demand forecasting

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Demand Forecasting

2605 Meridian ParkwaySuite 100Durham, NC 27713

Phone: 800.652.4274Fax: 919.484.4450www.clarkstongroup.com

Copyright 2000 Clarkston Group, Inc. All rights reserved.

This document may not be reproduced without the written consent of Clarkston Group, Inc.

Demand Forecasting 3

Contents

2

EXECUTIVE SUMMARY............................................................................................................... 4

WHAT IS DEMAND FORECASTING?........................................................................................ 5

II. A (HUMBLING) FACT ABOUT FORECASTING ....................................................................... 6III. DEMAND CHARACTERISTICS............................................................................................. 6IV. FORECASTING METHODS AND TECHNIQUES..................................................................... 6

Judgment Methods .................................................................................................................. 6Time Series Analysis ............................................................................................................... 8Causal Methods ....................................................................................................................... 9

DEMAND FORECASTING BEST PRACTICES........................................................................ 11

I. SALES AND OPERATIONS PLANNING (S&OP) PROCESS..................................................... 11II. ENABLING SOFTWARE ........................................................................................................ 11III. USE DIFFERENT FORECASTING TECHNIQUES FOR DIFFERENT PRODUCTS ..................... 11IV. FORECAST DEMAND VERSUS SHIPMENTS (OR DEMAND VERSUS SALES) ....................... 12V. TRACKING FORECAST ERROR OR OTHER PERFORMANCE MEASUREMENTS...................... 12VI. SALES AND MARKETING INVOLVEMENT ........................................................................ 12

SELECTED REFERENCES IN SUPPLY CHAIN MANAGEMENT......................................... 13

ABOUT CLARKSTON........................................................................................................................14

Note: Demand forecasting is one component of effective demand management. The other majorcomponents of demand management are (1) order processing, (2) making availabletopromisequotes, and (3) link between production and the marketplace. This paper will focus on theforecasting element of demand management.


Executive Summary

Demand forecasting is the process of determining what products are neededwhere, when, and in what quantities. It can be a significant source of competitiveadvantage by improving cost structure and customer service levels, anddecreasing backorders and lost sales. The goal of forecasting is to minimize theerror between the forecast and the actual demand. A successful forecastingstrategy includes selecting the appropriate forecasting technique.

A successful forecasting strategy includes:

Sales and operations planning (S&OP) process Enabling Software Using different forecasting techniques for different products Forecasting customer demand versus actual shipments Tracking forecast error and other performance measurements Involvement of sales and marketing in the forecasting process.


What is Demand Forecasting?

Forecasting customer demand for products and services is one of the mostfundamental tasks that a business must perform. It is a proactive process ofdetermining what products are needed where, when, and in what quantities.Consequently, demand forecasting is a customerfocused activity.

Demand forecasting is also the foundation of a companys entire logistics process.It supports other planning activities such as capacity planning, master productionscheduling (MPS), inventory planning, and even overall business planning.

I. Benefits of Effective Demand Forecasting

As we all learned in Business 101, profit is revenue less expenses, and theultimate goal of business is to maximize profit. Effective demand forecasting willhelp your business reach its goals. From the customers perspective, their needswill be satisfied by having your goods available when they want them. From thebusinesss perspective, revenue will increase since lost sales are minimized. Also,your businesss cost structure will improve due to more efficient use of capital.The companys income statement, balance sheet, and cash flow statements can allbe positively influenced by effective demand forecasting.

In addition to the raw financial benefits mentioned above, there are otheradvantages to improved performance in demand forecasting. They include:

Improved customer service levels. Fewer backorders and lost sales. Less inventory investment in safety stock. Improved production planning processes. Earlier recognition of marketplace trends.


II. A (Humbling) Fact About Forecasting

A simple fact about forecasting is that they are usually wrong. What is importantis the direction and amount the forecast deviates from that which actually occurs.Because forecasts are looking into the unknown future, some level of errorbetween demand forecasts and actual demand is expected. The magnitude of theerror, referred to as forecast accuracy, is what determines whether your forecastsare good or bad.

Careful selection of appropriate forecasting techniques and following bestpractices will help your business improve forecast accuracy.

III. Demand Characteristics

Before a discussion of popular forecasting methods, it is important to be familiarwith the five major characteristics of demand:

1. Average: Demand tends to cluster around a specific level.2. Trend: Demand consistently increases or decreases over time.3. Seasonality: Demand shows peaks and valleys at consistent intervals. These

intervals can be hours, days, weeks, months, years, or seasons.4. Cyclical: Demand gradually increases and decreases over an extended period

of time, such as years. Business cycles (recession and expansion) and productlife cycles influence this component of demand.

5. Random Error: Variations that cannot be explained or predicted.

IV. Forecasting Methods and Techniques

There are three types of forecasting methods available to predict demand: (1)judgment methods, (2) time series analysis, and (3) causal methods. Each of thesemethods are described below with the techniques used to apply the method.

Judgment MethodsThese methods utilize opinions to develop forecasts and are generally usedwhen historical data is not available. The basis for judgement methods is thatthe decision maker(s) possess sufficient experience to establish forecasts. Ingeneral, they are low cost and have a rapid development time. However, theyare not consistently accurate and are subject to bias by the group creating theforecast.


1. Sales Force Estimates

The sales force, which is closest to the customer and immersed in the marketplace,creates estimates of customer demand.

Because of their proximity to the consumers, information from the sales forcecan be very reliable. However, individual biases and can decrease theeffectiveness of this technique.

2. Executive Opinion

Executives, who usually have a good understanding of the broadbased factorsthat influence demand, create estimates of customer demand.

This can be a good method to use when creating forecasts for new or verystrategic products. However, it consumes valuable executive time and requiresconsensus between executives.

3. Delphi Method

The Delphi method is a facilitated process of gaining consensus within agroup of anonymous participants. The facilitator sends a forecastquestionnaire to each member of the Delphi group. Anonymity is critical inthis method to prevent a few group members from dominating the decision.When the questionnaire is returned, the responses are statistically summarizedand then sent back out to the group. Each Delphi member has the choice tomodify their previous responses based on the responses of the group. This is areiterative process that continues until a consensus is obtained.

The Delphi method is used for new products or for very longrange forecasts.However, it is a timeconsuming process that is highly dependent on thequality of the questionnaires. Furthermore, participants may provideinadequate responses because there is no accountability.


Time Series Analysis

Time series forecasting methods, also known as intrinsic methods, use priordemand history to generate a forecast. The techniques that make up thiscategory assume that the past patterns of demand will continue into the future.In the shortterm, they perform very well; in the longterm, they generallyperform poorly.

A commonly used analogy is that using time series analysis techniques is likedriving a car by looking at the rearview mirror. On a straight road this is anacceptable approach, but on windy roads, this would lead to disastrous results.As a result, these techniques are best used for (1) mature products, (2)products with a large amount of historical data, or (3) products with smooth,random demand variations.

Demand forecasting software easily systematizes these forecasting methods.Because of this, forecasting large numbers of items by these methods is easilyperformed with software.

1. NaveThe nave technique is simply the forecast for the next period is equal to thedemand for the current period. This is a simple, lowcost method that onlytakes trend into account. However, if demand is variable, this is a poor methodto use.

2. Moving Average (simple and weighted)The simple moving average method is used to determine the average amountof demand over a given time period. The weighted moving average techniqueallows each period in the calculation to carry its own weight. The user canspecify that recent demand will be weighted more heavily in the calculationthan older data.

Moving average techniques are most appropriate for stable demand patterns,when demand does exhibit trend or seasonality components. Unfortunately, iftrend or seasonality exists in the demand, the forecasts will lag in their abilityto predict future demand.


3. Exponential Smoothing (single, double, triple)Exponential smoothing methods require userspecified factors to calculate theforecast:

smoothing (averaging): reactivity of forecast to current demand period trend: dampen or accelerate trend in the forecast seasonality: dampen or enhance the effect of seasonality on the forecast

Single exponential smoothing uses the smoothing factor only in the forecastcalculation. Double exponential smoothing uses the smoothing and trendfactors in the forecast calculation. Triple exponential smoothing uses thesmoothing, trend, and seasonality factors in the forecast calculation.

These techniques are popular and are easily calculated with demand planningsoftware. Unfortunately, a great deal of effort may be required to estimate thethree factors used in forecast calculations. Also, while they perform well inshortterm forecasts, they do not perform as well on products with lowdemand.

Causal Methods

Causal methods create forecasts by determining a causeeffect relationshipbetween independent variables and the demand for the product. Two exampleswhere causal methods could be used to create forecasts will help to illustratethis: Forecasting snowblowers: Long range forecasts for winter snowfall can be

used to forecast snowblower consumption. Promotion planning: If the consumer price of a product is reduced by

$1.50 per unit, how will it affect demand?

One of the benefits is that causal methods provide good longterm forecastaccuracy. For instance, if a weatherman predicts that it will be a cold winter,then the demand for heavy jackets will be greater.

Perhaps their most beneficial aspect is that causal methods support Whatif? analyses. However, the forecasts are only as good as the independentvariables identified and the model created. They require careful thought andinsight into the variables that effect demand.

1. Linear Regression (simple and multiple)Simple linear regression generates a forecast by linking one independentvariable to the demand for a product. For example, sales of ice cream may bedependent on the price that is charged for the product. A model would bedeveloped which described this relationship. Given a specific price, a demandforecast for ice cream will be generated.


Multiple linear regression generates a forecast by linking two or moreindependent variables to the demand for a product. For example, sales of icecream may be dependent on the price that is charged for the product, thetemperature, and the number of hours of daylight. A model would bedeveloped which described this relationship. Given a specific price, atemperature, and a number of daylight hours, a demand forecast for ice creamwill be generated.


Demand Forecasting Best Practices

The six points below are the primary source of best practices in demandforecasting. While this list is not comprehensive of all the best forecastingpractices, these processes are characteristic of worldclass demand forecastingestablishments.

I. Sales and Operations Planning (S&OP) Process

The sales and operations planning process is perhaps the most important and oneof the best practices. In this process, decision makers from a variety ofdepartments hold a regular meeting to arrive at a one number forecast.Typically, representatives from logistics, forecasting, production, marketing,sales, and finance meet in the process. This ensures that multifunctional input isobtained for the forecast.

The S&OP process also provides an avenue for accountability because the grouphas agreed to the forecast value. If the actual demand does not match the forecast,the group is accountable for understanding the cause of the error.

II. Enabling Software

Enabling software is a critical element to generating quality forecasts. However, itis up to each business to determine the appropriate forecasting methods that are tobe used in the software. Companies that can use software properly, whileunderstanding the drivers for their demand, will be more successful in theirforecasting initiatives.

III. Use Different Forecasting Techniques for Different Products

Aspirin is a good medicine for a headache, but not for an upset stomach.Likewise, a specific forecasting technique is not appropriate for all types ofproducts. For example, moving average may be a good technique to use formature products with little demand variability, but would be a poor choice fornew products in a growth stage.


IV. Forecast Demand versus Shipments (or Demand versus Sales)

Demand and shipments (or sales) are different. Although shipments are typicallytracked and easy to access, they are not a true measurement of what the customersreally want. Shipments tend to be less than demand, because of backorders, lostsales, manufacturing delays, and other constraints in the business.

V. Tracking Forecast Error or Other Performance Measurements

There are a number of different ways to monitor forecasting performance. One ofthe most popular methods is to track forecast error, which is the differencebetween the forecast and the actual demand. Good forecasting practices include adisciplined system of monitoring forecast error and understanding the underlyingcauses of error.

VI. Sales and Marketing Involvement

It is essential that the sales and marketing organizations of the company beinvolved in the forecasting process. Unlike many of the other departments in thebusiness, sales and marketing have the ability to directly affect customer demand.They also tend to understand the customers and marketplace better than othergroups. Because of this, they can validate forecast values and assist in makingadjustments.


Selected References in Supply ChainManagement

Fisher, Marshall L., Janice H. Hammond, Walter R. Obermeyer, and AnathRaman. Making Supply Meet Demand In An Uncertain World, HarvardBusiness Review (MayJune 1994), pp. 8393. HBR reprint 94302.

Magretta, Joan. The Power Of Virtual Integration: An Interview With DellComputers Michael Dell, Harvard Business Review, (MarchApril 1998), pp.7384. HBR reprint 98208.

Chambers, John C., Satinder K. Mullick, and Donald D. Smith. How To ChooseThe Right Forecasting Technique, Harvard Business Review, (JulyAugust1971), pp. 4569. HBR reprint 71403.

Barnett, F. William. Four Steps To Forecast Total market Demand, HarvardBusiness Review, (JulyAugust 1988), pp. 39. HBR reprint 88401.


About Clarkston

Clarkston is a high technology ebusiness consulting firm that helps its clientsSeize the Advantage by providing comprehensive, strategic business solutionsdriven by ebusiness. The company harnesses the power of the Internet to unleashits clients' business potential by providing a full range of services includingstrategy, implementation and application support. Through a client partnershipapproach, Clarkston collaborates with its clients to articulate a competitivebusiness vision and assemble a highenergy deployment team to build businessmodels with Coherent Speed. Clarkston guides clients in the pharmaceutical, hightechnology, telecommunications, and manufacturing markets. The company, withheadquarters in Durham, N.C., has locations in Atlanta, Boston, Chicago, Detroit,San Francisco and India.

To learn more about Clarkston, contact Sara Sondecker at 800.652.4274 or visitour website at http://www.clarkstongroup.com.

demand forecasting

Documents