt to
enter Ending Year. The default is the previous year.
The “Ending” and “Forecast Year” fields are used in conjunction with
  each other to create either an Annual or Rolling Forecast.

Annual: The Ending year must be earlier than the Forecast year.

Rolling: The current year is both the Forecast Year and the Ending year. Sales history through the month
     before the current month is analyzed to calculate forecast quantities for the current month and 11
     months forward.
Identify Template
                                                                    Define
                                                                   Forecast
                                                                    Criteria




          Yearly Forecast               Current Date: Jan. ‘99
                                       Forecast Year: 1999
                                        Ending Year: 1998
   Quantity:
         Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
   1999   120 125 127 126 140 195 200 210 190 130 120 120



          Rolling Forecast             Current Date: Apr. ‘99
                                      Forecast Year: 1999
   Quantity:                           Ending Year: 1999
         Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
   1999                 126 135 175 195 210 190 140 130 120
   2000   120 125 127

                               uses history up through Mar. 1999
Forecast Method: Two-digit integer field. The system maintains several
statistical methods (there are six explained in more detail below).
Numbers between 00-50 are reserved for QAD. Users may define their
own forecast method using numbers 51-99 in User Forecast Method
Maintenance.
* Method 01 - Best Fit (default) This uses all predefined methods (02-06) and selects
the results with the least Mean Absolute Deviation.

* Method 02 - Double Moving Average This method is the simplest of forecasting
techniques. It uses a set of simple moving averages based on historical data and
then computes another set of moving averages based on the first set. The moving
averages are based on four months of data. This method produces a forecast that
lags behind trends effects.
* Method 03 - Double Exponential Smoothing. This method is the most popular forecasting
technique. It is similar to the Double Moving Average with the addition that it weights the most
recent sales data more heavily than than the older sales data. This method produces a forecast
that lags behind trends effects.

* Method 04 - Winter’s Linear Exponential Smoothing. This method produces results similar
to Double Exponential Smoothing, with the added advantage of incorporating a seasonal/trend
adjustment factor. This method can be used to forecast based on sales history that contains
both trends and seasonal patterns.
* Method 05- Classic Decomposition This method recognizes three separate portions of underlying
patterns within sales history: the trend factor, a seasonal factor, and a cyclical factor. The trend is
assumed to be a straight line that eliminates all random fluctuations in sales. The cyclical factor
follows the patterns of a wave ranging between high and low sales values. The cyclical factor spans a
period greater than one year. Classical decomposition is usually the preferred method to forecast
seasonal, high-cost items.
* Method 06- Simple Regression This is also called the least squared method. It analyzes the
relationship between the objects (sales) and time span (month). It ensures that the forecasted quantity
is equally likely to be higher or lower than the actual quantity sold.
Alpha factor: The Alpha factor determines the relative importance of sales history. It
must be greater than zero and less than one. The default is 0.40. A large alpha (close to
1) gives more weight to recent history. A small alpha (close to 0) gives increasingly
equal weight to all sales history. Recent sales data is more indicative of the future than
older data. Alpha is a weighing factor calculation only used with Methods 03 or 04.
Selection of an appropriate alpha value depends on your business environment. For a
new product with rapidly changing sales quantities, select a large alpha to produce
more accurate forecast. If the product’s sales history is long and stable, select a small
alpha to produce smoother forecast results.
Trend: Trend determines the degree to which sharp increases/decreases in sales history
are weighted by the forecast calculation. It must be greater than zero and less than one.
The default is 0.10. A large trend (close to 1) weighs heavily any sharp changes in sales
history. A smaller trend (closer to zero) begins to ignore sharp changes. Trend is only used
for Method 04. Sales history often includes seasonal patterns. Trend factor allows you to
determine how well the forecast should predict prior seasonal effects.
Sales History Patterns
                                     Define

       Trend             Seasonal   Forecast
                                     Criteria




New!




  Horizontal             Cyclical
Underlying Patterns                                                             Analyze
                                                                                History




                      Double         Winter’s Linear
 Double Moving      Exponential       Exponential          Classic           Simple
  Exponential       Smoothing         Smoothing         Decomposition      Regression
      02                03                 04                05                06



         Method                      02        03      04        05       06

         Cyclical                                               yes
         Trend            lags lags                    yes      yes
         Seasonal                                      yes      yes
         Horizontal                                                       yes
         Years of History   1    1                      2        2-3       1
         Trend Factor                                  yes
         Alpha Factor           yes                    yes
Note that it is suggested to use specific Forecast Methods with specific Demand Patterns.
Note the Years of History required for each Forecast Method.
Note that a Trend Factor can only be used for Forecast Method 04.
Note that an Alpha Factor can only be used with Forecast Methods 03and 04.
User factor [1] and User factor[2]: User factor[1] and User factor[2] are associated with
user defined forecast methods - Methods 51-99. This forecasting module is designed
to allow you to easily incorporate your own forecast methods. User factor[1] and User
factor[2] are similar to Alpha Factor and trend. They are reserved for you to interface
to your own forecast method.
 User factor[1] and User factor[2] can be negative and greater than one.
 User factor[1] and User factor[2] do not work with any of the predefined forecast
 methods (Methods 01-06)
Item Number: Item codes uniquely identify items or products - raw materials,
purchased or manufactured intermediates, finished items, packing materials,
whatever. Item codes also identify planning items, service kits, and repair parts used
in service activities.
Product Line: Product line codes identify major item and product groupings.
A forecast is calculated for every item within the product line.
Group: An optional code maintained in Item Master Maintenance you can use to
categorize similar items. A forecast is calculated for every item within the product
group.
Item Type: An optional code maintained in Item Master Maintenance you can use to
categorize similar items. A forecast is calculated for every item of this type.
Order Line Site: This field may be left blank. The default is the order line site specified
on the sales order. You can specify a Site to further delineate what sales history data is
retrieved by the forecast calculation. This may or may not be the site to which you ship
items. This site code displays as the default for each line item of the order, but you can
change it manually based on inventory availability or transport requirements.
Use Ship To/Sold To: This field is used to retrieve sales history based on the
customer’s Ship-To or Sold-To address. It is used with the customer range to extract
the sales history for the forecast calculation. Sales history is maintained on the
system by the customer’s sold-to and ship-to addresses. The user must determine
how to use the customers sales history to calculate the forecast. When Ship-To and
customer regions are defined to select the sales history, only permanent Ship-To
addresses in the address master file are referenced for customers who are in the
selected region range. The default is Sold-To.
Customer: The customer range combined with Use Ship-To/Sold-To determines
which sales history is used in the forecast calculation. The address code uniquely
identifies a specific customer. It identifies their name and address, and the
customer data and credit information applicable to this name and address.
List Type: Identifies a subset of customers by the type of address the List Type
code represents. Only the sales history for customers with this type of address
is analyzed in the forecast calculation. (List Type is used to differentiate addresses. It’s
useful when selling to many divisions of a large conglomerate. Assign all addresses to the
same List type).
Region: An optional code classifying customers by region. Region selects a
group of customers whose sales history is to be analyzed by the forecast
calculation. The forecast calculation sums all sales history for each item’s sold-
to/ship-to customers within the selected region range.
Use Simulation Criteria Inquiry to view the Forecast ID template.
                       Processing Forecast Simulation




This function calculates a forecast. Sales history data is analyzed to predict
sales quantities for the year specified. The calculation requires a criteria
template. You can use the criteria template defined in 22.7.1, Simulation
Criteria Maintenance and stored in the system by forecast ID, or you can
define the template here.
•   You may want to calculate a rolling forecast each month for high-cost
    items to minimize expensive over and under production.
•   Recalculate when sales data changes.
•   Items must exist in the item master file.
•   When there is insufficient history to create a valid forecast, a forecasting
    master record is not created and the item is printed out as “Items with
    Insufficient History to Generate Forecast”.
•   Negative results are shown as zeroes.
•   When the calculation is run, the criteria template is updated with the
    values used to produce the forecast.
•   If an annual forecast was calculated (forecast year > ending year),
    previously created records with the same forecast ID are deleted and
    replaced with the newly created records. If a rolling forecast was
    calculated (the current year is both the forecast and ending year),
    previously created records with the same forecast ID, year, and item are
    updated. The criteria template is frozen and can be further modified only
    when doing another calculation using Simulated Forecast Calculation.
•   The Forecast Method cannot be 00. Methods 01- 06 are predefined.
    Methods 07- 50 are reserved. Incorporate your own forecasting method
    using Forecast Method Maintenance with method numbers 51- 99.
Use this function to manually create or manipulate forecast results. Forecast
results are the detail forecast records produced by the forecast calculation or
uploaded through the CIM interface. Results may require manipulation to be more
reflective of future market demand, especially when forecast results are based on
unprecedented sales.
Retrieve forecast records by forecast ID and year. When the forecast record is for
a product group, define a specific item number. Forecast records are displayed
in three columns: month, original quantity and adjusted quantity. Initially, the
adjusted quantity is equal to the original. You change the adjusted quantities.
Whenever there is reason to suspect future demand will not be similar to sales
history, make adjustments manually to the forecast records. Make adjustments
required prior to copying forecast to MRP.
When loading forecast results with CIM interface or creating results here, you must
provide a forecast ID, year and item number. The system sets the forecast method
to 00 and creates a criteria template. The template is stored under that same
forecast ID. Alterations to forecast records are permanent. Run the calculation
again to reproduce the original forecast. Before manipulating forecast records,
archive the original forecast or copy the record to another forecast ID. This
function should be password-controlled.
Use Detail Forecast Inquiry to view manually created or manipulated forecast
results.
Use Detail Forecast Report to view manually created or manipulated forecast
results.
Use this function to replace or combine criteria templates and forecast
records. You cannot separate combined forecast records. During a
Combine or Replace, the original target forecast record and criteria
template is overwritten. The source record is not altered. Forecast
records are copied only in terms of units. Source and target items need
identical units of measure or a unit of measure conversion. This is done
for all items in the source record. This function should be password
controlled.
If the Target Forecast ID has a Forecast Method other than “00”, you receive:
ERROR: Target Forecast ID must have forecast method 00.
If your Target Forecast was created in 22.7.5 Simulated Forecast Calculation,
the system would have changed the Method to some value other than zero.
Therefore, the Target Forecast could not have been created in 22.7.5 Simulation
Forecast Calculation. It must be created in 22.7.7 Detail Forecast Maintenance
prior to this Replace/Combine transaction or the Target Forecast can be
created here.
You may increase or decrease the resulting record. This is useful when you
have prior knowledge of unprecedented demand, such as a future sales
promotion.
* Base - You enter the percentage by which the forecast quantity is
increased or decreased. A negative percentage means the forecast quantity
is decreased.
   Combining a source quantity of 100 with a target quantity of 200 and
   adding a 10% base results in 310 units per month.
   [source quantity 100 + target quantity 200 + 10% of source quantity 100 = 310]
Scale - You enter the percentage by which the forecast quantity is scaled.
Scale cannot be negative.

    Combining a source quantity of 100 with a target quantity of 200 and adding a
    10% scale results in 210 units per month.
    [target quantity 200 + 10% of source quantity 100 = 210]
Trend - You enter the percentage by which the forecast quantity is increased
or decreased. A negative percentage means the forecast quantity is
decreased.

 Combining a source quantity of 100 with a target quantity of 200 and adding a
10% trend results in the above.
[1st month - source quantity 100 + target quantity 200 + 10% of source quantity = 310]
[2nd month - source quantity 100 + target quantity 200 + 20% of source quantity = 320]
Use this function to create forecast for a new item based on historical
sales of another Item. This function is similar to Simulation To
Simulation Copy but here forecast records are copied for a single item.
Roll detail forecast records into the summary forecast file accessed by MRP.
The summary forecast file provides MRP with demand to calculate gross
requirements, project quantity on-hand and produce action messages.
You may create new summary forecast files, replace old files or combine with
existing files. The forecast detail records (created in the Simulation Forecast
Module) are identified by forecast ID, year and item. The summary forecast
files (22.1 Forecast Maintenance) are identified by item, site and year.
MRP is site specific, so select a summarized site for updating. The
summarized site may be different from the order line site specified on the
criteria template. Forecast records are updated to the summary forecast files
only in terms of units.
There are three loading methods to de-aggregate the monthly forecasts
records into the weekly summary forecast files:
1. Loading Method 1 - Auto spread: (default method) Monthly forecast is
broken into daily averages for the whole month. Then the system sums the
daily averages up into weekly buckets starting on the start date entered.
2. Loading Method 2 - Load Last Week: Monthly forecast is loaded into a
week that contains the last Monday of that month.
3. Loading Method 3 - Load First Week: Monthly forecast is loaded into a
week that contains the first Monday of that month.
Example Loading Methods
                                                                           Modify
                                                                          Forecast



            Autospread

                                                             April 2000
                         Sun       Mon   Tues   Wed   Thur    Fri   Sat
Load First Week                                                      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



                  Load Last Week
Loading Method selection depends on when you ship. For some
business environments, it usually ships a big bulk of orders at close of
the end of the month. It is more reasonable to use loading method 2 that
loads whole month forecast sales into a single week. It is the same for
loading into the first week. But generally it is more accurate to use auto
spread method loading the calculated forecast results into summarized
forecast files fcs_sum which is accessed by MRP. It is also important to
know no matter which method is used, you can go into Forecast
Worksheet Maintenance (22.2) to modify these figures.
Start: The default is today or the first day of the forecast year, whichever is
latter. It’s the earliest date the system considers in calculating and loading
weekly forecast quantities. The start date and forecast year are used by the
system to determine the week the forecast quantities are to be loaded. If the
start date is Wednesday, the system calculates the daily average from the
simulated forecast quantity for the calendar month in which the start date falls.
It accumulates five days of forecast into the weekly bucket that begins on
 Monday of that week.
Start (continued): The system does not process a forecast load if the
beginning date of the forecast year and the start date are not in the active
forecast period. It does not stop you from entering a past forecast year
and a currently active start date. You can also enter a future forecast year
and start date in a past period. However, if you enter a past forecast year,
no forecast quantities load; if you enter a future forecast year, the system
loads all buckets for the forecast year starting on the first day of that year.
Replace/Combine: It you select “Replace,” the five-day forecast
quantity places the entire quantity that was previously in the bucket.
“Combine” adds it to the existing forecast quantity.
Monthly data from Forecast Summary is loaded into weekly buckets for MRP.
This was loaded using Method 1 - Autospread
Monthly data from Forecast Summary is loaded into weekly buckets for MRP.
This was loaded using Method 2 - Load Last Week.
Monthly data from Forecast Summary is loaded into weekly buckets for MRP.
This was loaded using Method 3 - Load First Week
You may write your own forecast method and incorporate it into
MFG/PRO. Your user-supplied PROGRESS program must be
named “ffcalcXX,p”. “XX” is the forecast method and must be
between 51 - 99.
Good forecasting technique is a skill acquired by experience.
MFG/PRO allows you to incorporate your own expertise by adding
new PROGRESS procedures. The PROGRESS procedure must be
written and accessible to MFG/PRO before the method number is
defined with this function.
Include the following files in your PROGRESS ffcalvar.I and ffvar.I.
Use an array named calc[1-60] as the historical data input. Use an
array named fcast[1-12] for the calculated output forecast record.
Use method numbers 51- 99.
Method 00 indicates the criteria template or forecast record was
copied, entered manually or loaded through CIM interface.
Methods 01- 06 are predefined. Methods 07- 50 are reserved for
the system.
User factors (1) and (2) are reserved for your forecast methods.
Required if you want to use other statistical methods in a forecast
calculation. Forecast method tells the system the PROGRESS
procedure to use when calculating a forecast. Different procedures
employ different statistical methods.
Used to delete or archive forecast records into a file. Criteria templates are a
deleted or archived.
                         Summary
1. Create a Forecast Template in 22.7.1 Simulation Criteria Maintenance or
   22.7.5 Simulation Forecast Calculation.
2. Create a forecast in 22.7.5 Simulation Forecast Calculation.
3. Modify the forecast in 22.7.7 Detail Forecast Maintenance. (optional)
4. Combine or replace forecast records in either 22.7.11 Simulation To
    Simulation Copy or 22.7.12 Single Item Simulation Copy. (optional)
5. Roll detail forecast records into the summary forecast file (22.1) using
   22.7.13 Simulation To Summarized Forecast.
6. If you write your own statistical method, update the system with
   the Progress procedure to use in 22.7.17 User Forecast Method
   Maintenance. (optional)
7. Delete/Archive forecast records in 22.7.23 Detail Forecast
   Delete/Archive. (optional)
                              Demonstration
In order to demonstrate the kind of results that 22.7.5 Simulated Forecast
Calculation generates for each kind sales pattern, I’ve created Trend, Seasonal,
Horizontal, and Cyclical sales history. Graphs of each can be found on the
following page.
Sales History Patterns
                                     Define

       Trend             Seasonal   Forecast
                                     Criteria




New!




  Horizontal             Cyclical
Here are the values of the graphs on the previous slide. At the top
of each section, the type of pattern, the item number and the method
that will be used in calculating a forecast are indicated.
Here are the results of the first calculation. Since the pattern of sales history
for 1999 to 2001 for item “f20” was “cyclical”, Forecast Method “05” was used
 in the calculation.
Here are the results of the second calculation. Since the pattern of sales
history for 1999 to 2001 for item “f21” was “seasonal”, Forecast Method “04”
was used in the calculation. Forecast Method “05” could have been used as
well.
Here are the results of the third calculation. The pattern of sales history for
1999 to 2001 for item “f22” was “trend”. Forecast Method “03” was used in
the calculation. Forecast Methods “02”, “04” and “05” could have been used
as well. Methods “02” and “03” lag behind trend effects. In this example the
quantity of 2890 for January 2002 “lags” behind the quantity of 3800 for
December 2001.
Here are the results of the fourth calculation. Since the pattern of sales history f
1999 to 2001 for item “f23” was “horizontal”, Forecast Method “06” was used in
calculation.
                          Demonstration




The last demonstration in this presentation will be to show how the
last four fields of this program can be used to manipulate forecast
results.
Recall what each of these four fields can do:
Replace/Combine - When the flag is set to replace, any existing
data in the target forecast record from the start date forward is
deleted and replaced with the values of the source forecast record
When set to combine, the source forecast data is added to the
existing target forecast record.
Base - You enter the percentage by which the forecast quantity is
increased or decreased. A negative percentage means the
forecast quantity is decreased.
Scale - You enter the percentage by which the forecast quantity is
scaled. Scale cannot be negative.
Trend - You enter the percentage by which the forecast quantity is
increased or decreased. A negative percentage means the
forecast quantity is decreased.
Eleven items (f1 -f11) were created in 1.4.1 Item Master Maintenance.
Eleven Forecast Ids were created in 22.7.1 Simulation Criteria
Maintenance.
Eleven forecast were created in 22.7.7 Detail Forecast Maintenance.
The forecast were created for the year 2001.
A forecast quantity of 100/month was created for item “f1”.
A forecast quantity of 200/month was created for each of the remaining
10 items. 22.7.12 was run 10 times with item “f1” as the Source
Forecast ID.

TARGET CRITERIA
F2 COMBINE BASE +10%
F3 COMBINE BASE   -10%
F4 COMBINE SCALE +10%
F5 COMBINE TREND +10%
F6 COMBINE TREND -10%
F7 REPLACE BASE    +10%
F8 REPLACE BASE    -10%
F9 REPLACE SCALE +10%
F10 REPLACE TREND +10%
F11 REPLACE TREND -10%
Combining a source quantity of 100 with a target quantity of 200 and
adding a 10% base results in 310 units per month.
[source quantity 100 + target quantity 200 + 10% of source quantity 100 =
Combining a source quantity of 100 with a target quantity of 200 and
subtracting a 10% base results in 290 units per month.
[source quantity 100 + target quantity 200 - 10% of source quantity 100 = 290]
Combining a source quantity of 100 with a target quantity of 200 and adding a
10% scale results in 210 units per month.
[target quantity 200 + 10% of source quantity 100 = 210]
Combining a source quantity of 100 with a target quantity of 200 and adding a
10% trend results in the above.
[1st month - source quantity 100 + target quantity 200 + 10% of source quantity =
310]
[2nd month - source quantity 100 + target quantity 200 + 20% of source quantity =
320]
Combining a source quantity of 100 with a target quantity of 200 and subtracting a
10% trend results in the above.
[1st month - source quantity 100 + target quantity 200 - 10% of source quantity =
290]
[2nd month - source quantity 100 + target quantity 200 - 20% of source quantity =
280]
Replacing a target quantity of 200 with a source quantity of 100 and adding a
10% base results in 110 units per month.
[source quantity 100 + 10% of source quantity = 110]
Replacing a target quantity of 200 with a source quantity of 100 and subtractin
10% base results in 90 units per month.
[source quantity 100 - 10% of source quantity = 90]
Replacing a target quantity of 200 with a source quantity of 100 and adding a
10% scale results in 10 units per month.
[source quantity 100 x 0.1 = 10]
Replacing a target quantity of 200 with a source quantity of 100 and adding a
10% trend results in the above.
[1st month - source quantity 100 + 10% of source quantity = 110]
[2nd month - source quantity 100 + 20% of source quantity = 120]
Replacing a target quantity of 200 with a source quantity of 100 and
subtracting a
10% trend results in the above.
[1st month - source quantity 100 - 10% of source quantity = 90]
[2nd month - source quantity 100 - 20% of source quantity = 80]
QUESTIONS

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:10
posted:7/27/2012
language:English
pages:75