(Warwick Woodard, Diego Olivarez, Shane Needham, Britain Auer, 1998)
Elk Corporation of Dallas manufactures premium asphalt roofing shingles. To this point the company has used historical demand data to generate ``ballpark" figures for how much of each type of product to produce in any given period. The objective of this project is to produce a monthly production schedule for Elk in order to maintain their inventory at such a level as to ensure a 95% fulfillment rate of orders within 24 hours of their placement, while minimizing inventory costs. The reason Elk needs such a high level of order fulfillment on such a short timetable is simple: During high demand periods, their customers will place orders for shingles with several shingle suppliers. Their orders are often highly redundant, and whichever supplier can fulfill the customers' needs first will get the order. The other orders are cancelled shortly thereafter, accompanied by the obvious loss of revenue for the suppliers that are slow to act. It is the policy of Elk not to ship partial orders, so if an order cannot be completely filled within 24 hours, it is considered canceled.
We chose to create a linear programming model using Generalized Algebraic Modeling Software optimization package. This choice was an easy one, as Dr. Barr had already begun this project as a consulting job, using this approach himself. We obtained a copy of the original skeletal model and went from there. This model only took into account three very broad categories of Elk's products. We expanded this model to produce a schedule that would be detailed to well over one hundred different product types. Elk was cooperative in granting us access to ample data concerning previous years' sales. This data was incorporated into our GAMS model and a viable schedule was created.
The managers at Elk could not be expected to take our production plan and just use it on faith that it would work. Therefore we decided that a simulation would be a powerful way to demonstrate the effectiveness of the production schedule. The simulation, written in Visual Basic, takes demand data from the past several years and introduces variability that follows a normal distribution. It generates orders and the details of those orders around a mean and then checks the availability of the products requested. The simulation ends with a summary of orders cancelled and gives a percent fulfilled. If 95% of the orders are fulfilled, then our production schedule is a success.