Skip to Content
Author's profile photo Former Member

Composite forecasting in APO: Roadmap to demand plan accuracy

Composite forecasting
in SAP APO:  A roadmap approach to demand
plan accuracy

Alan Hendry,
Principal, SCMO2 Inc.

Achieving accurate demand plans and forecasts may often seem
like an unattainable “holy grail.”  One
way to generate more stable demand plans for products over time is to use SAP
APO demand planning to build composite forecasts.

The conceptual basis for composite forecasts is twofold:

  1. Your sales patterns are constant and known
    within your company. Deviations in apparent “best fit” forecasts are a result
    of data variability, not changes to sales patterns.
  2. There are best-fit patterns for your product
    ranges. The roadmap approach will find these patterns, and the composite
    forecasting run will sort your products according to their pattern

I recommend a 6-step process for setting up composite
forecasting in SAP APO as follows:

STEP 1:  Try to stratify your products

Use knowledge and
judgment to identify products that should have similar sales patterns.  Product groups or families may be arranged by
sales segmentation today, not by sales pattern. For example, printer and
printer paper may be in the same “product group” for sales management, but will
have different sales patterns.

STEP 2: Choose hero

Pick representative
products with clean or corrected history.
High selling products are good for many reasons:             Data fluctuations less significant if volumes are higher;
supply chain probably works best for high selling products; and purchase
pattern by customers more closely matches their sales pattern.  Don’t forget to inspect the sales history for
aberrations and confirm that the sales history to be used in analysis looks
valid and representative, and covers entire history horizon.

STEP 3:  Find best fit profiles

To identify best fit
profiles for hero products: run auto model methods and record results.  Use a high level of aggregation.  Repeat for different historical ranges and
confirm that similar results are obtained.  A handful of profiles is sufficient to start;
for most businesses, around 5 profiles will cover a large number of products.  Ideally, you will have profiles that are
distinctly different, seasonal profiles will have correct season patterns, and
trend profiles reflect long-term trends.

STEP 4:  Set up composite forecasting

Set up each relevant
model as a profile. Each profile must have the forecast error flag checked.  You
can use constant (moving average) as a profile.
Then, assign profiles to the composite forecast profile. Finally, select
the mode to determine the lowest error.

STEP 5: Run forecasts

Run forecast for
selected products and verify results. The system runs each of your chosen
profiles and picks the one with the lowest error in the category you defined.

STEP 6: Analyze and

Analyze the results
for a few cycles. Most products should be assigned to the same profile each
cycle.  This confirms you have found the
correct profiles. Some products may frequently move between profiles, or return
the constant profile.  This is often the
result of data or supply chain issues, not a fundamental shift in demand.  Assign products to best fit profiles where
possible to ensure the stability of forecast results.

By using a roadmap approach to composite forecasting, you
should see optimum forecast profiles for your business over a product’s life

Source:  Everything
you need to know about improving forecast accuracy with SAP APO functionality
for demand planning
, to be presented by Alan Hendry  at Logistics & Supply Chain Management a
collaboration of SAP and SAPinsider, March 4-8, 2013 in Las Vegas.  Visit for more information.

Assigned Tags

      Be the first to leave a comment
      You must be Logged on to comment or reply to a post.