Northcutt Bikes: The Forecasting Problem Chapter 8 Case Study Teaching Notes Introduction Northcutt Bikes is a relatively basic forecasting situation. While the major thrust of the case is a quantitative analysis of the demand data in the case, it also provides an opportunity to discuss with students the issues of applying forecasting solutions in a growing company, and in particular how the quantitative analysis of the forecasting data can be use in conjunction with the qualitative issues of the actual company situation. As with other cases involving small but growing companies, Northcutt is faced with the possible need to formalize their approaches to business if they are to continue with successful growth. Quantitative Analysis The first thing students should do in the case is plot the data to obtain a \"picture\" of the situation, as is indicated by the first case analysis question. The basic plot is attached to this note. Students should readily see that the data shows clear seasonality (as should be expected from the type of product), and also shows a clear upward overall growth trend. This is important for them to note, as any effective forecasting approach should include capturing both trend and seasonality components. In the second question, some students may use as one of their methods a \"collapse\" of the data into quarterly rather than monthly data points (as suggested by case question 4). If so, you can use this opportunity to discuss that while such an approach may provide \"better\" forecasts (in that the average MAD or MSE may be smaller), the method will also result in a loss of detail. You can use this opportunity to discuss the trade-offs involved and the type of conditions where such a collapse of data may or may not be appropriate. Included with this note are two examples of approaches used. The first is a simulation using Winter's Model, which is a smoothing model incorporating both trend and seasonality smoothing constants. The simulation was set up to search for the best smoothing constants, starting at 0.01. The criterion for selection was the smallest MAD. In this case the simulation selected an alpha of 0.23093, a beta of 0.01036, and a gamma of 0.41165. The resulting MAD was 194.84. Using this model, the forecast for the next four months (rounded) was: 755 1101 1041 1217 Students may want to know why only four months was asked for. If so, this question is a good opportunity to discuss with them the assumptions of time-series models and how the accuracy of such models falls off when used for long-term forecasts. The second model used was simple regression. Regression captures trend very well, but seasonality must be dealt with separately. At least three approaches can be used here. One is to use multiple regression with dummy variables. The second is to compute seasonality multipliers, apply them to the data (to \"de-seasonalize\" it), then reapply the multipliers after the forecast is made. The third approach (used here) is to regress each month separately. The forecasts using this method were: 789 1480 1207 1250 It is noteworthy that the February forecast is so much larger using the regression. Looking at the raw data shows that the demand in February is growing with a larger trend (in fact the regression shows a trend growth of 231 units per year for February. In contrast, the trend for January is 69.3, March is 130.3, and April is 81.1. Students should be expected to be able to discuss the differences in the models, how those differences come about, and what should they do about it if this was their company. They should also be capable of discussing (and not just showing) methodologies to determine which method was \"best.\" Typical approaches use computation of MAD or MSE. Qualitative Issues The last three case questions address some of the qualitative issues in the case. The first of these addresses the inclusion of the qualitative knowledge into forecasts. While the time series analysis of the data help provide patterns of demand, there are still many issues that cannot be captured. They include, of course, issues of competitive moves, economic factors, advertising and promotional activities, and so forth. Clearly Jan's knowledge of these qualitative issues can and should be used in combination with the quantitative analysis to provide the best overall forecast for the business. The last two questions address the issue of the size and the general management approaches that the business size suggests. Some might suggest the business is too large to continue running it as it has been run in the past. That may in fact be true, but the best solution is seldom to shrink. Instead she should look to formalize approaches to manage the growth and size. She needs to not only incorporate the forecast methodology into the business, but also needs to develop approaches to capacity and inventories that will best serve both her customers and the needs of her business. The alternatives can be discussed in general at this point in the case, but can then be addressed more explicitly in the second case in Northcutt Bikes, dealing with aggregate production planning. \f\fRunning Head: 1 My personal tutor Name: Course: Tutor: Date: Hello, I have seen how the tutors are disappointing you and would want to be helping you directly in all your academic assignments, projects, proposals and thesis. Kindly send the work to me directly to tunyahelp@gmail.com and get quality work. Regards tunyahelp@gmail.com Prof. Antonio Klin