Glossary
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
ABF: Account based forecasting.
AEI: Automatic Equipment
Identification technology which is used to collect shipment tracking information
quickly and effectively.
Aggregate Forecast: Sales forecast of a company as a whole.
Auto Regression or Auto-Regressive
Process: Where sales of one period are regressed on the previous period.
Auto-Regressive Moving Average (ARMA) Process (Model): Where the Auto-Regressive and Moving Average Processes are combined. It is often called ARMA model or the Box-Jenkins Model.
ATP: Available to promise.
Autocorrelation: Correlation
within a series. For example sales of 1996 is related to the sales
of 1995, and sales of 1995 is related to the sales of 1994.
Autocorrelated Time Series: A time series in which the current value of a series depends the past
value.
Back Forecasting: Making
forecasts of periods for which actuals are known. Also, known as
ex-post forecasts.
Base Period: A period in
time from which comparisons of other time periods are made.
Best Linear Unbiased Estimator (BLUE): The criterion used in regression modeling to select the best estimator
from a number of Unbiased Estimators.
Bias: It is often referred
to an error resulting from an error in data gathering, faulty program
design, mistakes on the part of personnel, or data sources.
Bottom-Up Forecasting: Forecasts
that originate from the bottom. For example, obtain sales forecasts from salespeople of different
territories and then add them together to arrive at the aggregate forecast.
Box-Jenkins Model: A time
series model named after the developers of this model. It combines
the Auto-Regressive Process with a Moving Average.
Bullwhip Effect: In case of stock out, customer tends to order more than it needs which corrupts the real pattern.
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
Categorical Variable: A qualitative variable created by classifying observations into categories. For example, a series of household incomes could be classified into categorical variables, low, medium, and high, based on specific ranges of income levels. Many statistical techniques are inappropriate for handling categorical variables. Also referred to as a Qualitative Variable.
Causal Model: A model that assumes that the Variable to be forecast exhibits a cause-and-effect relationship with one more other variables. Regression / econometric models are causal models.Census: A complete enumeration of the universe (population), e.g. the Census of Population and Housing, the Census of Agriculture, and the Census of Manufacturing. In contrast, sample is a portion of the universe.
Census X-11: It is one of the decomposition models. It decomposes time series data into secular (long-term) trend, seasonal index, cyclical index, and irregular component.
Classical Decomposition: A time series model which decomposes the data into trend, cycle, seasonality and randomness.
Coefficient Term: It is a slope of the line. It shows how the dependent variable, on the average, changes with a one unit of change in the independent variable.
Coefficient of Determination: A common measure of the "Goodness of Fit" used in regression modeling to assess the degree of causation between two variables or between one or more independent variables and dependent Variable. It is a square of the correlation coefficient and shows the percent of variations in the dependent (explained) variable that can be explained by the variations in the independent (explanatory) variable(s). Its value varies between 0 (0%) and 1 (100%). The higher the value is, the better.
Consensus Forecasts: Forecasts which are jointly agreed upon. Or average of forecasts given by different individuals.
Correlation Coefficient: A standard measure of relationship between a dependent and independent variable. Its value varies between -1 and 1. Zero means that there is no correlation whatsoever, and one means, perfect correlation. Positive value means that they are positively related, that is when one goes up, the other also goes up. Negative value means that they are negatively correlated.
Customers: Customers of a vendor are distributors, wholesalers and/or retailers.
Cycle: Cyclical fluctuations are those which occur regularly but not periodically. The length of a cycle is always more than one year. For example, economy goes up for a certain number of years and then goes down for a certain number of years. These functions don't occur regularly because in one cycle upswing (downswing) may take three years and in other cycle 5 years.
Data Warehouse: Where data are stored. Date may be stored at the mainframe.
Delphi: This is a judgmental technique of forecasting where a panel of experts are asked to give their own forecasts which is distilled to arrive at the final forecast.
Demand: Booking orders.
Dependent Demand: It represents the demand of vendor's factory (raw material, etc.), vendor's distribution center demand which depends on the customer distribution center's demand, customer retail store's demand which depends on the demand of final consumers.
Dependent Variable: A variable we wish to forecast. In regression analysis the variable being predicted is the dependent variable.
Dissaggregate Forecasts: Breaking up the total company forecast into categories and SKUs.
DRP: Distribution requirements planning --- planning about shipment, transportation, and warehousing.
Durbin-Watson Test: Diagnostic tool used to test a regression model. Its value varies between 1 and 4. The model is the best if its value is 2. Normally, the value between 1.5 and 2.5 is acceptable.
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
Eache: It represents consumer unit of purchase, e.g., a 12 oz bottle of soda and one box of cereal.
Econometric Forecasting: Where a model encompasses more than one equation to make a forecast.
Economic Indicator: It provides an indication of how the economy is behaving.
EDI: Electronic Data Interchange is used for transmitting documents such as invoices, orders, status of order, from one computer to another. With this one computer talks with another.
Efficient Consumer Response (ECR): Synchronizing consumer demand with production.
End-User: Ultimate user of a forecast.
Ex-Ante forecast: Preparing forecasts for periods for which actuals are not known.
Endogenous Variable: The variable which is internal and can be changed. For example, advertising expenditure which can be changed. If the X amount of advertising expenditure does not give the sales the company needs, it can raise it to achieve its desired goal.
Exogenous Variable: The variable which is external and is not within the control of a forecaster. For example, the state of the economy, the forecaster has to accept it as it is because he/she cannot change it.
Explanatory Variables: The variables that drive the sales. For example, advertising outlay, price and state of the economy. They are used to predict values of a dependent variable. They are also called independent variables.
Ex-Post: Preparing forecasts for periods for which actuals are known.
EVA: Economic value added.
Fitted Values: The predicted values derived from a regression model by applying the regression coefficients to the independent variables.
Forecast Horizon: The number of time periods out to be forecasted, i.e., one month out, one quarter out, and one year out.
Forecasting Process: The process outlines who will provide the information used for preparing forecasts; how it is gathered; after information is obtained; how it is processed and used for preparing statistical forecasts; and one statistical forecasts are prepared who participate in the process to arrive at consensus forecasts.
Forecast Receivers (Customers): Ones who receive forecasts (end users).
F Statistics: In a regression model it is used determine the overall performance of a model. Though the value which validates the model depends on the number of observations and number of independent variables used in a model, the higher value, the better.
Forecast Suppliers: Ones who supply forecasts (forecasters).
Forecast System: Mechanizing the forecasting process including the use of software and hardware.
Forward Buy: Occurs when an account buys extra quantity during the deal period to be sold after the deal has ended.
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
Independent Demand: Represents consumption demand. For example, POS based consumption data.
Independent Variables: The variables that drive the sales. For example, advertising outlay, price and state of the economy. They are used to predict values of a dependent variables or drivers.
Intermittent Demand: These are the products that have no demand for many months and sporadic demand in other months.
IRP: Integrated resource planning model.
Lead Time: Time needed to make any change in production plan or ordering raw materials. Or, amount of time required to provide (or produce) a product to an inventory location. Or, time needed to make any change in production plan.
Leading Indicator: Economic indicator whose peaks and troughs during the business cycle tend to occur before the general economy. Stock market prices are generally considered as a leading indicator of the economy.
Macro-forecasts: Forecasts of the economy as a whole. For example, forecasts of GDP and employment.
MAPE: Mean Absolute Percent Error, the average percent error with signs ignored.
Matured Products: Products that have passed their growth stage in terms of demand.
Micro-forecasts: Company level forecasts. For example, sales forecast.
Multicollinearity: When two independent variables are highly associated (correlated) with each other. It is not considered good in regression modeling.
MSE: Mean squared error. Here errors are first squared and then their average is computed.
Naïve Forecast: Next year forecast is the same as the current year actual.
Observations: Number of periods used in a forecasting model.
Operational Forecasts: Short term forecasts, usually of less than one year.
Outlier: A value that is unusually to large or too small.
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
Pooling Effect: When a consumer is not able to find the actual size desired, a high probability exists that the consumer may purchase another size in the same product family prior to switching to a competing brand.
Phasing: Percent of annual sales realized in a given month.
Price Elasticity: How sensitive is the sales to price. Highly elastic if a small change in price leads to a large change in demand. Highly inelastic, if a large change in price leads to a small change in demand.
Product Life Cycle: Refers to a life cycle of a product. The product forms a S curve with four stages of development ---- introduction, growth, maturity, and decline.
Qualitative Forecasting: Refers to judgmental approach to forecasting.
Quantitative Forecasting: Refers to statistical approach to forecasting.
Regression: It is a casual method of forecasting which assumes that the variable to be forecasted exhibits a cause/effect relationship with one or more variables (factors).
Residual: It is equivalent to a forecast error ---- the actual minus the fitted forecast value.
Safety Stocks: Buffer stock used to compensate for uncertainties in demand during lead time.
Scenario Forecasting: A judgment technique of forecasting where several set of circumstances are constructed which form the boundaries within which the actual number is expected to lie.
Seasonality: Seasonal fluctuations are those which occur regularly and periodically and the length of a cycle is always less than one year. For example, the sales of department store reaches peak during November and December because of Christmas. This happens every year and at the same time.
Sell in Forecast: Forecast of shipment from the manufacturer to retailer.
Sell Through Forecast: Forecast of sales to end-use consumers.
Shipping Data: Data of merchandise shipped.
SKU: Stock Keeping Unit (item). For example, shirt, size 15”, half sleeves, white color.
Spatial Autocorrelation: Often arises in a cross sectional data where a change in one region may cause a change in the activity in other region because of close economic linkages.
Standard Deviation: Measure of variations within a series. For example, how errors vary over different periods.
Stock-Out: When inventory isn’t available to meet orders in a timely manner.
Strategic forecasts: Long term forecasts, usually of more than one year.
Index: Top | A-B C-D E-F G-O P-S T-Z Back Home
Tactical Forecasts: Short term forecasts, usually for less than one year.
Time Series Models: Where it is assumed that past pattern will continue in the future. Here one needs only the data of series to be forecasted.
Top-Down Forecasting: Here the forecast is first prepared of the company as a whole which is then disaggregated into category and SKU levels forecasts.
T Test: It is used to determine in a regression model whether the impact of certain independent variable is significant or not. In other words, whether we should keep the variable in or throw it out. The variable is normally considered significant if its value is 2 or higher.
Trend: It is statistically computed. It shows how, on the average, sales is moving, upward or downward.
Unconstrained Demand: What could have been sold if there were no problem in production or anything else which might have cut down the sales.
Univariate Models: Here one need only the data of series to be forecasted. Time series models are univariate models.
Validation:
The process of testing whether the model is valid or not.
REFERENCE
Business and Economic Research Group, Executive Offices,
6275 Neil Rd. Reno, Nevada 89511