BUS606 Study Guide

Unit 5: Demand Forecasting

5a. Explain the difference between qualitative and quantitative forecasting 

  • What is forecasting, and what is the difference between forecasting and planning?
  • What methods can forecasters use to predict demand?

Forecasters develop scenarios of what the future might look like; planners develop the actions that need to take place based on the forecast. Forecasting is hindered by external uncertainty (external to the organization), and planning is hindered by internal uncertainty (internal to the organization).
Forecasters can use both qualitative and quantitative forecasting in making their forecasts. Qualitative forecasting uses non-statistical tools; these forecasters rely on past experiences and inferences. Qualitative forecasting may use experts, upper management, and market research in the forecasting process. The main disadvantage of qualitative forecasting is subjectivity. Quantitative forecasting, on the other hand, relies on data to make predictions. Statistical confidence intervals are used to predict the confidence of these predictions.
A third option is hybrid forecasting, a combination of statistical and non-statistical methods. A hybrid model can result in a forecast that evens out the uncertainties of using only qualitative or quantitative forecasting models. Hybrid models are less effective in high-demand volatility markets.
To review, see Forecasting and Measuring Forecast Accuracy in a Pharmacy.

5b. Create quantitative forecasting analyses using linear regression, moving average, and exponential forecasting techniques

  • How can we use linear regression for forecasting in OSCM?
  • How do correlation and covariance apply to linear regression?
  • When can we use the moving average to forecast demand?
  • How does the moving average differ compared to exponential smoothing?

Regression analysis allows us to infer a relationship between two variables. In economic terms, we can use regression analysis to infer the relationship between supply and demand and use that relationship to predict supply based on a certain level of demand. In OSCM, we can infer relationships between unit costs and demand. Testing the relationship between two variables can greatly help decision-making and strategic planning.
As part of regression analysis, understanding correlation and covariance are important. Correlation tells us how closely the variables are related to each other. Covariance tells us how variables vary together. Regression analysis, correlation, and covariance allow us to infer the relationship between variables. Because regression analysis is neither precise nor perfect, we can only make inferences based on the results of the analysis.
We can also use the moving average to forecast demand. Moving average is only used when the time series of data points is constant; that is, seasonality cannot be accounted for in this model. All historical data points have the same "weight", so changes in demand might show more slowly in this model. Exponential smoothing applies exponentially decreasing weights to older data to "smooth out" the effects of trend-like or seasonal-like data.
To review, see:


5c. Identify possible trends from a scatter plot that indicate positive, negative, linear, or non-linear relationships in data 

  • What is a scatter plot?
  • How does a scatter plot indicate a positive association?
  • How does a scatter plot indicate a negative association?
  • How can a scatter plot show a linear versus a non-linear relationship?

We can show the association between two different variables using a scatter plot. If the data we plot on the scatter plot shows an increase in both the y variable and the x variable (as variable y increases, variable x also increases), we have a positive association between these two variables. If variable y decreases and variable x increases, we have a negative association. We can also determine that the two variables are not associated with each other.
A linear relationship means that we can draw, or calculate, an almost straight line through the scatterplot data. A non-linear relationship means that we cannot draw a straight line – the line may be curved.
We can calculate how closely related the two variables are; this calculation provides an R value between -1 and 1. An R value of 1 shows a strong positive association; an R value of -1 shows a strong negative association; an R value of 0 indicates the two variables are not associated.
Can you see how comparing variables in a production process might be helpful? You could calculate the association between the number of workers on the assembly line and the quality of the finished product. In a retail store, you could calculate the association between the number of salespeople on the sales floor and the sales value.

5d. Interpret forecasting errors and use them to quantify the uncertainty in a forecast 

  • What are forecast errors, and how do these errors influence an organization?
  • What are the common forecast errors we see, and how are they calculated?

Forecasting errors happen all the time; after all, we are not working with perfect data. However, the size of the forecasting error is important. A forecasted/actual error of $50 might not be a problem, but what about a forecasted/actual error of $5,000? Large errors in forecasting can lead to shortages or overstock of products, less-than-ideal cash flow, and other costly errors.
We can identify several types of forecasting errors: Bias, Mean Absolute Deviation (MAD), Mean Absolute Percentage Error (MAPE), and Mean Squared Error (MSE). Each of these types of errors can be easily calculated.
Types of forecast errors include bias. This type of error can be a positive bias or a negative bias error. What are the consequences to a firm if they suffer from long-term positive or negative bias? Mean Absolute Deviation (MAD) is concerned with the absolute value of the error, not with whether the error is calculated to be positive or negative. The Mean Absolute Percentage Error (MAPE), like the MAD, does not give directional movement in the errors – it simply provides a percentage value rather than a number value. The Mean Squared Error (MSE) penalizes the larger error values over the small error values, allowing us more information with which to adjust our forecasting model.
To review, see Forecasting Errors.

Unit 5 Vocabulary 

This vocabulary list includes the terms that you will need to know to successfully complete the final exam.

  • bias
  • correlation
  • covariance
  • exponential smoothing
  • forecasting
  • forecasting error
  • Mean Absolute Deviation (MAD)
  • Mean Absolute Percentage Error (MAPE)
  • Mean Squared Error (MSE)
  • moving average
  • negative bias
  • planning
  • positive bias
  • qualitative forecasting
  • quantitative forecasting
  • regression analysis
  • weighted moving average