5.4 Demand management and forecasting
Demand management and forecasting is recognizing all demand for goods and services to support the marketplace. Demand is prioritized when supply is lacking. Proper demand management facilitates the planning and use of resources for positive and profitable results and may involve marketing programs designed to increase or reduce demand in a relatively short time.
5.4.1 Planning horizon
The planning horizon is how far a plan extends into the future and is dictated by tactical and strategic degrees of uncertainty. The tactical horizon may be based on the cumulative lead time needed to procure or produce low-level components. The strategic horizon is based on the time needed to adjust capacity. A greater degree of uncertainty requires a longer planning horizon.
Forecasting is attempting to predict or project future statistics—typically, demand or sales. It requires that all factors surrounding the decision-making process are recorded. Factors that affect forecasting include sales demand patterns, economic conditions, competitor actions, market research, product mixes, and pricing and promotional activities. Forecasts can be made at strategic, tactical, and operational levels.
.1 Types of forecasts
Forecasts can be categorized by technique, such as subjective, causal, and time-series. Subjective forecasting is a qualitative technique, while the causal and time series methods are quantitative, statistical models.
.2 Forecasting process
The forecasting process predicts demand and the use of products and services so that the right quantities are ordered in advance. In forecasting, either historical demand data are transformed into future projections or a subjective prediction of the future is made—or some combination of the two.
.3 Pyramid forecasting
Pyramid forecasting, or rationalizing high- and low-level forecasts, enables management to review and adjust forecasts made at an aggregate level and keep lower-level forecasts balanced. In the process, item forecasts first are aggregated by product group. Management then makes a new forecast for the group. The value is then transferred
to individual item forecasts so that they are consistent with the aggregate plan.
.4 Forecasting models
Many models are used to predict demand, such as regression analysis, time series, the Delphi method, and market research—or some combination of these quantitative and qualitative methods. Forecast data may be broken down in an attempt to uncover the components of demand, such as trend, seasonality, and cyclical and random patterns. The base component reflects the demand for an item without applying the patterns.
Baseline methods. Baseline demand is the percentage of a company's demand derived from continuing contracts and existing customers. It usually is a predictable component of demand.
Time series. Time series is a technique that projects historical data patterns by looking at past forecasts and forecast errors. A time series may contain seasonal, cyclical, trend, and random components.
Exponential smoothing. Exponential smoothing is a forecasting technique using a weighted moving average, where past observations are adjusted according to their age. The most recent data typically are weighted the heaviest. A smoothing constant is applied to the difference between the most recent forecast and critical sales data, avoiding the necessity of maintaining historical sales data. Alternatives to exponential smoothing include moving average and weighted moving average models.
Trend. Trend is the general upward or downward movement of demand over time.
Seasonality. Seasonality is a cyclical pattern of demand, where some periods of the year are higher or lower than others.
Regression models. Regression models are statistical techniques used to determine the best mathematical expression that describes the relationship between a dependent variable, such as demand, and one or more independent variables.
Focus forecasting. Focus forecasting is a system that allows a user to simulate the effectiveness of numerous forecasting techniques and select the best method.
.5 Error measurement
Error is the difference between a forecast and the actual demand.
Bias. Bias refers to consistent errors that cause a forecast to go either too high or too low. A forecast is biased if the current forecast errors are greater or less than zero.
Standard deviation. Standard deviation is a measurement of the dispersion of data or of a variable. It is calculated by finding the differences between average and actual observations, squaring each difference, adding the squared differences, dividing by n minus 1 (for a sample), and finally taking the square root of the result.
Mean absolute deviation (MAD). MAD is the average of the absolute values of the deviations of observed values from forecast values. It is the arithmetic mean of past absolute errors.
Mean absolute percent error. Mean absolute percent error is computed by taking the MAD, dividing by the average demand, and then multiplying by 100.
Tracking signal. Tracking signal refers to the ratio of the cumulative algebraic sum of the deviation between forecasts and actual values to the mean absolute deviation. This is used to alert that the forecast model is biased.
.6 Special situations
Some situations impact forecast accuracy by creating unique circumstances that change demand.
Promotions and de-promotions. These are products subject to wide fluctuations in sales, often being sold at a reduced price or as part of another sales incentive.
Cannibalization. Cannibalization is the reduction of demand for a product due to the introduction of a new or similar product.
Substitution. Substitution is the use of a different product or component not originally specified on an order but serving the same purpose, which impacts
.7 Data warehouses
Data warehouses are repositories of data specially prepared to support decision-making applications. In forecasting, they are where quantitative and qualitative data are collected for future needs.
.8 Matching methods and uses
Each forecasting method has a unique set of characteristics and applies differently to specific situations. For example, exponential smoothing might be appropriate for a very short-term forecast predicting demand during lead time, but not for a long-term product group demand forecast (where regression might be a better fit). Selecting the best method or methods for specific needs is important in successful forecasting.