Chapter 8 Forecasting
Which statement about forecast accuracy is true? A) A manager must be careful not to "overfit" past data. B) The ultimate test of forecasting power is how well a model fits past data. C) The ultimate test of forecasting power is how a model fits holdout samples. D) The best technique in explaining past data is the best technique to predict the future.
A manager must be careful not to "overfit" past data
A tracking signal greater than zero and a mean absolute deviation greater than zero imply that the forecast has: A) no bias and no variability of forecast error. B) a nonzero amount of bias and a nonzero amount of forecast error variability. C) no bias and a nonzero amount of forecast error variability. D) a nonzero amount of bias and no variability of forecast error.
A nonzero amount of bias and a nonzero amount of forecast error variability
Describe some of the managerial considerations required to utilize big data effectively.
Answers will vary but students may start by describing the 3Vs of big data, which are volume, variety, and velocity and how the complexity of big data creates managerial challenges for harnessing it. Some of the specific managerial considerations are: 1. The need for adequate computing power and server capacity to handle the load of big data, which can be alleviated by public cloud servers. 2. The skills required to clean, organize, and analyze big data. 3. The culture and support of top organizational leaders in accepting findings from big data projects. 4. The managers with knowledge of problems that big data can tackle.
________ is a collection of data from traditional and digital sources and is characterized by volume, variety, and velocity.
Big data
What is the difference between mean absolute deviation (MAD) and mean squared error (MSE)?
Both MAD and MSE are measurements of the amount of forecast error, and smaller values of both metrics reflect superior forecasting methods. The difference between the two is that MAD places less emphasis on an outlier while MSE is more sensitive to one. A forecast technique that seeks to minimize MSE will have overall forecast accuracy hurt by one extreme outlier more than a forecast developed using a MAD-minimizing technique.
________ methods use historical data on independent variables to predict demand.
Causal
A forecasting system that brings the manufacturer and its customers together to provide input for forecasting is a(n): A) nested system. B) harmonically balanced supply chain. C) iterative Delphi method system for the supply chain. D) collaborative planning, forecasting, and replenishment system.
Collaborative planning, forecasting, and replenishment system
Andy took what he liked to call "the sheriff without a gun" approach to forecasting. Every period he tried a number of different forecasting approaches and simply averaged the predictions for all of the techniques. This overall average was the official forecast for the period. The more formal name for this technique is: A) grand averaging. B) focus forecasting. C) simple average. D) combination forecasting.
Combination forecasting
________ are produced by averaging independent forecasts based on different methods or different data, or both.
Combination forecasts
Describe the combination forecast techniques and discuss how they have been shown to perform in recent studies.
Combination forecasts are forecasts that are produced by averaging independent forecasts based on different methods, different sources, or different data. Research during the last two decades suggests that combining forecasts from multiple sources often produces more accurate forecasts. It is intriguing that combination forecasts often perform better over time than even the best single forecasting procedure. Combining is most effective when the individual forecasts bring different kinds of information into the forecasting process. Forecasters have achieved excellent results by weighting forecasts equally, and this is a good starting point. However, unequal weights may provide better results under some conditions
In the winter, Handyman Negri repaired snowblowers and in the summer he earned extra money by repairing lawnmowers, a classic example of: A) promotional pricing. B) complementary products. C) mixed model service. D) yield management.
Complementary products
Which one of the following is most useful for measuring the bias in a forecast? A) cumulative sum of forecast errors B) standard deviation of forecast errors C) mean absolute deviation of forecast errors D) percentage forecast error in period t
Cumulative sum of forecast errors
Which one of the following basic patterns of demand is difficult to predict because it is affected by national or international events or because of a lack of demand history reflecting the stages of demand from product development to decline? A) horizontal B) seasonal C) random D) cyclical
Cyclical
There are historically three 32-month periods of generally rising prices in the stock market for every one 9-month period of falling prices. This observation leads you to conclude that the stock market exhibits a: A) random pattern. B) trend pattern. C) seasonal pattern. D) cyclical pattern.
Cyclical pattern
Which of the following statements about bid data is not true? A) Data technicians must be the ones to identify problems to be tackled with big data. B) Companies employing data-driven decisions tend to be more successful than others. C) Data scientists and skilled professionals are a necessity to execute big data projects. D) Public cloud providers are an option for hosting bid data projects that may swamp single servers.
Data technicians must be the ones to identify problems to be tackled with big data
The ________ is a process of gaining consensus from a group of experts while maintaining their anonymity.
Delphi method
The ________ variable is the variable that one wants to forecast.
Dependent
Which one of the following is an example of causal forecasting technique? A) weighted moving average B) linear regression C) exponential smoothing D) Delphi method
Exponential smoothing
A bias error results from unpredictable factors that cause the forecast to deviate from actual demand.
False
A weary traveler shows up at a hotel desk at midnight without a reservation. The desk clerk informs him that there is a room available, but sadly it is marked up 80% higher than the usual price. This is an example of promotional pricing
False
Aggregating products or services together generally decreases the forecast accuracy
False
Combination forecasting is a method of forecasting that selects the best from a group of forecasts generated by simple techniques.
False
Judgment methods of forecasting are quantitative methods that use historical data on independent variables to predict demand.
False
Judgment methods of forecasting should never be used with quantitative forecasting methods.
False
The Delphi method is a process of gaining consensus from a group of experts by face-to-face, non-anonymous, debate and voting throughout several rounds of group discussion led by a moderator.
False
The closer the value of the sample correlation coefficient is to -1.00, the worse the predictive ability of the independent variable for the dependent variable
False
It is now near the end of May and you must prepare a forecast for June for a certain product. The forecast for May was 900 units. The actual demand for May was 1,000 units. You are using the exponential smoothing method with α = 0.20. The forecast for June is: A) fewer than 925 units. B) greater than or equal to 925 units but fewer than 950 units. C) greater than or equal to 950 units but fewer than 1,000 units. D) greater than or equal to 1,000 units.
Fewer than 925 units
When forecasting total demand for all their services or products, few companies err by more than: A) one to four percent. B) five to eight percent. C) nine to twelve percent. D) thirteen to sixteen percent.
Five to eight percent
Barney took what he liked to call "the shotgun approach" to forecasting. Every period he tried a number of different forecasting approaches and at the end of the period he reviewed all of the forecasts to see which was the most accurate. The winner would be used for next period's forecast (but he still made forecasts all possible ways so he could use the system again for the following period). The more formal name for this technique is: A) combination forecasting. B) post-hoc forecasting. C) focus forecasting. D) shotgun forecasting.
Focus forecasting
What are reasonable criteria for selecting one time-series method over another?
Forecast error measures provide important information for choosing the best forecasting method for a service or product. They also guide managers in selecting the best values for the parameters needed for the method: n for the moving average method, the weights for the weighted moving average method, alpha for the exponential smoothing method, and when regression data begins for the trend projection with regression method. The criteria to use in making forecast method and parameter choices include (1) minimizing bias (CFE); (2) minimizing MAPE, MAD, or MSE; (3) maximizing r2; (4) meeting managerial expectations of changes in the components of demand; and (5) minimizing the forecast errors in recent periods. The first three criteria relate to statistical measures based on historical performance, the fourth reflects expectations of the future that may not be rooted in the past, and the fifth is a way to use whatever method seems to be working best at the time a forecast must be made
Using salesforce estimates for forecasting has the advantage that: A) no biases exist in the forecasts. B) statistical estimates of seasonal factors are more precise than any other approach. C) forecasts of individual sales force members can be easily combined to get regional or national sales totals. D) confusion between customer "wants" (wish list) and customer "needs" (necessary purchases) is eliminated.
Forecasts of individual sales force members can be easily combined to get regional or national sales totals.
The electricity bill at Padco was driven solely by the lights throughout the office; everything else was driven by alternative energy sources. The office was open roughly 8 hours a day, five days a week and the cleaning crew spent about the same amount of time in the offices each week night. The kilowatt hour usage for the office was best described as a: A) horizontal demand pattern. B) random demand pattern. C) seasonal demand pattern. D) cyclical demand pattern.
Horizontal demand pattern
Which forecasting technique would you consider for technological forecasts?
I would consider the Delphi method because technological change takes place at a rapid pace and often the only way to make forecasts is to get the opinion of experts who devote their attention to those issues.
________ are assumed to "cause" the results that a forecaster wishes to predict.
Independent variables
________ methods of forecasting translate the opinions of management, experts, consumers, or salesforce into quantitative estimates.
Judgment
The Delphi method of forecasting is useful when: A) judgment and opinion are the only bases for making informed projections. B) a systematic approach to creating and testing hypotheses is needed and the data are usually gathered by sending a questionnaire to consumers. C) historical data are available and the relationship between the factor to be forecast and other external or internal factors can be identified. D) historical data is available and the best basis for making projections is to use past demand patterns.
Judgment and opinion are the only bases for making informed projections
Which one of the following statements about forecasting is true? A) The five basic patterns of demand are the horizontal, trend, seasonal, cyclical, and the subjective judgment of forecasters. B) Judgment methods are particularly appropriate for situations in which historical data are lacking. C) Casual methods are used when historical data are available and the relationship between the factor to be forecast and other external and internal factors cannot be identified. D) Focused forecasting is a technique that focuses on one particular component of demand and develops a forecast from it.
Judgment methods are particularly appropriate for situations in which historical data are lacking
The judgment methods of forecasting are to be used for purposes of: A) making adjustments to quantitative forecasts due to unusual circumstances. B) generating data for use in time-series approaches. C) providing the calculations necessary for quantitative forecasts. D) calculating the forecast error for quantitative methods.
Making adjustments to quantitative forecasts due to unusual circumstances.
The manufacturer developed and tested a questionnaire, designed to assist them in gauging the level of acceptance for their new product, and identified a representative sample as part of their: A) salesforce estimate. B) market research. C) executive opinion. D) Delphi method.
Market research
________ is a systematic approach to determine consumer interest in a product or service by creating and testing hypotheses through data-gathering surveys.
Market research
Why are forecasts for product families typically more accurate than forecasts for the individual items within a product family?
More accurate forecasts are obtained for a group of items because the individual forecast errors for each item tend to cancel each other
A(n) ________ forecast is a time-series method whereby the forecast for the next period equals the demand for the current period.
Naive
Polly Prognosticator was the greatest quantitative forecaster in recorded history. A skillful user of all techniques in your chapter on forecasting, she knew better than to try and develop a forecast for data that exhibited a: A) random pattern. B) horizontal pattern. C) seasonal pattern. D) cyclical pattern.
Random pattern
One aspect of demand that makes every forecast inaccurate is: A) trend variation. B) random variation. C) cyclical variation. D) seasonal variation.
Random variation
The ________ measures the amount of variation in the dependent variable about its mean that is explained by the regression line.
Sample coefficient of determination, r-squared
A(n) ________ measures the direction and strength between the independent variable and the dependent variable.
Sample correlation coefficient, r
Professor Willis noted that the popularity of his office hours mysteriously rose in the middle and the end of each semester, falling off to virtually no visitors throughout the rest of the year. The demand pattern at work is: A) cyclical. B) random. C) seasonal. D) trend.
Seasonal
With the multiplicative seasonal method of forecasting: A) the times series cannot exhibit a trend. B) seasonal factors are multiplied by an estimate of average demand to arrive at a seasonal forecast. C) the seasonal amplitude is a constant, regardless of the magnitude of average demand. D) there can be only four seasons in the time-series data.
Seasonal factors are multiplied by an estimate of average demand to arrive at a seasonal forecast.
________ is a time-series method used to estimate the average of a demand time series by averaging the demand for the n most recent time periods.
Simple moving average
Your team has been asked to develop a forecast for the need for storage in the company's communication devices ten years from now. What method would develop the best forecast? Why? How would you execute this method?
Since you are tasked with developing a technological forecast ten years into the future, and for a product that has evolved significantly over the last ten years, it is doubtful that a quantitative approach is suitable. Among the judgment methods discussed in the text; salesforce estimates, market research, executive opinion, technological forecasting and the Delphi method, the latter three would hold the most potential for a forecast. Answers will vary as to implementation depending on the approach chosen.
What are some of the principles organizations can observe to improve their forecasting process?
Some principles organizations can observe to improve their forecasting process include: 1. Better processes yield better forecasts. 2. Demand forecasting is being done in virtually every company, either formally or informally. The challenge is to do it well-better than the competition. 3. Better forecasts result in better customer service and lower costs, as well as better relationships with suppliers and customers. 4. The forecast can and must make sense based on the big picture, economic outlook, market share, and so on. 5. The best way to improve forecast accuracy is to focus on reducing forecast error. 6. Bias is the worst kind of forecast error-strive for zero bias. 7. Whenever possible, forecast at more aggregate levels. Forecast in detail only where necessary. 8. Far more can be gained by people collaborating and communicating well than by using the most advanced forecasting technique or model.
Pho Bulous, a Vietnamese restaurant in the bustling metropolis of Edmond, has had great success using forecasting techniques to predict demand for their main menu items ever since they opened their doors. Their forecast for last month was grossly inaccurate and so far this month, their forecast appears to be just as bad as last month's. It's already time to prepare the forecast for next month, what should they do about their model?
The answer depends on whether Pho Bulous believes that last month's and this month's results are aberrations or the start of something new. Both causal and time-series techniques assume that there has been no change in how the world works, that is, independent factors of time or other variables will permit the forecaster to make accurate predictions about the future. If Pho Bulous believes that there is a significant change in the system, for example, a new competitor in the Edmond restaurant scene, a significant change in population or in their disposable income, then they might try multiple regression to include these factors or weight more recent data more heavily in a time-series model (the scenario isn't specific about which technique they have used thus far). Pho Bulous might also try a combination approach if they feel their situation has changed significantly. On the other hand, if Pho Bulous feels that these two months are not reflective of any major paradigm shift for the restaurant crowd in Edmond, they could continue to use the model(s) they have had success with in the past.
What are the steps of the forecasting process as described in the text?
The authors describe a six-step forecasting process. Step 1. Update the history file and review forecast accuracy. Enter the actual demand and review forecast accuracy. Step 2. Prepare initial forecasts using some forecasting software package and judgment. Adjust the parameters of the software to find models that fit the past demand well and yet reflect the demand manager's judgment on irregular events and information about future sales pulled from various sources and business units. Step 3. Hold consensus meetings with the stakeholders, such as marketing, sales, supply chain planners, and finance. Arrive at consensus forecasts from all of the important players. Step 4. Revise the forecasts using judgment, considering the inputs from the consensus meetings and collaborative sources. Step 5. Present the forecasts to the operating committee for review and to reach a final set of forecasts. Step 6. Finalize the forecasts based on the decisions of the operating committee and communicate them to the important stakeholders.
How is a typical forecasting process similar to the Plan-Do-Study-Act (PDSA) cycle?
The authors indicate that forecasting is a process that should be continually reviewed for improvements; the PDSA cycle provides one vehicle for continuous improvement. The authors present a six step cycle for forecasting: 1) adjust the history file, 2) prepare initial forecasts, 3) consensus meetings and collaboration, 4) revise forecasts, 5) review by the operating committee, and 6) finalize and communicate the forecasts. The history file adjustment in step 1 provides a check of forecast accuracy; if results have been less than stellar, then planners and forecasters will explore different techniques and/or independent variables to prepare future forecasts. This approach closely parallels the PDSA cycle of methodically trying a new approach and checking results before acting system-wide.
Assume that a time-series forecast is generated for future demand and subsequently it is observed that the forecast method did not accurately predict the actual demand. Specifically, the forecast errors were found to be: Mean absolute percent error = 10% Cumulative sum of forecast errors = 0 Which one of the statements concerning this forecast is true? A) The forecast has no bias but has a positive standard deviation of errors. B) The forecast has a positive bias and a standard deviation of errors equal to zero. C) The forecast has no bias and has a standard deviation of errors equal to zero. D) The forecast has a positive bias and a positive standard deviation of errors.
The forecast has no bias but has a positive standard deviation of errors
The number of #2 pencils the bookstore sells appears to be highly correlated with the number of student credit hours each semester. The bookstore manager wants to create a linear regression model to assist her in placing an appropriate order. In this scenario: A) the dependent variable is student credit hours. B) there are two independent variables. C) there are two dependent variables. D) the independent variable is student credit hours.
The independent variable is student credit hours.
Which one of the following statements about forecasting is false? A) The method for incorporating a trend into an exponentially smoothed forecast requires the estimation of three smoothing constants: one for the mean, one for the trend, and one for the error. B) The cumulative sum of forecast errors (CFE) is useful in measuring the bias in a forecast. C) The standard deviation and the mean absolute deviation measure the dispersion of forecast errors. D) A tracking signal is a measure that indicates whether a method of forecasting has any built-in biases over a period of time.
The method for incorporating a trend into an exponentially smoothed forecast requires the estimation of three smoothing constants: one for the mean, one for the trend, and one for the error
A forecaster that uses a holdout set approach as a final test for forecast accuracy typically uses: A) the entire data set available to develop the forecast. B) the older observations in the data set to develop the forecast and more recent to check accuracy. C) the newer observations in the data set to develop the forecast and older observations to check accuracy. D) every other observation to develop the forecast and the remaining observations to check the accuracy.
The older observations in the data set to develop the forecast and more recent to check accuracy.
A linear regression model is developed that has a slope of -2.5 and an intercept of 10. The sample coefficient of determination is 0.50. Which of the following statements is true? A) The sample correlation coefficient must be 0.250. B) The sample correlation coefficient must be -0.707. C) The sample correlation coefficient must be -0.250. D) The sample correlation coefficient must be 1.00.
The sample correlation coefficient must be -0.707.
Explain how the value of alpha affects forecasts produced by exponential smoothing.
The smoothing constant alpha allows recent demand values to be emphasized or deemphasized depending on how the forecaster wishes to incorporate previous values. Larger alpha values emphasize recent levels of demand and result in forecasts more responsive to changes in the underlying average. Smaller alpha values treat past demand more uniformly and result in more stable forecasts.
What is the difference between a reservation and an appointment? A) There is no difference between the two terms. B) The term reservation implies that the customer has paid in advance. C) The term appointment implies that the customer has paid in advance. D) The term reservation is issued when the customer occupies the facility to receive service.
The term reservation is issued when the customer occupies the facility to receive service
Which one of the following statements about the patterns of a demand series is false? A) The five basic patterns of most business demand series are the horizontal, trend, seasonal, cyclical, and random patterns. B) Estimating cyclical movement is difficult. Forecasters do not know the duration of the cycle because they cannot predict the events that cause it. C) The trend, over an extended period of time, always increases the average level of the series. D) Every demand series has at least a random component.
The trend, over an extended period of time, always increases the average level of the series.
Which one of the following statements about forecasting is false? A) Causal methods of forecasting use historical data on independent variables (promotional campaigns, competitors' actions, etc.) to predict demand. B) Three general types of forecasting techniques are used for demand forecasting: time-series analysis, causal methods, and judgment methods. C) Time series express the relationship between the factor to be forecast and related factors such as promotional campaigns, economic conditions, and competitor actions. D) A time series is a list of repeated observations of a phenomenon, such as demand, arranged in the order in which they actually occurred.
Time series express the relationship between the factor to be forecast and related factors such as promotional campaigns, economic conditions, and competitor actions
________ analysis is a statistical approach that relies heavily on historical demand data to project the future size of demand, and it recognizes trends and seasonal patterns.
Time-series
A regression equation with a coefficient of determination near one would be most likely to occur when the data demonstrated a: A) seasonal demand pattern B) trend demand pattern. C) cyclical demand pattern. D) random demand pattern.
Trend demand pattern
A linear regression model results in the equation Y = 15 - 23X. If the coefficient of determination is a perfect 1.0, the correlation coefficient must be -1.
True
A naive forecast is a time-series method whereby the forecast for the next period equals the demand for the current period.
True
A simple moving average of one period will yield identical results to a naive forecast.
True
A water ski manufacturer believes they can double their sales by producing snow skis during the other half of the year. This approach to demand management is an example of complementary products
True
Aggregation is the act of clustering several similar products or services
True
An exponential smoothing model with an alpha equal to 1.00 is the same as a naive forecasting model.
True
Better forecasting processes yield better forecasts.
True
Bias error causes the greatest disruption to planning efforts
True
Focus forecasting selects the best forecast from a group of forecasts generated by individual techniques.
True
Technological forecasting is an application of executive opinion to keep abreast of the latest advances in technology.
True
The causal method of forecasting uses historical data on independent variables (such as promotional campaigns and economic conditions) to predict the demand of dependent variables (such as sales volume).
True
The repeated observations of demand for a product or service in their order of occurrence form a pattern known as a time series
True
Time-series analysis is a statistical approach that relies heavily on historical demand data to project the future size of demand.
True
A weary traveler shows up at a hotel desk at midnight without a reservation. The desk clerk informs him that there is a room available, but sadly it is marked up 80% higher than the usual price. This is an example of: A) promotional pricing. B) yield management. C) backlogs. D) backorder.
Yield management
Which one of the following statements about forecasting is false? A) You should use the simple moving-average method to estimate the mean demand of a time series that has a pronounced trend and seasonal influences. B) The weighted moving-average method allows forecasters to emphasize recent demand over earlier demand. The forecast will be more responsive to change in the underlying average of the demand series. C) The most frequently used time-series forecasting method is exponential smoothing because of its simplicity and the small amount of data needed to support it. D) In exponential smoothing, higher values of alpha place greater weight on recent demands in computing the average.
You should use the simple moving-average method to estimate the mean demand of a time series that has a pronounced trend and seasonal influences.
It would be most appropriate to combine a judgment approach to forecasting with a quantitative approach by: A) having a group of experts examine each historical data point to determine whether it should be included in the model. B) combining opinions about the quantitative models to form one forecasting approach. C) adjusting a forecast up or down to compensate for specific events not included in the quantitative technique. D) developing a trend model to predict the outcomes of judgmental techniques in order to avoid the cost of employing the experts.
Adjusting a forecast up or down to compensate for specific events not included in the quantitative technique
Variations in demand that cannot be predicted are said to be a(n) ________ pattern.
random
A systematic increase or decrease in the mean of the series over time is a(n) ________.
trend
In an exponential smoothing model a ________ value for alpha results in greater emphasis being placed on more recent periods.
Larger
________ is a causal method of forecasting in which one variable is related to one or more variables by a linear equation.
Linear regression
The trend projection with regression model can forecast demand well into the future.
True
The naive forecast may be adapted to take into account a demand trend.
False
Traditional data processing applications are capable of handling big data.
False
If forecast errors are normally distributed with a mean of 0, the relationship between σ and MAD is: A) 1.25MAD ≈ σ B) MAD ≈ 1.25σ C) MAD ≈ 0.5σ D) 0.8MAD ≈ σ
1.25MAD ≈ σ
14) "Well if you're out of Duff I'll just take my business elsewhere!" the customer shouted as he stomped out of the Quickie Mart. This unfortunate incident could be described as: A) a stockout. B) a backorder. C) a backlog. D) yield management.
A stockout
The dispersion of forecast errors is measured by both MAD and MSE, which behave differently in the way they emphasize errors. ________ gives larger weight to errors and ________ gives smaller weight to errors.
MSE, MAD
Nathan managed to level the customer requests for his valuable services by offering reservations, deploying some promotional pricing, and engaging in yield management, all forms of ________.
demand management
Combination forecasting is most effective when the techniques being combined contribute different kinds of information to the forecasting process.
True
Forecast error is found by subtracting the forecast from the actual demand for a given period.
True
Forecasts almost always contain errors
True
One of the basic time series patterns is random
True
8) When the underlying mean of a time series is very stable and there are no trend, cyclical, or seasonal influences: A) a simple moving-average forecast with n = 20 should outperform a simple moving-average forecast with n = 3. B) a simple moving-average forecast with n = 3 should outperform a simple moving-average forecast with n = 15. C) a simple moving-average forecast with n = 20 should perform about the same as a simple moving-average forecast with n = 3. D) an exponential smoothing forecast with a = 0.30 should outperform a simple moving-average forecast with α = 0.01.
A simple moving-average forecast with n = 20 should outperform a simple moving-average forecast with n = 3.
Market research is a systematic approach to determine consumer interest by gaining consensus from a group of experts while maintaining their anonymity.
False
A(n) ________ is a portion of data from more recent time periods that is used to test different models developed from earlier time period data.
Holdout set