Forecasting Part 2

Forecasting – Part Two

Introduction

Having covered the basic elements of using the Demand Forecasting module in
MerciaLincs PC, the training can now move onto more advanced features of the
functionality in this module.

This section illustrates how to review forecasts applying the various techniques in the
most efficient way.

Block 2 Forecasting Part 2.doc Release 14 Page 1

Forecasting Part 2
Core Concepts

 Flexible selections, where a number of criteria are used to select items from the
full range of products to arrive at the final list of items.

 Archive lists, which are system generated lists that specifically report on whether
items and SKUs are flagged on the database as active or inactive.

 Sorting lists for convenience, so that the list can be reviewed in a pre-determined
logical order, based upon a wide range of conditions.

 Applying criteria to selections of items to filter down to a more specific range.

Filtering can be based upon a range of criteria.

 Saving Item and SKU lists for future use.

 Grouping, to aggregate items for review or action as a group using Edit graph.

 Expand the item selection options, by selecting ranges or groups of lists or items
and using different types of selection criteria.

 Weighting the information, to relate the view to values, weights, volumes or other
meaningful units held within the database.

 Forecast Summary and Reporting.

 Using ‘Vertical’ functions to review and edit forecasts, to allow

increased efficiency of reviewing and editing (if appropriate).

 Assessing the changes in statistical models before committing to

saving.

 Scaling – copying models from one SKU to another.

 Futuremodel – creating models from a manually entered forecast.

Block 2 Forecasting Part 2.doc Release 14 Page 2

 Forecasting Part 2

(a) Flexible selections

Selecting items in MerciaLincs PC can be achieved in a number of ways. Retrieval of

single or ranges of items can be based upon inclusion in Lists, selection of File index
numbers and selection of part numbers defined on the database.

(i) List Types

The most common list TYPES recognised by MerciaLincs PC are My Lists (MY),
Exception Lists (EL), CAtegory Lists (CA) and Program Lists (PL). The different
lists relate to the method of creation or the module and environment during creation.
The differentiating features of the list types are the prefix, which are indicated in bold
type and in brackets above and the last two characters of the user id number (e.g. 55)
which follow the prefix. See the Database Administration training notes for a more
detailed explanation of list types.

(ii) Retrieving Lists

To load a number of items into Demand Forecasting there are two methods available.
The first is to use the “Select items” option from the “File” menu. The second is to
use the Select Items icon as illustrated below.

A window is presented to allow the list type to be keyed in. Typing in the prefix of
the list type will allow all lists with the same prefix to be reviewed and selected. In
the example below the prefix EL is used.

Block 2 Forecasting Part 2.doc Release 14 Page 3

Forecasting Part 2
MerciaLincs will display all current lists with the selected prefix.

MerciaLincs will then load the items within the selected EL prefix lists, displaying a
progress bar whilst the operation is being executed.

It is possible to select from ALL available lists by clicking the drop down box in the
Select items window

Clicking on ‘CHOOSE FROM ALL LISTS’ will allow the appropriate list to be
selected.

Block 2 Forecasting Part 2.doc Release 14 Page 4

 Forecasting Part 2

(iii) Retrieving Items

Retrieving items using file index numbers can be carried out by entering file index
numbers (with spaces or commas between each one, if more than one item is
required) into the item selection window.

Alternatively, if the required index numbers are sequential as in the above example
the following convention can be used:

Block 2 Forecasting Part 2.doc Release 14 Page 5

Forecasting Part 2
Also this type of item selection by index number can be linked using a comma:

(iv) Retrieving product codes

To retrieve items by the product code (for example code 100-77D), type an
apostrophe (‘) followed by the product code into the select items window, as
illustrated below:

This type of retrieval can also be combined with ‘wild card’ symbols, * and ? to select
items matching the wildcard and product code structure used. The example below
uses a search for all product codes with a leading ‘100-‘ and any other characters by
keying ‘100-* into the select items field. The * symbol accepts any text following the
specified prefix.

Block 2 Forecasting Part 2.doc Release 14 Page 6

 Forecasting Part 2

The above would retrieve 100-77D, 100-77A, 100-99, 100-9 etc but would not
retrieve 1008, 1009, etc.

The selection can be made more specific by using the ? to substitute for any character
in a given position in the product code, as below.

The above would retrieve all items that had a seven character part number,
commencing with 100- and ending with D. Any letters, numbers or spaces would be
accepted in positions 5 and 6.

Combining options can extend this type of retrieval, using a comma to separate each
part number range entered.

E.g. entering ‘??????A,’??????Z would retrieve all the items with a seven character
product code ending with A or Z.

(v) Combining methods of item retrieval

The following screen shows how each of the item selection options can be combined
by use of a comma to separate each item selection criteria.

The above will offer a selection from all of the lists with prefix CA, followed by a
selection from all lists with prefix EL, along with all product codes that start with the
three characters 100 plus the first ten items on the database.

To summarise, any numbers or letters (including * and ?) prefixed with an apostrophe

will be treated as product codes, any numbers entered will be treated as index

Block 2 Forecasting Part 2.doc Release 14 Page 7

 Forecasting Part 2
numbers unless prefixed with an apostrophe, any letters entered will be treated as the
start of list names

(b) Retrieving more than one List

It is possible to select more than one list by using Shift/Ctrl click or click and drag
techniques. Previously, a number of items were simultaneously selected by choosing
an individual list. This section now advances the list retrieval concept a stage further
to illustrate the principles of combining item retrieval methods.

For selecting multiple lists, click the left-hand mouse button on the first list required,
position the mouse over the next list required, and press and hold the Control key in
conjunction with the left mouse button. To select a range of lists, select the first item
using the left mouse button, then position the cursor over the end of the range and
hold the shift key in conjunction with the left mouse button to select the end of the
range.

(i) List Types

List Categories have been mentioned previously, where the list prefixes are used to
identify the environment and type of list created.

Within the list management functionality in MerciaLincs PC, the system also
differentiates between two types of lists, Items and Stock Keeping Units (SKUs).

The two types of list reflects the fact that some item properties (or attributes) relate
specifically to the item code, (e.g. brand) and other properties are unique to an item at
a particular account, at the SKU level (e.g. forecasted annual usage at a node).

Item lists share a common reference at the item level and so are usually based upon
the contents of a Master Record field. This can be achieved by using Filters or
Sorting from a larger list or from all items. They may also be created during the
execution of a program elsewhere in MerciaLincs PC, to trap items with exception
conditions.

SKU lists differ in that they are items linked by features that are not item specific, but
account specific. They are usually created based on the contents of an Account
Master Record field and may also be created during program execution to trap
exceptions.

As an example, consider a database with 5 accounts, Total (Source) and North, South,
East, West (Nodes).

MY55ITEMLIST contains 3 items, index numbers 1,2,3. These items are considered
active at all accounts.

MY55SKULIST contains 6 SKUs, these are index numbers 4 and 5 (two items) but at
each of three accounts, i.e.: index 4 Total, index 4 North, index 4 South, index 5
Total, index 5 East, index 5 West so this list contains 6 SKUs.

Block 2 Forecasting Part 2.doc Release 14 Page 8

 Forecasting Part 2

MY55ITEMLIST

Account Total North South East West

Item 1 X X X X X

Item 2 X X X X X

Item 3 X X X X X

MY55SKULIST

Account Total North South East West

Item 4 X X X

Item 5 X X X

If MY55ITEMLIST is selected, there will be 15 SKUs available for use within the
current session. This is because each of the 3 items will be loaded and all 5 of the
available accounts will be used, since no limit was specified for which accounts are to
be worked on.

If the MY55SKULIST is selected, there will only be the 6 item/account combinations

available within the current session. This is because the list explicitly defines which
item/account combinations are valid for the condition/criterion used to build the list.

It should be noted that if an item and a SKU list were retrieved together, all accounts
would be available for all items.

(ii) Archive lists

As referred to previously, MerciaLincs produces lists automatically sometimes based

upon exceptions and sometimes based upon the values and criteria held on items in
the database. Archive lists are another category of list in the system, which are
system generated and maintained. They are automatically produced each time period
end is run. The archive lists can also be regenerated at any point in time if changes
are made to the active item and active SKU flags.

The archive lists produced by MerciaLincs are:

 ALLACTIVEITEMS – reporting all items with a 1 in MR [49]

 ALLDEADITEMS – reporting all items with a 0 in MR [49]
 ALLACTIVESKUS – reporting all items with a 1 in AMR [;63]
 ALLDEADSKUS – reporting all items with a 0 in AMR [;63]
The values in these fields should be routinely maintained.

Block 2 Forecasting Part 2.doc Release 14 Page 9

 Forecasting Part 2

The archive lists are available when using select items by either typing the name into
the dialogue box or by typing AL and then selecting the particular list required from
those available.

(iii) Sorting the Current Product Selection

Once the relevant listed items have been selected it can be extremely useful and
convenient to be able to sort the list into a more logical order.

When products are first selected in the Demand Forecasting module, they will appear
in file index order, as they are retrieved from the database as shown below.

Sorting the selected products is based on Master Record Fields. A sort may be applied
before or after item selection.
In this example the current selection of products will be sorted into ascending
alphanumeric product code order.

Select “File” and “Sort items” from the Main Menu or click on the Sort icon,

A prompt will then appear to verify that the currently selected items are to be sorted.

The sort key must now be selected and the sort order specified.

Block 2 Forecasting Part 2.doc Release 14 Page 10

 Forecasting Part 2

The default selection in the Sort fields dialogue box is Product Code and Sort order
“Ascending”. Accept these by clicking “OK”.

To view the results of the sort use the item drop down box.

To turn off the Sort, click the Sort icon and select “No” when prompted for “Sort
current selection?” and then “Cancel” in the “Sort fields” dialogue box.

Block 2 Forecasting Part 2.doc Release 14 Page 11

Forecasting Part 2

(c) Filtering Product Selections

Applying filter criteria to reduce the current selection is possible in MerciaLincs by

using the Filter option. This option allows the user to reduce the number of items to
review in the item box by applying a wide range of criteria based upon data entered or
held on the database.

We can filter before or after retrieving products. In this example we will filter the
current selection of items and retrieve only those items whose Standard Cost in
Master Record 5 is greater than 1.00. Select “File” and “Filter” from the Main Menu
or click on the Filter icon .

MerciaLincs then requires the user to define the information upon which the filter is
to be based. Options are selected from the following list:

For example, for a “Master Record” filter, click the radio button and “OK”.

From the list of available fields, click to highlight the required field and click “OK”.
In this example, select Std cost.

Block 2 Forecasting Part 2.doc Release 14 Page 12

 Forecasting Part 2

The options are then open to further choices of conditional criteria, as shown above.
Select the required test to perform by clicking the radio button. In this case the ‘Is
greater than’ operator has been selected.

At the Available fields dialogue box, enter the appropriate cut off point in the Value
box.

A further option is then presented, to make additional conditional criteria for the
selection or to select “End of tests”. In this example, the test criteria entry is now
finished.

Block 2 Forecasting Part 2.doc Release 14 Page 13

Forecasting Part 2

A final check is then made, to confirm the selections before carrying out the specified
filter.

After confirmation, a progress window appears detailing the number of items

processed. The number of items satisfying the filter and percentage processed figures
are both indicated.

The status line indicates the filter used to test items.

The filter applied remains active until turned off. The filter can be de-activated by
using the filter icon, or the Filter option from the File menu.

If a filter is de-activated the item selection remains the same; i.e. products in the items
window will only change if a new group of products is selected.

If a new group of items is selected with a filter still active, only those items that meet
the filter test will be retrieved.

(d) Filtering on Description Records

When filtering on descriptions the following four filters are available. These are
apparent on the Test to perform options menu. The options are:

 Is equal to
 Is not equal to
 String search (case insensitive)
 String search (case sensitive)

Block 2 Forecasting Part 2.doc Release 14 Page 14

 Forecasting Part 2

The first two available options filter on the value of the description field. These are
most often used when selecting items based on a relatively short field length, such as
ABC class or Unit of Measure. The third and fourth ‘string’ options filter the contents
of a description field, regardless of the position of the specified string of characters.

To filter a selection for items that include ‘paracetamol’ in the item description field,
apply a filter based upon ‘Description fields and use a ‘case insensitive’ filter. Type
paracetamol into the value field and run the filter as illustrated.

The system displays all the description fields held on the database.

Select Description and click “OK”.

Choose the case insensitive radio button and click OK.

Block 2 Forecasting Part 2.doc Release 14 Page 15

Forecasting Part 2

In the Value box type paracetamol and click “OK”.

Click on the “End of tests” radio button.

Block 2 Forecasting Part 2.doc Release 14 Page 16

 Forecasting Part 2

Select Yes, to Filter current product selection.

This filter will trap all items with the word paracetamol or Paracetamol or
PARACETAMOL anywhere in the item description field.

When completed, click on the Item drop down box to examine the results of the filter.
There will only be items with ‘paracetamol’ in the description.

If the case sensitive option had been selected then the filter would only have trapped
items with the word paracetamol (all in lower case) in the description and would have
discarded items with Paracetamol or PARACETAMOL, etc.

If a previous filter or partial selection of items had been made, the filter would only
have been based upon items currently selected.

(e) Creating Item Lists

Once any range of items has been selected and/or filtered/sorted, it is possible to store
the currently selected items as a personal list for future use. This can be achieved by
using two different methods. The first is by using the “File” option on the Forecast
Workbench menu and then the “Save item list” from the file menu. The second is by
using the Save Item List icon from the main Demand Forecasting module as
illustrated overleaf.

Enter the name and type of list created.

Block 2 Forecasting Part 2.doc Release 14 Page 17

Forecasting Part 2

This action creates a ‘MY’ list using the name entered in the “List name” field as a
suffix.

This option saves the currently selected items only as a list of file index numbers and
hence saves no information about which accounts have been selected. This means
that whenever this list is retrieved, all accounts for each item will be loaded.

(f) Creating SKU Lists

Creating SKU lists can be especially helpful when used in conjunction with filters to
produce a focussed view of a very specific range of item/account combinations. The
choice of SKU lists allows the currently selected items to be saved at account level.
For example, the current selection may have relevant information stored for a specific
account.

It is important to maintain the accuracy of the lists stored on the database, to ensure
that up to date information is being used and also to ensure that the size of the
database is kept to the most appropriate level. This subject is covered in detail in the
Database Administration module, but consideration should be given too the relevance
and accuracy of the lists being created.

(g) Basic Grouping

For reviewing details of history and forecast and also for editing, MerciaLincs
provides a facility to aggregate the selection of a number of items. This is very
simply achieved in the Demand Forecasting module.

Once an appropriate group of items is selected (in the illustration, CA55LINCTUS is

used), all that needs to be done is to check the Group flag, to the extreme right of the
item drop down box, as illustrated below.

Block 2 Forecasting Part 2.doc Release 14 Page 18

 Forecasting Part 2

In the example below the CA55LINCTUS group of 6 items is viewed at the TOTAL
account and is weighted by Standard Cost (Master Record 5) and viewed in Plot.

The group can also be viewed in the relevant Master Record or Account Master
Record weighting at any particular account.

For example, to view the South account information at standard cost, the account
must be selected before the group check box is flagged. The system will not allow
navigation across accounts when the group check box is flagged. The weighting for
the selected account can be carried out before or after grouping the items together.

Block 2 Forecasting Part 2.doc Release 14 Page 19

Forecasting Part 2

Weighting a group of items can be particularly useful for providing top level
summaries of the forecast. For example, all items can be selected, grouped and
weighted by any MR or AMR field to provide an overall view of the weighted
forecast.

In this example, the forecast for all items at the source account (Total) is grouped and
weighted by Selling Price. This provides a forecast of sales revenue for all items
defined in MerciaLincs as well as relating the historically posted information in the
same units to portray the overall trend.

The resultant projection is then available graphically or in tabular form.

Block 2 Forecasting Part 2.doc Release 14 Page 20

 Forecasting Part 2

The grouping and weighting functions are also available in Edit Graph. The
weighting of information is carried out using the weighting field to the right of the
account window. Grouping can be carried out after selecting the items, before
entering Edit Graph or using the icon available in Edit Graph, as illustrated below.

Block 2 Forecasting Part 2.doc Release 14 Page 21

Forecasting Part 2
Summarising Forecast Information

Reporting on the item history and forecast information within MerciaLincs has been
reviewed in previous use of the Item History and Forecast option, from the Show
menu within the Demand Forecasting module.

Generating a hardcopy list of this type of information for a range of items and
accounts is more conveniently generated using the four year report.

(a) Four Year Report

The Four year report is accessed in the Demand Forecasting module either from the
Reports option on the toolbar or by using the Four year icon, as illustrated below:

This action can only be carried out with items already selected. The user is then
prompted to confirm the current item selection or to make a new selection, via the
standard ‘item selection’ dialogue box.

Selecting “Yes” advances the process to the next stage. Selecting “No” takes the user
through the select item process. A prompt is generated to indicate which accounts
are to be incorporated into the Four year report. The usual methods of selecting
accounts are available.

Block 2 Forecasting Part 2.doc Release 14 Page 22

 Forecasting Part 2

Once the accounts have been selected and confirmed the select the options from the
Run report window, as illustrated.

The weighting fields allow the display to be weighted as reviewed previously. The
forecast options allow the various components of the forecast to be included in the
information displayed. Printing options allow the output to be presented to the screen,
the clipboard or direct to a printer.

Block 2 Forecasting Part 2.doc Release 14 Page 23

Forecasting Part 2
Grouping allows the user to present the items and accounts linked by a relevant
selection, for example, Buyer. The History option allows either Demand or Sales
history to be shown.

Details can also be selected, to illustrate or omit various text and the scaling option
allows the user to present the output in a condensed numerical format by reducing the
displayed unit value. The default values for details are to print product and account
details. The Print ‘breakpoint totals’ option will produce a summary against the
groupings used for the accounts selected. The Print totals across accounts option
produces grand totals for each account selected at the bottom of the report.

When scaling is used the scaling factor selected will be displayed on the output, but
the effects of rounding should be considered when using the scaling option.

At this point it is possible to modify the Font and Margins for the output before
selecting OK. At any time, Quit closes the Four year report and returns to the start
point.

A report similar to the one below is presented:

Block 2 Forecasting Part 2.doc Release 14 Page 24

 Forecasting Part 2

When displayed to the screen the information can be reviewed before printing. The
print option directs the output directly to the printer with no ‘print preview’
functionality. To print the screen output, simply choose “Print” form the “File”
menu.

Once completed the Four year report closes and returns to the main Demand
Forecasting module screen.

(b) Summarising Forecast - Account Structure

To recap, history is only stored at nodes and not at aggregates or sources. The effects
of this feature upon the forecasts generated by MerciaLincs will now be explored.

If statistical forecasts have been calculated in multi-level account structures it is

important to note that the sum of node forecasts may not equal the source forecast, if
the source has had Leastsquares run to fit a forecast model.

This example provides insight into the difference between a statistical forecast at a
source compared to the aggregation of statistical forecasts produced at several nodes.

For reference, the plots of the 3 node accounts follow the plot for the Total account.

This illustration clearly shows the presence of seasonal effects in the statistical model.

Block 2 Forecasting Part 2.doc Release 14 Page 25

Forecasting Part 2

The statistical model at the West account also illustrates the presence of seasonality in
the history.

Block 2 Forecasting Part 2.doc Release 14 Page 26

 Forecasting Part 2

The forecast at the East account on the previous page illustrates that the analysis of
the history did not reveal statistically significant seasonality and hence, a simple two
term model with level and trend only is produced.

The North account provides another example where no statistically significant

seasonality has been detected, again resulting in a simple “two term” model.

To illustrate the comparison of the Total account versus the sum of the nodes a table
which uses the details from each item for 6 periods of history and 6 periods of
forecast has been produced.

Account 11/04 12/04 1/05 2/05 3/05 4/05 5/05 6/05 7/05 8/05 9/05 10/05
North 151 26 207 106 111 150 107 107 107 107 107 108
East 114 90 119 46 92 115 83 83 83 83 83 83
West 66 85 319 25 73 65 87 36 70 61 69 47
Sub 331 201 645 177 276 330 277 226 260 251 259 238
Total 331 201 645 177 276 330 314 226 245 120 258 250
Diff 0 0 0 0 0 0 37 0 -15 -131 -1 12

Throughout the 6 months of history there is an exact match in the sum of the nodes
versus the total account. This is because no history is stored at aggregates (including
the source).

From the period 5/05, the sum of the nodes differs from the statistical model produced
at the Total account. Reviewing the graphs above, it may be observed that the two
accounts producing a two term model (the first two in the table) produce no seasonal
patterns, whereas the total illustrates seasonality due to a more statistically complete
sample of history. It should also be noted that the deviation of the sum of the nodes
from the Total is in both directions and in the most extreme period differs by nearly
110% of the Total.

Block 2 Forecasting Part 2.doc Release 14 Page 27

 Forecasting Part 2
Using Advanced Functions

(a) Leastsquares

In the Demand Forecasting module there is a facility to run Leastsquares on a

selection of items. This feature can be especially useful for generating new statistical
forecast models for new items or for running Leastsquares to update statistical
forecast models for a selected range of items.

The Demand Forecasting module has an option to build and manage lists that can
also be processed by Leastsquares selectively as a group. It is now possible to treat
new items and refitting models for existing items in the same way in one go. For
example, several of the period end exception lists (see Training Block 5 for details)
can be selected and processed together by Leastsquares, to generate new statistical
forecast models.

It must be remembered that Leastsquares directly amends the statistical forecast

for the selected items and therefore it is expected that the user will follow good
database practise and make a backup of the database before using this option.

The facility exists in Demand forecasting module in the Edit option.

Once the appropriate items have been selected run the Leastsquares function.

Block 2 Forecasting Part 2.doc Release 14 Page 28

 Forecasting Part 2

As stated, it is mandatory to confirm that a backup has been taken.

The options concerning the history length are to not recalculate the history length, to
recalculate for all items, irrespective of the values used in the history length or to
recalculate only for items with a default history length of zero.

The other option allows the removal of all market intelligence from the forecast when
running LeastSquares.

Once the fields are appropriately checked, the OK button then begins the processing.
A progress indicator appears while the program is running. Selecting Cancel will end
the programme if required.

Once Leastsquares has been successfully run a range of exception lists may be
created. These lists appear in the Save Lists dialogue box.

In this example Leastsquares has created 6 lists as shown, most containing a number
of SKUs, but only 3 lists contain more that 1 item. Each of the lists created by
Leastsquares is a new list.

Block 2 Forecasting Part 2.doc Release 14 Page 29

Forecasting Part 2
If any of the lists already existed, the column on the left-hand side of the display
would contain an asterix (*) and the user would have to rename the list or replace the
old one.

The error conditions that can be generated by Leastsquares are detailed in the module
entitled “Block 1 - Forecasting Part 1”.

(b) ‘What-If’ Model Fit Tests

To enable some quantitative assessment of the change made whilst investigating

forecast changes, MerciaLincs PC includes a feature to compare the current statistical
model with the one on file. This information guides the user towards making changes
that improve the forecast.

The Model Information window portrays the following key statistical information:

 Standard Deviation
 Standard Deviation / Standard Deviation from the Variance Law
 Standard Deviation / Level
 Annual Usage

The SD/Exp SD ratio indicates whether the current model is better than (<1) average
(1) or worse than (>1) average when compared to the other statistical models used to
compute the variance law values.

SD/Level values less than 0.5 indicate a relatively low volatility and hence a statistical
forecast that closely describes the historical demand. Values approaching and in
excess of 1 indicate a high level of volatility which does not accurately describe the
historical demand pattern.

The information window will include the values plotted for any selected range of
periods in the upper part of the window and also includes the values stored in the
database for the fields defined above in a lower window.

Each time a forecast simulation is carried out the values listed in the example overleaf
are recalculated to provide a comparison between the statistical model data on file and
the latest values.

Block 2 Forecasting Part 2.doc Release 14 Page 30

 Forecasting Part 2

In the following example the model has had the Amplitude per Standard Deviation
(AMR [;60]) changed to 3 and then Leastsquares has been run. The changes are
updated in the Model Information Window.

Block 2 Forecasting Part 2.doc Release 14 Page 31

Forecasting Part 2
The updated model illustrates an apparently much better model fit against history and
therefore shows the commensurate improvements in the statistical values in the
window.

(c) Scaling

If a new product is added to the database it will not be expected to have any demand
or sales history to support the statistical analysis carried out by Leastsquares. The
Scaling function is a tool for establishing a statistical model for such items.

Scaling allows the statistical model from an established item (the reference item) to be
copied and scaled (multiplied up or down from the reference item) to create a forecast
with appropriate level, trend and seasonality coefficients.

Scaling can be used two ways, the first being when a forecast is required with
immediate effect and the second when a forecast is required to support a future launch
date.

The program resides in the Edit Graph function of the Demand Forecasting module
and is operated as shown, from the Model menu or by using the Scaling icon.

Select the Scaling options, which allows a future date of effectivity to be used for the
new forecast and to scale seasonal patterns present in the reference item forecast.

Block 2 Forecasting Part 2.doc Release 14 Page 32

 Forecasting Part 2

The next step is to define details of the reference item and account, along with
options for the Plug level (anticipated average monthly demand) and the expected
trend. If the desired reference item is not present in the drop down box, using the
“Search” facility allows the user to key in the part number of the desired reference
product to be added to the drop down menu before selecting.

The “Display” option shows a view of the plot of the reference product with the
standard 3 years history and 2 years forecast to assist selection of an appropriate
reference product.

Block 2 Forecasting Part 2.doc Release 14 Page 33

Forecasting Part 2

To apply the scaled statistical forecast, click “OK”. Selecting “Cancel” returns the
user to the starting point.

It should be noted that when using Scaling the Mean Squared Error used to generate
this forecast is calculated from the Variance Law. For further information relating to
the calculation of the Variance Law, please refer to the module entitled “Block 7
Database Administration – Part 1”. Initial values for the safety stock and standard
deviation over the lead-time are also calculated by Scaling. The item is set to use fast
smoothing (MR[3]=2), so that the statistical model will be affected more by period
end processing when actual sales are posted to the database.

When using Scaling to generate a statistical forecast model for a future launch, the
Scaling option ‘Trend calculated as of introduction date’ should be selected after an
effective date has been previously applied.

When this option is selected, the plugged level entered by the user is applied as at the
introduction date and the trend will grow the level from then on. In the following
example taken from a ‘Calendar Months’ database, the introduction date of 1/10/05
was set and plugged level of 100 and trend of 1 were entered.

Block 2 Forecasting Part 2.doc Release 14 Page 34

 Forecasting Part 2

Without the “Trend calculated as of introduction date” selected, the plugged level of
100 is applied from the “now” date so that by the time the introduction date is
reached, the trend has already grown the forecast from 100 to 106 units.

Block 2 Forecasting Part 2.doc Release 14 Page 35

Forecasting Part 2

(d) Generating a Statistical Model from the Current Forecast – Futuremodel

The “Futuremodel” function within Edit Graph in the Demand Forecasting module
fits a statistical forecast model to a net forecast, that is the sum of the existing
statistical forecast plus market intelligence. It performs similarly to Leastsquares, but
uses the forecast values as its data points. This is used to establish a new level, trend
and seasonality patterns if these model terms can be detected. Hence, this program is
typically used to convert market intelligence forecasts to statistical forecasts.

There is no checking of outliers, significant changes in pattern, all time supply, lumpy
demand or significance of the seasonal pattern within Futuremodel.

Again, the Mean Squared Error and related fields are computed based on the variance
law and the product is moved onto fast smoothing (MR[3]=2). Futuremodel is
accessed via the Model option of the Edit Graph or by using the icon, as shown.

Prior to using Futuremodel to generate a forecast it is essential to load in the forecast

for each period that the system is to model. In the example below figures have been
added to a base model as Market Intelligence and it is these figures that are required
to be included in the statistical model.

Block 2 Forecasting Part 2.doc Release 14 Page 36

 Forecasting Part 2

When using Futuremodel the user is required to indicate the number of terms, the
number of periods and whether the current net forecast is to be preserved by applying
Market Intelligence to the statistical forecast.

Number of terms 13 is the maximum number of model terms. Using less than the
maximum number generally means the resulting model will not
map seasonality very well. (e.g. If only two terms are specified
then only a basic level and trend model will be fitted)

Periods to use The user must define how many periods of forecast to use when
calculating the model. Best results are obtained when one or two
full fiscal years are used.

In this example, the net forecast has been set to mimic the market intelligence, with 2
years data and 13 terms. When selecting OK MerciaLincs presents the user with an

Block 2 Forecasting Part 2.doc Release 14 Page 37

Forecasting Part 2
information box to confirm the outcome. Here the analysis has only been able to fit
12 terms to the model. Cancel takes the user to the starting point once again.

The result of running Futuremodel can be seen below:

The statistical forecast model is now clearly fitted out into the future and the net
forecast has been maintained with market intelligence.

(e) Refer History

This function provides an alternative to Scaling and Future model, for creating a
statistical forecast model for a new product.

The technique involves copying the demand history from one or more predecessor
items. This technique is often used in cases of product succession, when the new

Block 2 Forecasting Part 2.doc Release 14 Page 38

 Forecasting Part 2

product is an enhanced version of the old or when an existing product is to be

launched at another account and requires a separate forecast.

There are two different Refer History functions within the Demand Forecasting
module.

The function within the Editgraph module operates at a single account.

The function available from the Edit option of the main menu operates at all accounts
for the active product.

(f) Referring History for One Product at All Accounts

This option is held in the Main Demand Forecasting module window.

Select a product with no history in the Item drop down box and select “Edit” and
“Refer History” from the Main Menu. The Refer History dialogue box will appear
prompting for the reference product, (the item to be referred from). Choose an
appropriate reference product from the available list and click on the “Add>>” button
so that the selected item appears in the right hand “Refer From” box:

Block 2 Forecasting Part 2.doc Release 14 Page 39

Forecasting Part 2

Errors are corrected by clicking the “<<Remove” button to de-select.

In this example, history is being referred from a product that is sold in 100ml bottles
to one which is sold in 1000ml bottles, so the sales of the new size pack volume will
be based on sales of the 100ml pack multiplied by 0.1.

There is an option to define which history type to copy. Selecting Demand avoids
duplication of sales history. The selected history is copied and added to any existing
demand data found for the new item. The option to overwrite the target item history
avoids this potential addition. This can be useful, if a mistake has previously been
made referring the history from another item. Leastsquares can also be run to produce
a new statistical forecast model at all of the accounts involved.

The target product has now adopted the scaled amount of the reference product’s
history at all accounts and a statistical model has been created on this product at all
accounts. The reference item has not been affected.

(g) Referring History for One Product at One Account

It may not be relevant to refer history at all accounts. Within Edit Graph the ‘refer
history’ icon also allows history to be referred from one product to another but it only
operates for the current item/account selection.

It should be noted that the Edit Graph function of Refer History will not run for
source accounts, aggregates or for grouped items.

The first prompt using this function is illustrated overleaf.

Block 2 Forecasting Part 2.doc Release 14 Page 40

 Forecasting Part 2

Where this Refer History for Account box differs is in the two additional options,
“Search” and “Display”.

Search allows the user to select any additional items that were not selected when
Refer History was run. This function works in the same way as the search option in
scaling. When search is selected the following dialogue box is displayed.

The appropriate part number is required to proceed.

Display also works in the same way as previously illustrated in scaling.

The results of this type of history referral are the same as those previously described.

(h) Editing Dates

The dates relating to statistical forecast models can be set and maintained through a
number of routes. These dates can be maintained through:
 Edit Dates in the Edit option in the Demand Forecasting module
 Edit Graph using the traffic light icons on the toolbar
 Model Parameters within the Edit option of the Edit Graph menu

Block 2 Forecasting Part 2.doc Release 14 Page 41

Forecasting Part 2

To edit the dates for a single item, at a specific account the Model Parameters option
from the Edit menu of Edit Graph is illustrated.

Here the opportunity is presented to set the history length for the item, the
Discontinue date (the date when the forecasts are to cease) or Effectivity date (the
date when the forecasts are to commence) options. Note, that if a date in the middle
of a forecast period is used, the period will be considered to be active.

The date format required is DDMMYY, so 261005 represents 26th October 2005.
Once the field has been completed, “OK” will modify the forecast appropriately. If

Block 2 Forecasting Part 2.doc Release 14 Page 42

 Forecasting Part 2

no model exists, no action will be taken at this time but the date will remain filed for
the SKU selected.

Checking the input is purely a case of repeating the steps taken, which allows the
dates to be reviewed.

The results of a discontinue date of March 2006, for example, is shown below.

The use of the Traffic Light icon for effective or discontinue dates uses opposed
arrows as a cursor, which can be positioned at the appropriate time frame. Clicking
the cursor at the relevant point applies that date as the required start/end date.

To edit dates for a number of items the “Edit Dates” option must be used from the
Edit menu of the demand forecasting module with the relevant items previously
selected.

Block 2 Forecasting Part 2.doc Release 14 Page 43

Forecasting Part 2

A spreadsheet type matrix is then presented for controlling the dates for the current
item selection. In this view all accounts are accessible for editing.

The dates are changed by clicking the cursor in the relevant cell and keying in the date
format DDMMYY, as previously described.

The “Type” field allows the Effective date of the Discontinue date to be applied or
reviewed. This matrix also allows scrolling through the items, making the process
quick and effective. “OK” applies the changes to the forecasts and Cancel returns to
the Forecast Workbench.

(i) Deleting Text

MerciaLincs provides a number of methods for removing text from the database. For
processing a large number of items, with the relevant items previously selected, the
Delete Text option from the Text menu within the Demand Forecasting module is
illustrated.

This method allows ‘from’ and ‘to’ dates to be used for clearing text for the current
item selection.

Block 2 Forecasting Part 2.doc Release 14 Page 44

 Forecasting Part 2

To add or remove text for one item at a time the Edit Graph function of the Forecast
Workbench can be used. There are two methods from this option, as illustrated.

When using the icon, the cursor changes appearance, matching the Text icon. The
cursor must be positioned over the period with the text to be deleted before clicking
the mouse button. A prompt appears to confirm whether to add text.

Choosing “Yes” would continue the process of adding another text reference for this
period. The “No” option over existing text allows the current text to be edited via the
Text Entry dialogue box.

Block 2 Forecasting Part 2.doc Release 14 Page 45

Forecasting Part 2

If the text message has been imported from the Off-line Forecast Workbench, the
‘OFW Text Message’ box will be checked.

The “Drop” option will then remove this text. Alternatively, the text can be modified
and “OK” will save the updated entry.

If adding new text the Text Entry dialogue box will appear with the keyword and free
text boxes empty. The “Text” option on the Edit Graph menu works the same as
clicking the text flag icon.

(j) Vertical Editing of Master Records and Account Master Records

To illustrate this function an example has been used where a filter for Annual usage
less than 1 has been previously applied to all items on the database.

The target field for vertically editing (using the same value in all items) is the ‘Active
SKU’ field. With the relevant items selected, the Account Master Record option
should be selected from the Edit menu within Edit Graph.

Block 2 Forecasting Part 2.doc Release 14 Page 46

 Forecasting Part 2

The destination AMR field must then be selected. In this example AMR[;63] is to
receive the value of 0 for all items, setting the active SKU flag to an ‘inactive’ state.
To post the value for all the SKUs the “Group edit” check box is selected.

“OK” then posts the new value to the selected field.

Horizontal edits are performed without the Group edit check flag selected if different
values are to be used or if each item is to be reviewed individually.

The editing of Master Records is similarly performed.

Block 2 Forecasting Part 2.doc Release 14 Page 47