Académique Documents
Professionnel Documents
Culture Documents
To subscribe to ERPtips, go to
www.ERPtips.com/Subscribe.asp.
This article begins a two-part series on the fundamentals of Demand Planning (DP) design and
implementation. Steve has six full DP implementations under his belt, and he's got a lot of insight
into the best practices of successful DP installs. In his first article on the most commonly
implemented APO component, Steve defines the key DP concepts and guides readers through a
typical (and effective) DP design.
APO system and then passed to it is not required to backup Sales limitations for use in a produc-
BW. Significant levels of such History as this is already available tive environment:
reporting may affect performance. in another InfoCube. However,
In practice, we have found that every key figure entered by a user • You can only create one new
the community wanting such via a Planning Book will need sav- combination at a time.
reports is small and this has never ing. The need for this backup
been an issue. Also, to help pre- comes from the possibility that • There is no history main-
vent this issue, a modest selection someone might inadvertently de- tained. Therefore, if you ever
of data should only be selected for initialise the DP Planning Area— deliberately or inadvertently
reporting at one time. Figure 5 something that is known to have deleted all the combinations
shows an example of a BW Query happened to a live system. It in the POS, you would have to
using a Remote cube. should also be remembered that recreate the new combinations
to make any change to the Plan- all over again—if someone
Forecast cube ning Object Structures or Plan- had kept a record!
Using the forecast cube method, ning Areas would require the
the plan is periodically extracted deletion of all DP data in Live- The technique I have used in all
in its entirety to an InfoCube. One Cache first. past DP implementations is to
extract is performed at the end of have the planners maintain the
every planning cycle, which rep- New Combinations Cube new combinations on one or more
resents the agreed, frozen demand Most new Characteristic Value spreadsheets. These are then
plan. Additional characteristics Combinations (CVC) are gener- saved to .csv format and loaded
are used to stamp the values in the ated from actual new Sales His- into a special InfoCube. The
cube with the planning cycle iden- tory as it is received in APO. This structure of this cube should fol-
tifier. Once this extract has been allows established products to be low the same principles as
performed, then queries on this planned for the customers who described for the Sales History
cube have no further impact on take them. But how do you plan cube earlier—the main reason is
LiveCache or the APO system at for new products and/or new that we want to fool the system
all if a separate BW system is customers before the first sale is into treating this data as Sales
April/May 2004 Volume II Issue 2
used. This cube can then be com- made? To be able to do this in History. The fields in the spread-
bined with the sales history cube DP, an appropriate CVC must be sheet can actually be much fewer
as a multi-cube to allow forecast created. Although within trans- than you might think, as most can
accuracy reporting. As it can take action /sapapo/mc62 there is a be coded as constants in the trans-
several hours to complete the function for creating new plan- fer rules. Many characteristics are
extract, it is not suited to report- ning combinations, it has serious actually attributes of other char-
ing the live DP plan as
it changes during the
planning cycle. It is
SAPtipsJournal
Backup Cube
Although originally
recommended because
early versions of Live-
Cache did not log
updates for DP data, it
is still advisable to
periodically save all the
data in the DP Plan-
ning Area that is not
recoverable from other
sources. For example, Figure 6: Generating CVCs from a New Combinations Cube
acteristics and can also be looked The data can then be trans- acteristic to code the plan
up during the load. ferred to the Planning Area using date.
transaction /sapapo/tscube. For
Once the spreadsheet is loaded this system to work, you need to 5. Your Planning Area will need
to the cube, you can then use follow a few simple tips: to include the time character-
/sapapo/mc62 to generate the istic specified in the external
Sales History (see Figure 6). 1. You must include 9AVER- plans.
Note that the “from” and “to” SION as a characteristic, as
dates are hard coded to match /sapapo/tscube expects to see 6. If there is a time characteris-
the data in the cube, which in it and, unlike the others, it tic you aren’t supplied with
this case is given a fixed date of cannot be remapped. in the file, then don’t include
01.07.1999. it in the cube. If it is present,
2. /sapapo/tscube can load data then /sapapo/tscube will
External Plans at a detail level or at an expect it to be populated with
On all but one implementation I aggregate level. The system a value. Whereas if it is
have seen, it has been a require- uses the disaggregation rules absent, DP will use time dis-
ment to load plans into DP that specified in the planning area aggregation to calculate it.
had been defined on other sys- for the key figure if required. For example, if the file con-
tems. These ranged from forecasts Note that this also means that tains month and the Plan-
supplied by customers to sales you can leave those charac- ning Area has weeks (and
plans supplied from remote parts teristics not required un-val- months), then only include
of the globe by users without uated in the cube—or even months in the cube. During
access to SAP. Several things need exclude them from the cube the transfer to DP, the data
to be noted about these externally altogether. will be disaggregated to
supplied plans: weeks.
3. You may store plans from
• The characteristic values multiple sources potentially 7. The plans in the cube can be
used may be different from at different levels of aggrega- displayed and analysed using
April/May 2004 Volume II Issue 2