# (Article) MANAGING PRODUCTS VIA DEMAND VARIABILITY AND BUSINESS IMPORTANCE (2007)

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

MANAGING PRODUCTS VIA DEMAND

VARIABILITY AND BUSINESS IMPORTANCE


By Richard Herrin

Since every company has produc- 1. Confidence Interval (For the calculation of safety stock, read
tion capacity and working capital my article “How to Calculate Safety Stock
2. Coefficient of Variation (CV)
for Highly Seasonal Products,” published
constraints, they cannot maintain 3. Coefficient of Determination (R2) in the Summer 2005 issue of the Journal.)
the same flexibility and service Based on experience, you can set a target
4. Mean Absolute Percent Error (MAPE)
levels across the complete prod- of safety stock for a given set of customer
uct portfolio … forecasters should Confidence Interval: There is a standard service and lead time. Let’s say that the
devote most of their efforts to method for computing the safety stock target is 2,500 units. Then, if the safety
based on the desired customer service, stock of a certain product comes to more
products that are low in fore- than 2,500 units, it is unforecastable;
deviation (forecast error), and lead time.
castability but high in business if it comes to 2,500 units or less, it is
importance … forecasts improve if forecastable.
done on an aggregate level.
Coefficient of Variation (CV): This shows

T
he main objective of Sales and the percent of variation in the data around
Operations Planning (S&OP) is the mean. The more variations in the data,
to deliver operating plans that the more difficult it is to forecast. Here
balance supply and demand in such a you can set your own number, say, 80%.
way that production assets are efficiently If a product has a coefficient of variation
utilized, customer service targets are of 80% or more, it should be designated as
met or exceeded, and working capital an unforecastable product. Otherwise, it
requirements are minimized. For optimal should be labeled as forecastable product.
results, the organization must quantify
uncertainty in forecasted demands. The Coefficient of Determination (R2): The co-
S&OP process accounts for demand efficient of determination can be computed
uncertainty by managing a balance among from the actual and forecast data. It gives
supply flexibility, customer service RICHARD HERRIN an indication how closely the forecasts
levels, and inventory. Since almost all are related to the actual numbers. It varies
companies have production capacity and Mr. Herrin works for Syngenta’s between 0 and 1. The closer it is to 1, the
working capital constraints, they cannot North American region where he is easier it is to forecast. Here again, you can
maintain the same flexibility and service responsible for implementing best set your own target, say, 0.70. This means
levels across the complete portfolio of practices in forecasting and demand that any product with a R2 of 0.70 or less
products. Therefore, the S&OP must use management. He is currently leading is unforecastable, and anyone with a value
segmentation strategies. The segmentation a project to implement statistical of more than that is forecastable.
strategy is best achieved by categorizing forecasting. His early career was spent
products into quadrants based on demand in production operations and Sales and Mean Absolute Percent Error: This is
variability (forecastability) and overall Operations Planning. He holds a BA in an average error with signs ignored. Here
importance to the business. Chemistry from Jacksonville University, again, you can set your own error target;
a BA in Chemical Engineering from the let’s say it is 20%. This then means any
There are several different ways University of Florida, and an MBA from product that gives a MAPE of 20% or
to measure forecastability. The most the University of North Carolina. He is more is unforecastable; one that gives less
important among them are: also a Certified Professional Forecaster. is forecastable.

6 THE JOURNAL OF BUSINESS FORECASTING, SPRING 2007


At Syngenta, we use a combination of
two different criteria. One is R2 and the other
FIGURE 1
is MAPE. Any product that has R2 of 0.70
or higher and a MAPE of less than 100% PRODUCTS BY IMPORTANCE AND VARIABILITY
is designated as a forecastable product
(For highly seasonal products, weighted
MAPE is more useful than MAPE). If
this criterion is not met, we aggregate the Q2 Q1
product to a different level (say, category). Stable Demand Stable Demand

High
At a higher level of aggregation, products and and
generally become forecastable.
Low Importance High Importance

Forecastability
The importance to business of each
product is based on a number of criteria
including sales volume, profitability, and
the stage of life cycle the product is in.
We have a weighting system that gives
a weight to each of these elements for a Q3 Q4
product and then sums them to find the Low Variable Demand Variable Demand
total weight for that product. and and
Low Importance High Importance
In Figure 1, Quadrant 1 has products with
stable demand and high business importance;
Quadrant 2, products with stable demand
and low importance; Quadrant 3, products Low High
with variable demand and low importance; Business Importance
and Quadrant 4, products with variable
demand and high importance.

This approach allows planners to FIGURE 2


differentiate supply and demand planning RESULTS OF CATEGORIZATION
processes to develop plans that efficiently
utilize production assets, exceed or meet
customer service targets, and minimize
15% of all products 85% of total
working capital requirements.
sales dollars sales dollars

RESULTS OF THE
CATEGORIZATION 14% of all products 31% of all products 45% of all products
High

9% of total sales dollars 50% of total sales dollars 59% of total sales
Forecastability

We categorized about 1,000 finished


products that are forecasted by month.
The results shown in Figure 2 indicate that
45% of all the products are most important
to the business and are highly forecastable. 17% of all products 38% of all products 55% of all products
Low

They bring in 59% of total sales. Among 6% of total sales dollars 35% of total sales dollars 41% of total sales
the other 55% of products, only 35% of
them are important to business and require
the most attention from the forecaster.
Low High
BUSINESS STRATEGY Business Importance
FOR MANAGING
DIFFERENT ITEMS Note: Forecastability results were determined by product and by month.

After all the items were sorted out by

THE JOURNAL OF BUSINESS FORECASTING, SPRING 2007 7


quadrant, we developed our best strategy Products in Quadrant 2 (Q2): The pro- VMI processes. If any product in this
for managing items of each quadrant. ducts in this quadrant have a stable demand quadrant is very costly and disruptive
(thus they are easy to forecast) but are of to manufacturing, we try to phase it
Products in Quadrant 1 (Q1): Products in low importance to the business. Below is out.
this quadrant have a stable demand and are how we manage these products.
of high importance to the business. Since 2. Forecasts are prepared at an aggregate
they have a stable demand, they are easy 1. We generate statistical baseline level, and then blown down to a
to forecast. The strategies we developed forecasts as well as their confidence lower level of detail, using historical
for these products are as follows: intervals. If, at any given time, ratios.
salespeople feel that the demand
1. Since these products are easy to pattern of a given item will be 3. The emphasis here is on supply plans
forecast, we generate a statistical significantly different from the that minimize the cost. Products are
baseline forecast at a 90% confidence historical pattern, they can overwrite produced in large lots approximately
interval. Confidence interval tells its statistical baseline forecast as once a year.
us how the forecast can vary at long as it is within 90% confidence
different levels of confidence. For level. For any change beyond that Products in Quadrant 4 (Q4): These
example, if the forecast is 100 units, confidence interval, they must products have an unstable demand (thus
we can say that at a 95% level of give the reason why, and need the they are difficult to forecast) and are
confidence (if we want to be right approval of S&OP. of high importance to the business. We
95 times out of hundred), the actual manage these products as follows:
will be as low as 85 units and as 2. Forecasts are updated at least
high as 115 units. Forecasts of items quarterly. 1. Products in this quadrant are, by
in this quadrant are based on the nature, very difficult to forecast. We
assumption that future demand will 3. Since these products are highly try to improve their forecast through
follow the same historical patterns. forecastable, the economy of intensive collaborative efforts (work-
If and when salespeople feel that production rather than flexibility is ing with our customers) and/or throu
the demand pattern of a given item more important. Here the emphasis gh aggregation. Forecasts generally
will be significantly different from is on minimizing the cost. As such improve when they are forecasted
the historical pattern, they can they are produced in large lots. We on an aggregate level. So to improve
overwrite its statistical baseline produce these products once or twice their forecasts, we can either
forecast. If the override is outside the a year during the off season when aggregate them at a higher level (e.g.,
bounds of the confidence interval, idle capacity is available. at a category level instead of product
they must give the reason why. It level), forecast them over longer-
may be that the assumptions made 4. On these products we provide time horizons (e.g., quarterly instead
in preparing forecasts are not valid. customer service levels of 90% or of monthly), or use a combination
Whatever the reason it should be less. of both. We often prepare their
well documented. forecasts at an aggregate level and
Products in Quadrant 3 (Q3): Products then allocate down among different
2. Since the demand for these products is in this quadrant have a variable demand products using either their historical
stable, production doesn’t have to be (thus they are difficult to forecast) and ratios or an agreed product mix.
very flexible and we can reserve spare are of low importance to the business. We
capacity for products in Quadrant 4 manage these items as follows: 2. For these products, we maintain
(which are of high importance but maximum production flexibility so
their demand is highly unstable). 1. Since these items are very difficult to that we can respond to short-term
For these items, we schedule enough forecast, their forecasts are not quite changes in demand. In season, such
production to cover a minimum of meaningful for actual replenishment. products are produced on a monthly
three-month forward demand. For them, we optimize cost over cycle. Flexibility can be achieved
flexibility in the supply chain. Supply either by carrying excess safety
3. Forecasts are prepared by items and will produce the most cost-efficient stock or by having idle capacity
are updated quarterly. production volumes and commercial available to produce quickly on short
will have the task of selling up to notice. We wind up having a trade off
4. On these products we target a but not beyond this volume. We between higher safety stock and idle
customer service level of 90% or also try to supply these products on capacity available (ability to adjust
higher. a make-to-order basis or through production quickly to demand).

THE JOURNAL OF BUSINESS FORECASTING, SPRING 2007 9


3. We target customer service on these
items at 90% or higher.

CONCLUSIONS
BE A MEMBER OF
By categorizing products on the basis of THE INSTITUTE OF BUSINESS FORECASTING
forecastability and business importance,
we can optimize demand and supply Benefits include:
plans and ensure customer service targets
are met with minimum working capital •Journal of Business Forecasting Forecasting,” “How to Select Forecasting Software/
Complimentary for active IBF Members, each issue Systems,” and more. Plus, you will have access to
investment. S&OP should set the targeted gives you a host of jargon-free articles on how to obtain, electronic copies of the latest journal. Moreover, you will
customer service levels and the business recognize, and use good forecasts written in an easy-to- also have access to our Action Templates, ready to use.
importance criteria. If the plans result understand style for business executives and managers. Currently, they include: (1) How to calculate forecast
in average inventory projections that Plus, it provides new, practical forecasting ideas to help error? (2) How to calculate how much money you will
you make vital decisions about sales, capital outlays, save by reducing specific amount of error? (3) How to
are above the targeted working capital credit, plant expansion, financial planning, budgeting, calculate safety stocks (forthcoming).
level, S&OP should address it by either inventory control, production scheduling and marketing
reducing customer service targets (lower strategies. A one- year subscription includes 4 issues.
Most of the articles are written for and by practicing •Events & Training (Discounts available)
safety stock, higher back order, higher risk IBF Conferences and Tutorials can raise your forecasting
forecasters.
of stock out), by increasing production accuracy to new levels. Get step-by-step training, hear
capacity (reduced cycle inventory, •Journal of Business Forecasting Past Articles case studies from forecasting professionals working in
increased capital assets), or by reducing well known companies, see demos of the latest software
NEW! Active Members will now have FULL access to all
packages and systems, network and make long lasting
demand variability (via pricing, promotion, Journal of Business Forecasting articles since inception.
connections with your forecasting peers, and more. Our
With active IBF Membership, you will have the ability
etc.). If these options are not acceptable, events are run in Europe, Asia, as well as in the U.S.A.
to download unlimited .pdf files of articles based on
the S&OP should raise the targets in line your set search criteria. This way you will have access to
Plus, we also offer online events through our Webinar
with the current projections. series.
research at your fingertips! You can access hundreds of
articles representing a multitude of industries, companies,
Join us at an IBF event today! For a full schedule of our
and topics including demand planning and supply chain
The highest level of forecast accuracy upcoming events and testimonials, visit us online: www.
management. This access will give you a step ahead in
can be obtained by implementing pro- ibf.org.
improving your forecasting performance. There is no
cesses that combine statistically generated other body of knowledge which is as extensive as this one
forecasts with the practical experience of and is geared primarily towards forecasting practitioners. In-House Training Seminars
sales and marketing personnel. We use (Discounts available)
•Benchmarking Research Reports Bring the IBF to your workplace. Enjoy the convenience
time-series based statistical forecasts as Our benchmarking reports will provide you with of a professionally developed forecasting training program
a starting point and allow management understanding of key metrics and how your company for your staff at a location of your choice anywhere in the
to overwrite with additional market measures up. The ultimate outcome of these studies world. Gain knowledge and hands-on training that can
intelligence (judgmental input). Judgment is to gain a solid understanding of the “best in class” be put to use right away. Companies that recently had
metrics most companies are achieving. Research In-House Training include: GAP, Cadbury, Wachovia,
forecasts within the statistical range includes: benchmarks of forecasting errors, forecasting Wyeth, GlaxoSmithKline, Nike, Molson, and more. Call
(confidence interval) are not challenged. software/systems, forecasting salary, and more. These in- us for further details today! Discounts are applicable for
If they go beyond the accepted confidence depth studies of topics are based on various surveys of Corporate Members.
interval they should be challenged. If forecasting professionals from IBF events as well as from
other sources.
accepted, the revised forecasts should Forecasting Books (Discounts available)
be documented along with reasons for Our books are geared toward helping professionals learn,
changing. Forecasters should devote •Knowledge & Action Templates process, interpret, and implement Business Forecasting
Our growing online knowledge base includes key issues information. In addition, if you miss one of our
most of their efforts to products that and information on forecasting. This covers issues such conferences, we offer manuals that detail each speaker’s
are low in forecastability but high in as “How to Win the Support of Top Management for presentation from all our conferences.
business importance. Products that
are low in forecastability and low in
business importance should be managed Individual Membership Corporate Membership
less frequently (quarterly or annually). (8 People Maximum)
Products that are highly forecastable $250 Domestic, $300 Foreign $1800 Domestic, $2000 Foreign
should not be revised unless the forecaster
has information on significant changes in
their demand drivers. These changes must Call 516.504.7576 or visit us on the web www.ibf.org to sign up!
be documented. g
(richard.herrin@syngenta.com)

10 THE JOURNAL OF BUSINESS FORECASTING, SPRING 2007


Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

You might also like