Supply Chain Analytics
The classic supply chain approach has been to try to forecast future inventory demand as accurately as possible, by applying statistical trending and "best fit" techniques based on historic demand and predicted future events. The advantage of this approach is that it can be applied to data aggregated at a fairly high level (e.g. category of merchandise, weekly, by group of customers), requiring modest database sizes and small amounts of manipulation. Unpredictability in demand is then managed by setting safety stock levels, so that for example a distributor might hold two weeks of supply of an article with steady demand but twice that amount for an article where the demand is more erratic. Then, using this forecast demand, a supply chain manufacturing and distribution plan is created to manufacture and distribute products to meet this forecast demand at lowest cost (or highest profitability). This plan typically addresses the following business concerns: - How much of each product should be manufactured each day? - How much of each product should be made at each manufacturing plant? - Which manufacturing plants should re-stock which warehouses with which products? - What transportation modes should be used for warehouse replenishment and customer deliveries? The technical ability to record and manipulate larger databases more quickly has now enabled a new breed of supply chain optimization solutions to emerge, which are capable of forecasting at a much more granular level (for example, per article per customer per day). Some vendors are applying "best fit" models to this data, to which safety stock rules are applied, while other vendors have started to apply stochastic techniques to the optimization problem. They calculate the most desirable inventory level per article for each individual store for their retail customers, trading off cost of inventory against expectation of sale. The resulting optimized inventory level is known as a model stock. Meeting the model stock level is also an area requiring optimization. Because the movement of product to meet the model stock, called the stock transfer, needs to be in economic shipping units such as complete unit loads or a full truckload, there are a series of decisions that must be made. Many existing distribution requirements planning systems round the quantity up to the nearest full shipping unit. The creation of for example, truckloads as economic shipment units requires optimization systems to ensure that axle constraints and space constraints are met while loading can be achieved in a damage-free way. This is generally achieved by continuing to add time-phased requirements until the loads meet some minimum weight or cube. More sophisticated optimization algorithms take into account stackability constraints, load and unloading rules, palletizing logic, warehouse efficiency and load stability with an objective to reduce transportation spend (minimize 'shipping air'). Optimization solutions are typically part of, or linked to, the company's replenishment systems distribution requirements planning, so that orders can be automatically generated to maintain the model stock profile. The algorithms used are similar to those used in making financial investment decisions; the analogy is quite precise, as inventory can be considered to be an investment in prospective return on sales. Supply chain optimization may include refinements at various stages of the product lifecycle, so that new, ongoing and obsolete items are optimized in different ways: and adaptations for different classes of products, for example seasonal merchandise. It should also factor in risks and unexpected constraints that often affect a global supply chain's efficiency, including sudden spikes in fuel costs, material shortages, natural disasters such as hurricanes, and instability of global politics.