Regardless of the claims of many MRPII and ERP vendors that their systems can optimize your production and inventory; regardless of how glitzy their sales' brochures are; regardless how sleek their salespersons' talk about their systems' capabilities  doubt them! I have yet to find one system that can truly optimize the inventory replenishment operation. All the systems currently on the market are fairly simplistic to the point of being naive. Lot size and economic production quantity, lead-time, demand patterns and order distributions are considered in gross by the system. As a result, MRPII/ERP systems do not optimize inventory operations.

At Sensient Technologies Corporation, we have implemented MRPII and ERP types of systems at various production sites to help manage our operations. This, however, does not translate automatically to better customer services and inventory situations. Sensient, formally known as Universal Foods Corporation, is a global leader in the manufacture and supply of high-performance flavor, fragrance and color products to enhance a variety of consumer goods, from ice cream and breakfast cereal to lipstick and ink jet printers. The company operates out of more than 20 countries with more than 40 plants and sells to customers in every corner of the world. As you can imagine, it is a challenge to manage the complexities of the supply chain and, therefore, the inventories that meet the need of all our customers. This is where OR/MS methodologies can help to add "intelligence" to the current MRPII/ERP systems to best manage inventory replenishment operations.

Common Problems with MRPII/ERP Systems


As mentioned earlier, all MRPII/ERP systems currently on the market are fairly "simplistic." Although these systems have provided a platform for the integration of plant resources and priorities, they have serious inherent weaknesses.

For example:

Arbitrarily fixed planning and stocking parameters. Planning parameters, such as lot sizes, lead times and safety stocks, are critical to the management of inventory. There are two basic problems with MRPII/ERP systems. First, these important parameters are arbitrarily set and are rarely validated. One simple reason is that there is no simple methodology that can readily be used to validate or "optimize" these parameters. In addition, each product or product group has different demand and replenishment characteristics. It is difficult to "generalize" and to use canned models/policies. Secondly, even if these important parameters were once "validated," the dynamic nature of the business dictates the need to regularly monitor the demands and, thus, constantly make changes to the parameters. This is rarely done in practice. As a result, MRPII/ERP systems perform poorly in managing inventories and their replenishments.

In addition, there is a general lack of decision analysis and evaluation capability with MRPII/ERP systems. This inherent limitation further handicaps the proper setting of important planning parameters. While major providers claim to have additional add-in "optimization" modules or packages, they are inadequate.

"Infinite" capacity. Since MRPII/ERP systems plan product items on an individual basis, they naively assume that production capacity is infinite and that there is no item-to-item competition for such capacity when the systems translate demand into planned orders. Obviously, this assumption is flawed considering the different bottlenecks that exist in real production situations. As a direct result of assuming infinite capacity, MRPII/ERP systems do not have product-mix optimization capability that would provide intelligent choices of which products to produce in order to maximize throughput. This is truly a major drawback.

Consequently, MRPII/ERP systems often produce unrealistic production scenarios that result in excessive inventories, sub-optimal utilization of resources and ultimately poor customer service.

To remedy the weaknesses of MRPII/ERP systems, we can either wait for the continual evolution of these systems or build customized add-ins or stand-alone programs to address specific optimization problems.

Even though all major players claim that they are improving their optimization capabilities, the pace of this improvement is fairly slow. Complicating the issue is that each production situation is very unique. It is very difficult  frankly, almost impossible  to have a generic and flexible system that will cater to all characteristics of diversified production situations of all industries.

A far better choice to overcome the weaknesses of current MRPII/ERP systems is to build add-ins or stand-alone programs that will help the systems optimize a specific operation. Following is a description of one particular effort at Sensient Technologies to prepare a stand-alone spreadsheet simulation module that adds "intelligence" to our MRP/ERP systems. The focal point of this discussion is on the inventory replenishment operations.

Spreadsheet Simulation Program


In our search of models that would "optimize" our inventory replenishment operations, we evaluated Economic Order Quantity (EOQ) types of models as well as some probabilistic models. None of these are useful for our situation. For example, in a typical EOQ model, it is assumed that there is a constant, known demand rate. I have yet to see any real situations that satisfy this condition. In addition, the model assumes that there is a constant, known lead-time for delivery. This assumption is equally unrealistic. Product items are competing for the same resources during production. Lead-time varies greatly. Probabilistic models (e.g., the newsvendor model) have stringent assumptions that make the model inapplicable to practical situations. Since all these canned models from textbooks were not suitable, we decided to obtain planning and stocking parameter values from simulation.

The beauty of simulation is that we can mimic real situations. Using specific planning and stocking parameter values, simulation allows us to easily examine inventory operation performances, e.g. stock-outs and frequencies, inventory amounts and carryovers, etc. The results from simulation provide a clear picture as to a best choice of planning and stocking parameters. In our simulation program, we have the following inputs for each product.

1. Product demand pattern. From order history, we construct a discrete histogram that represents the demand distribution. In addition, we calculate the order inter-arrival time. These two pieces of information uniquely defined the product demand pattern. Each product or product group has a unique demand pattern. Since historical information may not be repeated in the future, we can easily change the histogram and inter-arrival time to accommodate any anticipated changes of demand in the marketplace.

2. Production/order lead-time. Lead-time is unique to the product item as well as to the work center that makes it. Our program allows us to simulate inventory situations using different lead-times. This enables us to examine lead-time's impact to the inventory performances and to the production capacity.

3. Safety stock/reorder point. This is an important parameter to decide. If we set the reorder point level high, the inventory level will be high and the amount of inventories to carry will be high, too. This also corresponds to the production resources that are tied up for this particular product. Since the work center has only limited capacity, building a high inventory for one product means "robbing" scarce resources that may be needed more urgently by other products. This is an important balance act  between building up inventory of one particular item vs. building up all products  to meet the needs of all customers.

4. Reorder/production quantity. The trade-off for this parameter is as follows. If we set this quantity high, e.g., huge lot size, the need to reorder/produce will be reduced. This is because a large order will suffice for a long period of time until its depletion. On the other hand, if the reorder quantity is small, this will dictate more frequent reorders. A higher reorder frequency will translate to higher set-up costs, smaller production lots, more movements, paper work, etc. This again calls for a delicate balance between a large order vs. a small order. Each carries benefits and disadvantages.

To run our program, we need first to quantify all the four items above for each scenario. After running a few scenarios, the best replenishment policy will quickly emerge. To alleviate any concerns of the robustness of the results due to the random variations inherent in simulation, each scenario is run 50 times representing 50 years, each year using 300 working days. In our spreadsheet program, each run for a fixed scenario for 50 repetitions/years takes only a few seconds. Running the program for all major products took only a few hours.

The program is very versatile, because demand patterns and planning and stocking parameters (items 1-4 mentioned earlier) can easily be changed in simulation. Given the capacity of the work center, the best replenishment policy for each product/product group can thus be obtained. This program takes guesswork out of the determination of the best inventory policy. It analyzes objectively the various replenishment scenarios and provides the user with the best choice. To schedule the production of planned orders derived from the above methodology, we map out the production sequence using a specialized scheduling program with critical chain capability. We apply the Theory of Constraints (TOC) methodology to our production scheduling operation. The practice goes hand-in-hand with the replenishment operation discussed above.

This experience clearly reveals to me that: 1. OR/MS professionals can contribute greatly to enhance the inherent shortcomings of current MRP/ERP systems, and 2. to truly optimize an operation, great amount of efforts are needed to revamp the MRPII/ERP systems where OR/MS professional must play a central role. Until that day, we should examine carefully the validity of all current MRPII/ERP outputs.