For modeling, it is understood that ship-to-customer locations are customarily aggregated into larger customer regions and individual products stock-keeping units (SKUs) are grouped into major product lines or categories. While customer aggregation is usually straightforward, product aggregation requires experience and skill to do properly. With respect to time periods, many models are built at an annual level of detail. However, it is sometimes desirable to construct models with multiple time periods, or "buckets," so that questions of seasonality, inventory pre-build or long-term strategic planning may be explicitly addressed.
Step 3: Obtain Customer Demand Data. Customer demands, measured in units of weight, cube or units (cases, pallets, gallons, liters, etc.) must be obtained for each customer region/channel/finished product/time period. While theoretically this can be done by ad hoc methods, it is generally much more efficient and accurate to obtain the data from an actual business system such as an invoice or shipment history file. These historical demands can then be scaled by means of suitable forecast factors to drive out-year analyses.
Step 4: Obtain Freight Costs. Many options are available here, including use of the firm's own rates, distance-based equations and sophisticated simulations of proposed traffic management policies. Commercially available databases are available for less-than-truckload (LTL), truckload (TL) and parcel rates for most North American locations. When properly used, these databases are a rich source of unbiased data and can greatly facilitate the model-building process. Unfortunately, for the rest of the world machine-readable tariffs are generally not available; the study team must rely on internal resources and/or ad hoc rate solicitations.
Step 5: Obtain Facility Data. Facility data consist of procurement, manufacturing, distribution center, cross-dock, port, etc., costs and capacities, as well as conversion recipes (bills of material) at manufacturing locations. Facility data are technically optional. However, from a managerial perspective they are essential to a well-developed supply chain model and are often the principal drivers for the recommended supply chain design. Sources include historical cost accounts, engineering standards, extant contracts and commercial databases, the latter useful for establishing and understanding inherent regional differences. As with any effort that involves accounting data, challenges here include fixed vs. variable classifications, allocations by product and understanding the reasons for reporting differences across facilities (region, mission, managerial and/or workforce competence and so on). Facility data are typically a small percentage of the database but are often the most difficult to prepare and can generate the most controversy.
Step 6: Validate the Model. At this point, the model structure and database are essentially complete. But before succumbing to the temptation to turn the optimization engine loose, it is essential that the database be validated by means of an historical cost/flow baseline. The traditional methodology consists of "locking down" all facility and transportation link volumes per historical values, multiplying these flows by the cost coefficients developed for the model (thereby generating an estimate of total historical costs) and comparing the results with accounting data. The objective, of course, is to reconcile (or at least explain) differences to the satisfaction of both study team and management, thereby building credibility for the analyses to follow. At first blush, model validation would appear to be a daunting task; however, it can be greatly simplified if the database has been properly constructed and the software package includes well-designed baseline preparation features that leverage historical business system-derived data.